United States EPA-600/2-84'096
Environmental Protection
Agency May 1984
&EPA Research and
Development
IN SITU FIELD PORTABLE
FINE PARTICLE
MEASURING DEVICE
Prepared for
Office of Environmental Engineering and Technology
Prepared by
Industrial Environmental Research
Laboratory
Research Triangle Park NC 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards.
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents necessarily
reflect the views and policy of the Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-84-096
May 1984
IN SITU FIELD PORTABLE FINE PARTICLE
MEASURING DEVICE
by
Robert G. Knollenberg
Particle Measuring Systems Inc.
1855 South 57th Court
Boulder, Colorado 80301
EPA Contract No. 68-02-2668
EPA Project Officer: D.Bruce Harris
Industrial Environmental Research Laboratory
Research Triangle Park, N.C. 27711
Prepared for:
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
ABSTRACT
During the past two years an in situ fine particle measuring device has
been developed by Particle Measuring Systems for the Environmental Protection
Agency. The resulting instrument is designated a Fine Particle Stack Spectro-
meter System. The instrument utilizes a laser fed optical system with detection
by near-forward light scattering. Sample volume is established through the use
of a high resolution optical system viewing particle images in a dark field
through a masked beam splitter. The instrument covers an 0.5 to 11.0 ym size
range with 60 channels resolution. Absolute theoretical accuracy is ±20% of
size for completely unknown refractive index. The instrument is designed to
operate continuously at in-stack temperatures up to 250°C at flow velocities
up to 30 m/sec. Flow velocities are determined from measured particle transit
times through the laser beam. Internal probe components are cooled by a water
jacket and external heat exchanger. The heat exchanger, probe pulse processing
electronics, data acquisition system and operating controls are housed in a
central electronics console. The data acquisition system utilizes a micro-
processor, EROM firmware, and a matrix printer/plotter to provide a variety
of processed data reports. It has been laboratory characterized and field tested
on coal-fired power plants at both the inlets and outlets of control devices.
Its performance indicates good agreement with impactors and excellent agreement
with opacity meters in computed mass loading and optical opacity. Its size
resolution is greater than other current known techniques. Its eventual use
will be directed at the characterization of particulate emissions of stacks
or other stationary sources and to qualitatively evaluate the performance and
collection efficiences of particulate control devices in operation.
ii
-------
CONTENTS
Abstract ii
Figures iv
Tables viii
1. Introduction 1
2. Discussion of the Stack Environment and FPSSS
Design Criteria 5
3. Overview of the FPSSS 10
3.1 FPSSS Optical System 15
3.2 FPSSS Head 15
3.3 FPSSS Support Beam and Port Bearing 19
3.4 Thermal Control System 20
3.5 FPSSS Electronics Console 20
3.6 Operation and Use 22
4. Examination of the Optical Properties of Particles
and their Implication for Stack Measurements .... 29
5. Results of Aerodynamic Tests and Thermal Modeling ... 54
6. Expanded Description of FPSSS Systems 82
6.1 The FPSSS Electro-Optical System 82
6.2 Electronic Subsystems 88
6.3 Basic Data Processing Equations 97
7. Calibration and Laboratory Test and Simulation
of Fly Ash Environments 102
7.1 Calibration and Evaluation of the FPSSS at
PMS 102
7.2 IERL Test Summary Ill
8. Field Testing 148
8.1 Initial Testing at Valmont Power Plant 148
8.2 Field Tests in Charlotte, North Carolina .... 158
8.3 Final Tests at Valmont 166
9. Conclusions 204
References 209
Appendix I 210
Addendum 216
iii
-------
FIGURES
Number
1 FPSSS block diagram ••••......... 12
2 Photograph of FPSSS components . 13
3 Photograph of FPSSS electronics console ......... 14
4 Optical system diagram . 16
5 Assembly diagram of the FPSSS probe head 17
6 FPSSS installation at IERL 18
7 Close-up of FPSSS head and split lateral bearing .... 18
8 FPR FPSSS heat exchanger assembly 21
9 FPSSS numerical testing output samples 25
10 FPSSS size distribution graphics samples 26
11 FPSSS mass distribution samples 27
12 Sample of FPSSS time series plot 28
13 Wavegroups contributing to light scattering 31
14 Collecting angle required to collect 89% of diffracted
light 32
15 Collecting angle required to collect 80% of diffracted
light 34
16a Average intensity vs. scattering angle for D = 0.6 ym . . 35
16b Average intensity vs. scattering angle for D = 2 ym . . . 36
16c Average intensity vs. scattering angle for D = 5 ym ... 37
17 MIE response of the Royco 218 particle counter for
particles with refractive indices 1.33, 1.70,
1.90 and 2.20 39
18 Response of the Royco 245 particle counter 40
19 MIE response for PMS CSAS-100 spectrometer model .... 41
20 MIE response of the Climet particle counter for
particles with refractive indices 1.33, 1.45,
1.70 and 1.90 42
21 MIE response for PMS LAS-200 spectrometer 43
22 MIE response of the Royco 218 particle counter for
particles with refractive indices 1.54, 1.54 - 0.051,
and 1.54 - 1.01 45
23 MIE response of the Climet particle counter for
particles with refractive indices 1.54, 1.54 - 0.051,
1.54 - 0.51 and 1.54 - 1.01 46
24 Sensitivity of 4° - 22° collecting angles to real
index dispersion 43
25 Sensitivity of 4° - 22° collecting angles to complex
index dispersion 49
26 Sensitivity of 2° - 11° collecting angles to real
index dispersion 50
27 Sensitivity of 2° - 11° collecting angles to complex
index dispersion 51
28 Sensitivity of 2° - 11° collecting angles to 1°
errors ............... 53
29 Conceptual FPSSS probe optical-mechanical packaging ... 55
IV
-------
Figures (Cont.)
Number Page
30 Test with 4 inch annular ring probe mock-up 56
31 Test with 6 inch annular ring probe mock-up 57
32 FPSSS probe head configurations during wind tunnel
tests 59
33 Test with breadboard FPSSS probe head scan along
optical axis from window to mirror 60
34 Test with breadboard FPSSS probe head transverse
scan across sample plane 61
35 Test with breadboard FPSSS probe head transverse
scan across sample plane behind standoffs 62
36 Head cross section for thermal model 65
37a FPSSS sensor head nodal layout 70
37b Nodal diagram of FPSSS coolant loop 71
38 Brassboard 80
39 Relative size and position of sample volume cross
section 86
40 Image sizes and positions on masked aperture detector . . 87
41 Image size as a function of sample volume position
for the case of 4X gain ratio of masked-to-signal
aperture 89
42 FPSSS electronics console block diagram 90
43a FPSSS probe electronics block diagram 92
43b FPSSS data acquisition system block diagram 95
44 Relative scattering response 104
45 FPSSS sample volume map 105
46 FPSSS interrange comparison 107
47 FPSSS/CSAS comparison (submicron particles) 108
48 FPSSS/CSAS comparison (medium sized particles) 109
49 FPSSS/CSAS comparison (large particles) 110
50 Schematic of IERL wind tunnel showing dust handling
equipment 112
51 IERL screw feeder and IERL fluidized bed aerosol
generator 114
52 SEM photomicrographs display the primarily spherical
Duke Power Plant fly ash (upper) and the clumped
aggregates of iron oxide (lower) 116
53 FPSSS/Coulter Duke power fly ash size distribution
comparisons 117
54 FPSSS/Coulter counter iron oxide size distribution
comparisons 118
55 FPSSS accumulative mass distribution for the iron
oxide for run number 14 122
56 Size distribution for IERL run number 1 125
57a Mass distribution for IERL run number 1 126
57b Accumulative mass distribution for IERL run number 1 . . 127
58 FPSSS size distribution for IERL run number 5 128
59a FPSSS mass distribution for IERL run number 5 129
59b FPSSS accumulative mass distribution for IERL run
number 5 130
v
-------
Figures (Cont.)
Number Page
60 FPSSS accumulative mass distribution for IERL run
number 7 131
61 IERL impactor/FPSSS mass distributions comparison . . . 133
62 Impactor/FPSSS mass distribution comparison for iron
oxide 134
63 FPSSS time series and number density plot in IERL
tunnel ..... 135
64 Time series for rotation of probe head in IERL tunnel . 136
65 Effects of rotation in windstream on mass loading . . . 138
66 Spatial Variations of number density and mass loading
in IERL tunnel 139
67 FPSSS size distribution for IERL heated tunnel using
Duke fly ash, for run ending at 11:05 on February 8 . 141
68 FPSSS mass distribution for IERL heated tunnel using
Duke fly ash, for run ending at 11:05 on February 8 . 142
69 FPSSS time series plot of number density and mass
loading for 6 minute run ending at 43:00 on
February 8 143
70 FPSSS size distribution for IERL test in fuel oil
boiler flue, ending at 13:00 on February 14 145
71 FPSSS mass distribution for IERL test in fuel oil
boiler flue, ending at 13:00 on February 14 146
72 Schematic drawing of a steam electric generating
station 149
73 Average in-stack size distribution 152
74 Average in-stack extinction cross section distribution . 153
75 Lear Siegler RM41 opacity record for April 24, 1978
between 14:00 and 15:00, during which the FPSSS
performed measurements in the stack 154
76a Average in-stack volume (mass) distribution 155
76b Average in-stack volume (mass) distribution from 0.4
to 0.85 ym 157
77 Valmont control device removal efficiency . . 159
78 Charlotte total mass comparisons 160
79 FPSSS number density for sample run ending at 14:55
on April 17 162
80 FPSSS mass loading for sample run ending at 14:55 on
April 17 163
81 Charlotte outlet mass distribution 164
82 Charlotte cumulative mass distribution 165
83 Valmont outlet velocity profiles . 168
84 Average FPSSS size distribution observed using ranges
3 and 4 during first impactor run (Valmont outlet) . . 171
85 Average FPSSS mass distribution observed using ranges
3 and 4 during first impactor run (Valmont outlet) . . 172
86 FPSSS size distribution for run ending at 14:11 on
September 13 (Valmont outlet) . 173
87 FPSSS mass distribution for run ending at 14:11 on
September 13 (Valmont outlet) 174
VI
-------
Figures (Cont.)
Number Page
88 FPSSS size distribution covering third impactor
run, ending at 10:14 on September 14 (Valmont
outlet) 176
89 FPSSS mass distribution covering third impactor
run, ending at 10:14 on September 14 (Valmont
outlet) 177
90 FPSSS size distribution for time period ending at
10:23 on September 14 (Valmont outlet) 178
91 FPSSS mass distribution for time period ending at
10:23 on September 14 (Valmont outlet) 179
92 FPSSS size distribution at 14:00 on September 14
(Valmont outlet) 181
93 FPSSS mass distribution at 14:00 on September 14
(Valmont outlet) 182
94 FPSSS size distribution for run ending at 16:49 on
September 14 (Valmont inlet) 183
95 FPSSS mass distribution for run ending at 16:49 on
September 14 (Valmont inlet) 184
96 FPSSS size distribution for run ending at 10:13:24
on September 15 (Valmont inlet) 186
97 FPSSS mass distribution for run ending at 10:13:24
on September 15 (Valmont inlet) 187
98 Valmont rapping sequence as observed in opacity and
mass median diameter 188
99a Rapping sequence for September 15th at approximate
mean mass size 190
99b Rapping sequence for September 15th at submicron
sizes (CSAS-HTS) 191
100 Lear Siegler RM41 opacity record for September 14 ... 192
101 Valmont - % opacity scattergram 193
102 Valmont rapping sequence as observed in mass loading
and number density 194
103 Valmont inlet mass distribution 196
104 Valmont inlet impactor run number 6 cumulative mass
loading 197
105 Valmont outlet mass distribution 198
106 Valmont outlet cumulative mass loading FPSSS/impactor
comparison 199
107 SEM photomicrographs at equal magnifications display
Valmont impactor samples 201
Al Photograph of PMS calibration wind tunnel facility . . 212
A2 Diagram of PMS calibration wind tunnel facility .... 213
A3 Wind tunnel monodispersed aerosol source subsystem . . 214
A4 Sample inlet tube installed in tunnel test section
with FSSP's 215
vii
-------
TABLES
Number Page
I FPSSS Component Power Dissipations ............ 64
II FPSSS Nodal Descriptions ................. 72
III FPSSS Radiation Conductances ............... 74
IV FPSSS Temperatures for Balanced 1 GPM Flow ........ 75
IV FPSSS Temperatures for Balanced 2 GPM Flow 76
IV FPSSS Temperatures for Unbalanced 1 and 2 GPM Flow .... 77
V FPSSS Thermal Analysis Results .... 78
VI IERL Wind Tunnel Mass Loading Comparisons 120
VII Mass Loading Increases 144
VIII Summary of Valmont FPSSS, Impactor and Opacity Data . . . 169
vxn
-------
1. INTRODUCTION
In September, 1977, Particle Measuring Systems, Inc. (hereafter PMS) was
contracted by the Environmental Protection Agency (hereafter EPA) to design,
fabricate, and evaluate an i,n situ Fine Particle Stack Spectrometer System
(FPSSS). The development of the FPSSS was intended for application to charac-
terize the particulate emissions of stacks and other stationary sources and to
qualitatively evaluate the performance and collection efficiences of particulate
control devices now in use or under development. The primary EPA design cri-
teria that the FPSSS had to satisfy were to provide in situ high-resolution
particle sizing over a size range from 0.5 to 5.0 jam diameter, with number
4 -3
densities in this range above 10 cm and integrated particulate loading from
0.3 to 3.0 g m . The FPSSS was also required to operate at temperatures from
20 to 250°C and at flow velocities from 1 to 30 m sec . An operational pro-
totype instrument has been designed, fabricated, and performance-tested. This
report describes the design, development, and testing of this new high-resolu-
tion in-stack particle size spectrometer system.
The basic design approach utilized existing technology. Three major areas
of previous work provided the underpinnings of the FPSSS design. These were:
a) The well-developed PMS technique of utilizing high-resolution imaging
systems with light scattering to perform in situ single particle
size measurements.
b) The experience by PMS in stack measurements using plumbed instruments.
c) The Pioneer Venus Large Probe Cloud Particle Size Spectrometer (LCPS)
development work, involving cooperative efforts between PMS and the
Ball Brothers Research Corporation (BBRC).
Probably the greatest desire of the research performing particle size
measurements is to do so without disturbing the particles and thereby making
the measurements in situ. In some respects the concept of in situ sampling is
an unattainable goal; what one desires is to minimize sample perturbation. As
far as single particle sizing devices are concerned, this ability to size
particles in situ generally requires additional imaging technology to dimension
sample volume. In PMS in situ light scattering spectrometers, sizing is
through pulse height analysis of scattered light and the sample cross section
-------
is defined through imaging. The sample volume is defined as the product of the
sample cross section and the particle velocity. The particle velocity can be
determined through independent airspeed measurement or directly from image tra-
jectory measurements. Particle number densities in stacks are extremely high,
particularly at submicron sizes. Such high number densities require the use
of reasonably high magnification imaging systems to localize a much smaller
region than is common practice with commercial aerosol counters employing similar
light scattering approaches.
Prior to the current efforts, PMS began to apply existing light scattering
instrument technology to stack measurements (Knollenberg, 1977). Our efforts
with in-stack measurements largely dealt with the development of an extractive
stack sampling duct and a high sample test section which allowed existing PMS
instruments to view the hot particle flow through protective windows and make
quasi in situ measurements in the hot flow. Because of similarities in the
aspiration of PMS spectrometers this provided a method of direct application of
several instruments covering different particle size ranges. It also allowed
for the probing of other hot particle sources requiring sample removal—partic-
ularly chemical process ovens.
Another area of work that provided a developed technology base was the
Pioneer Venus Large Probe Cloud Particle Size Spectrometer (LCPS) involving
both PMS and BBRC. Results from this successful experiment are given in
Knollenberg and Hunten, 1979a, 1979b, and 1980, and a full instrument description
is given in Knollenberg and Gilland, 1980. From the beginning, we realized the
design of an instrument to measure the particle size and number density in the
Venusian atmosphere and continuous cloud canopy would face difficult environ-
mental requirements. Operating temperatures ranged from -50°C to 450°C, pressures
up to 100 atmospheres, sample velocities to 50 m sec , and the reentry G-load
was on the order of 400 G's. The instrument had to cover a size range of from
0.5 to 5.0 Um by in situ optical single particle methods. Such requirements are
similar to, although more rigid than, those being sought by the EPA for an in-
stack instrument. The FPSSS was intended for operations to only 250°C. The in-
stack pressures are also much closer to ambient. The size range of 0.5 to 5 ]_im
-------
was equivalent to the scattering subrange of the LCPS, which was one of four
separate size ranges. The LCPS maximum sample velocities were nearly twice as
high as those expected in-stack. The important point is that the LCPS instru-
ment had to stay in alignment through the mission profile (two hour descent)
and remain relatively contamination-free. These problems were also paramount
in the FPSSS design. The Venusian environment has another similarity to that
found in stacks. The clouds of Venus are composed largely of sulfuric acid and
thus the atmosphere is hot and highly corrosive. Heaters had to be installed
on the LCPS windows to prevent condensation. Aerodynamic shields had to be
developed to provide additional protection from wetting.
The FPSSS design draws heavily from these areas of previous work. In
essence, the FPSSS is the result of using existing technology rather than the
result of basic research and development. An early prototype operational device
was constructed and tested within the first 9 months of the contract. The
final prototype FPSSS instrument would appear to meet the original EPA design
objectives. It has a near-forward light scattering optical system with an ex-
panded 60 channel 0.4 to 11.0 \im size range divided into four subranges of 15
size channels each. The instrument is capable of relatively accurate measure-
4 -3
ments at number densities up to 5 x 10 cm , without significant sensitivity
to refractive index. An optical velocimeter has been designed and incorporated
in the FPSSS. The instrument can operate continuously at 250°C temperatures
utilizing a water-cooled head design and external heat exchanger.
Extensive theoretical modeling of thermal as well as optical performance
has been utilized in configuring the FSPSS probe head. Wind tunnel facilities
at PMS have played an extremely important role in measuring aerodynamic impacts
of the FPSSS sampling section. Calibration has included laboratory and wind
tunnel tests (at both PMS and the IERL facilities at the EPA) on particulates
having known or independently verifiable size distributions. Field tests on
three operating coal-fired power plants were conducted during the course of
this work and are discussed in detail.
The next section provides a discussion of the design criteria established
by the EPA for the FPSSS. Section 3 introduces the FPSSS. The fourth section
gives an in-depth description of the tradeoffs imposed by various optical col-
-------
lecting geometries to sizing performance. Section 5 describes the results of
aerodynamic testing and thermal modeling. Section 6 details the prototype
FPSSS system, followed by its initial calibration and laboratory test results
discussed in Section 7. The field test results are discussed in Section 8,
followed by our general conclusions.
-------
2. DISCUSSION OF THE STACK ENVIRONMENT AND FPSSS DESIGN CRITERIA
Optical instruments for sizing aerosol particles have been around for 20
years or more. However, they have difficulty performing measurements in stack
environments. The following is a list of problem areas that can generally be
attributed to stack environments:
1) High particle number densities
2) High temperatures
3) Corrosive environment
4) Contamination problems, generally from condensable vapors
5) Poor accessibility
6) Relatively high and variable particle velocities
7) Irregular shaped particles
8) Mixed particle composition and refractive index
9) Electrical charging
The above listing is, in general, in order of importance and difficulty.
Obviously, one is not worried about effects of irregular shape or refractive
index if his instrument is being consumed by hot sulfuric acid. We will expand
on some of these problems.
Particle number densities in stacks are extremely high, particularly at sub-
micron sizes. Such high number densities require the use of imaging systems to
localize a much smaller region of optical uniformity than is common practice
with commercial aerosol counters, where particles can be plumbed through an
illuminated volume of one mm or so.
The high temperatures encountered in stacks generally restrict one's
ability to actually place an instrument, with its electronics, optical systems,
etc., inside the stack. A partial solution is to remote the passive portions of
the optical system in the stack environment and couple the input and output of
the optical system through fiber optics or other light communication devices to
the electronics. However, a cooling jacket is required if electronics are placed
within the environment, and the risk of damage to an usually expensive piece of
apparatus must be recognized.
The hot corrosive environment requires the use of heated or purged windows
and relatively inert adaptive hardware. Contamination problems always exist in
a stack environment, and it is generally best to recess optics and flood their
-------
surfaces with clean purge gases. Because the stack velocity is quite high and
variable, the instrument design cannot be sensitive to particle velocities. An
accompanying problem is turbulence-induced vibrations.
Stacks and control devices are generally fitted only with small access
ports—typically 3 to 5" I.D. The poor accessibility to the particulate en-
vironment is often compounded by placement at locations requiring traverses
along catwalks, staircases, and exposure to outside environments.
The fact that the particle species within the stack environment are often
of irregular shape, have mixed particle composition, and unknown refractive
index (with considerable absorption) suggests imaging methods might be applied.
They are, in general, much less subject to error than light scattering methods.
Some imaging systems can provide particle morphology information as well. How-
ever, imaging techniques are unsatisfactory below a few microns due to lack of
resolution. Light scattering systems are thus the only easy alternative. In
order for light scattering systems to perform at all well, they must collect a
large portion of diffracted light as compared to refracted light (strong forward
scattering). This generally limits their application to sizes less than 5 ym
diameter in which it is possible to collect nearly all of the diffracted light
while minimizing the amount of refracted light collected. This generally means
collecting the light scattered down to angles of a few degrees.
On stacks using electrostatic precipitator control devices, the particles
exhausted in the stack are highly charged and represent a potential Electromag-
netic Interference (EMI) problem. Static discharge from conductive apparatus
is sufficient in frequency and magnitude to pose a threat to integrated circuits-
particular ly MOS and CMOS devices.
The EPA delineated eight objectives that the FPSSS should be designed to
meet. These are listed below, followed by a discussion of our interpretation,
areas of concern, and other considerations that our design was based upon. The
objectives specified by the EPA were:
1. In situ operation.
2. Particle size range: 0.5 to 5.0 ym.
3. Particulate loading: 0.03 to 3.0 g m
/• Q _O
4. Number density: 10 -10 particles cm .
-------
5. Temperature range: 20°-250°C.
6. Free stream gas velocity: 1-30 m sec
2
7. Flow duct area: 0.1-40 m .
8. Real-time operation: maximum allowed integration time is 3 min.
9. Portability: The device shall be designed such that a two-man
sampling team could easily move and operate the apparatus.
1) We have already touched upon the elusive aspect of truly in situ measure-
ments. In the context used here it is interpreted as measurement in the stack
without extraction and with minimal disturbance such that the size distribution
measured in the presence of the FPSSS is not altered by more than 10% from that
existing at the same point of measurement in its absence. The primary disturb-
ing effect is aerodynamic and the problem is addressed in Section 5.
Because of the environment encountered in stack measurements, one's ability
to pipe particles around in small bore tubing and force particles to interact
with a small uniformly illuminated light region required by more standard single
particle light scattering methods is almost nil. Only in situ methods are
really applicable. The only way we have found that measurements can be truly
in situ using light scattering is to employ a good imaging system in combination
with the light scattering optics or to use multiple-beam or other signature-
producing illumination sources. We thus reemphasize the definite tie between
imaging instruments and light scattering instruments through the necessity for
a well corrected high performing imaging optical system in both cases. In the
light scattering techniques the imaging system is used to define the sample volume,
although sizing is through measurements of the amount of light scattered. In
imaging techniques, sizing as well as sample volume definition is inherently by
imaging.
2) The particle size range of 0.5-5.0 ym is close to standard size ranges
on several PMS particle size spectrometers as well as being the exact range of the
LCPS scattering subrange. We anticipated and found no problems in meeting the
detectivity required and provided an extended range capability of 0.4-11 ym. We
did find, however, that a tradeoff existed between minimum detectable size and
maximum number density. Questions of best optical design for high accuracy are
addressed in Section 4.
-------
3) Examination of the particle mass loading and number densities showed
7 -3
the two to be incompatible. An inquiry to the EPA revealed that the 10 cm to
o __ o
10 cm number densities were meant to apply to all particles down to sizes
as small as 0.01 ym. They suggested much lower number density values for the
restricted 0.5 to 5.0 vim range. From our own work in stack environments and
from worst-case considerations for the mass peaking out at sizes in the 0.5 to
/ ^ S — 3
5.0 ym range (nearer 5 ym than 0.5 ym) a suggested range of 10 cm" to 10 cm
is a more appropriate upper limit. It is this upper limit that is important
to single particle methods as it establishes related coincidence error. The
high number densities were recognized as a "design driver"—particularly since
even slightly increased sensitivity to below 0.5 ym would be accompanied by much
higher number densities.
4) The 250°C upper temperature limit was also an important design driver.
It was sufficiently high to necessitate thermal-modeling work and played an
important part in the FPSSS design. This subject is treated in detail in
Section 5.
5) The range of velocities encountered in stacks is for the most part
in between those encountered by airborne spectrometers and aspirated laboratory
instruments. Our primary area of concern here was the low end of 1 m sec and
the maintenance of satisfactory airflow characteristics to satisfy the -in situ
sampling requirements. We also found turbulence to be significant in actual
field testing.
6) Assuming the ducts or stacks are round, the duct areas specified
translated to diameters ranging from about 0.4 to 7 meters. In our design, this
determined the overall distance the probe head would be maneuvered. It also
established cable and cooling line lengths. We did not judge it to be an in-
fluential design driver, although we did find that in our design, which involved
a boom-supported probe head, oscillations were induced at supporting lengths
greater than 12 feet. However, on most stacks opposing ports precluded the need
for maximum boom extension.
7) The high number densities encountered invariably lead to particle
count rates approaching several KHz. Samples of 10,000 particles or more have
statistical sampling uncertainties of less than 1% CVL/N). Thus, sampling runs
of much less than 3 minutes would be sufficient to characterize the number
-------
distribution. Quantifying the mass would ordinarily require greater sampling
times, since the larger more infrequent particles have much greater weight,
but it would in any case be stable for sample sizes accumulated in much less
than 3 minutes.
Data acquisition was an area where we already had good experience. Existing
PMS system capabilities appeared to be more than adequate to meet design goals.
An existing production data acquisition system specifically designed for such
intended usage was utilized in initial work. We found no reason to modify it
until actual stack experience dictated specific desired changes and specialized
outputs. The final FPSSS design incorporated a computer-directed data acquisition
system that embodied an on-site data analysis capability; it is described in
detail in Section 6.
8) The question of portability means portability from stack to stack and
also maneuverability within the stack. Since EPA compliance methods for stack
sampling currently require as many as 48 sample points (U.S. EPA Method 5), the
eventual similar use of the FPSSS dictated a multipoint sampling or stack-
scanning capability. We recognized the difficulty in handling unwielding appar-
atus in the design and restricted the total weight to 150 Ibs. , with no single
subsystem weighing more than 60 pounds.
-------
3. OVERVIEW OF THE FPSSS
The FPSSS is described in this section as it was finally delivered to the
EPA. Since it is a one-of-a-kind device we still must regard it as a prototype.
In the following sections we will refer (chronologically) to certain initial
configurations; however, it will benefit the reader to have a basic familiarity
with the final instrument during these presentations and analyses. A more in-
depth treatment of primary FPSSS hardware and its performance is given in Section
6.
The development of the FPSSS was not considered a high risk task, although
it did require the application of diverse and demanding technology. The per-
formance requirements were within the range of standard PMS instruments (with
the exception of the high temperatures involved). Prior successes in developing
technology to perform measurements in even more severe environments were used
to good advantage in the development of the FPSSS—for instance, the LCPS
developed for NASA and the Pioneer Venus payload withstood 450°C. This tech-
nology base was applied to the FPSSS design. An 80-node thermal model for the
probe head was developed by BBRC. It provided predictions of cooling require-
ments, thermal gradients, and mechanically induced changes in optical alignments.
The FPSSS design reflects efforts to minimize the alignment error predicted
by the model.
Our early analysis of the design requirements for the FPSSS suggested
that an approach be followed utilizing as much existing technology as practical.
We had previously designed an extractive sampling train to adapt another PMS
instrument to take stack measurements. The instrument used was a classical
scattering aerosol spectrometer (Model CSAS-100) unit; it makes measurements
while viewing particles a few inches away from its collecting optics. Those
results showed that existing optical system designs were capable of handling
the projected high particle number densities associated with stack flue gas at
sizes larger than 0.5 ym diameter. However, those measurements also revealed
size distribution spectra where the number density was rapidly increasing with
decreasing particle size. Obviously, with such a size distribution, at some
smaller size any instrument finds a limitation due to coincidence error.
Sampling ahead of control devices might further limit use—even at 0.5 ym—
10
-------
but this PMS basic detection system appeared to be the most adaptive of existing
technology and the least problematic.
The real problem was performing the measurements in the hot environment.
Stacks are often well above boiling temperature, and, if continuous measure-
ments are desired, an instrument of this type must be cooled to survive. Our
approach was to provide a reasonable thermal environment for a modified PMS
Model CSAS-100 electro-optical system using an active cooling system. The re-
sulting FPSSS design utilizes a He-Ne laser-fed high-resolution imaging optical
system with particle detection and sizing by strong forward light scattering.
The imaging system provides accurate sample volume definition and enables
particle transit time measurements to extract flow velocity.
The FPSSS has four size ranges covering 0.4-1.15, 0.5-2.0, 1.15-5.65, and
2.0-11.0 ym. Each size range has 15 size classes. In normal operation, two
size ranges are sampled concurrently (e.g., 0.5-2.0 ym and 2.0-11.0 ym), pro-
ducing 30 classes from 0.5 to 11.0 ym. The maximum number density that can
4 -3
be measured is 5 x 10 cm . A block diagram of the various FPSSS subsystems
is given in Figure 1.
Figure 2 is a photograph of the instrument head, heat exchanger, and two
lateral support bearings (mounted to port flanges). The bearings allow the
operator to extend the head on a segmented boom up to 6 meters into the particu-
late environment. The water-cooled head can operate continuously at temperatures
above 250°C. The head contains the laser, condensing and imaging optics, photo-
detectors, and programmable preamplifiers.
Figure 3 shows the FPSSS electronics console, housing the signal processing
electronics and the data acquisition and display system. Data acquisition is
accomplished using a microcomputer with firmware programs and random access
memory. Both CRT and hard copy displays are generated. Sufficient memory
capacity exists to generate size, area, mass and accumulative mass distributions
for up to 7 individual samples. In addition, numerical listings and time series
plots of selected parameters (e.g., mass loading and number density) may be
generated. Calibration parameters can be entered manually; for instance, the
particle density is invariably entered to compute mass and aerodynamic diameter.
The latter can be chosen as the relevant size parameter for various outputs.
11
-------
FRONT PANEL CONTROLS
FPSSS DAS
ELECTRONICS CONSOLE
Figure 1. FPSSS block diagram.
-------
HEAT EXCHANGER
LATERAL SUPPORT BEARINGS
INSTALLMENT HEAD
Figure 2. Photograph of FPSSS components. Two sizes of lateral support bearings are
shown for maximum boom extensions of 4 and 6 meters respectively. The heat exchanger
also contains the purge air pump. Boom sections are not shown.
-------
Figure 3. Photograph of FPSSS electronics console.
-------
3.1 FPSSS Optical System
The optical system became the primary design driver of the FPSSS. Its
alignment is the single most important optical parameter requiring control.
The FPSSS optical system shown in Figure 4 is quite similar to that originally
proposed to the EPA. The output of a 5 mW high-order multimode laser is directed
towards a 75 mm radius condensing mirror by a pair of plane mirrors. The con-
densing mirror focuses the laser beam down to approximately 125 urn diameter at
the object plane of the light collecting/imaging system. The laser beam then
expands to approximately 3 mm diameter before being absorbed by a "dump" spot
(or central field stop) on the FPSSS window. Particles in the vicinity of the
object plane scatter energy through the window which is collected by the prime
objective and relayed at 2:1 magnification to the secondary objective, which is
a microscope objective operating at 10X. The total magnification is thus 20X.
The particles are imaged onto a pair of exit faces of a beam splitter. The
reflected image face is overlaid with a narrow opaque mask aligned in the direc-
tion of particle flow. The second transmission prism face is unmasked. A pair
of photodiodes and their associated preamps view the particle images through
each prism face.
The capabilities of this optical system are of a fundamental nature in that
they provide a means for defining a desired sample volume, thus effecting the
i-n S'Ltu measurement capability (Knollenberg, 1975 and 1976a) . The masked beam
splitter derives two signals, providing (in conjunction with double pulse
height analysis) a means of determining if a particle's position is in the de-
sired sample volume. The relative size of the sample volume cross section varies
inversely with the magnification used in the collecting optics. The light
transmitted on axis through the beam splitter is the signal used to size the
particles.
3.2 FPSSS Head
The head of the prototype FPSSS shown in the photograph in Figure 2 is
further illustrated in Figure 5. The head is approximately 20" long, has a
2.25" x 3.5" elliptical cross section, and weighs 7 pounds. The head body is
fed by two water lines which form a pair of circuitous paths through the head
body, making several internal bends before running back to the rear and exiting.
Thus, four water lines are fed through the boom. Two high voltage laser leads,
15
-------
ADJUSTABLE BEAM
STEERING MIRRORS
DETECTOR
MODULE
7cm RADIUS
CONDENSING BEAM
FOLDING MIRROR
SECONDARY
20 x OBJECTIVE
AH COATED N
WINDOW \ LASER BEAM DUMP
SPOT
Figure 4. The above optical system is housed entirely
within the FPSSS head, except for the 7 cm radius mirror.
Only this mirror and the outside face of the window are
exposed to the particle media and these surfaces are
flooded with purge air.
-------
STAINLESS STEEL BOOM
REFERENCE DETECTOR
LASER MODEL 80-ST HIGH REFLECTIVITY FLAT
(COHERENT RADIATION) FIBERGLASS INSULATION MIRROR (2)
HIGH VOLTAGE
LASER LEAD
INPUT
WATER
LINE
STAINLESS
STEEL INSULATION
RETAINER AND COVER
PHOTO DETECTOR
ASSEMBLY
BEAM
SPLITTER
LOW LOSS AH COATED
WINDOW
HIGH REFLECTIVITY
MIRROR 7.5mm
RADIUS
Figure 5, Assembly diagram of the FPSSS probe head,
-------
Figure 6. FPSSS installation at IERL. Test tunnel showing the
attachment of the lateral bearing to the port flange and typical
proximity of thermal control system and electronics console.
Figure 7. Close-up oE FPSSS head and split lateral bearing. The
probe head must pass clear of the bearing before clamping second
bearing half.
18
-------
a multiconductor signal cable, and an air line also feed through the boom to
the FPSSS head.
The optical system just described demanded careful mounting. The laser is
mounted on '0' rings without adjustments. The only adjustments are x-y adjust-
ments on the first beam-folding mirror and the 75 mm mirror. The adjustment
on the 75 mm mirror was not planned initially; however; it was found necessary
when we discovered that adjustment using the beam-folding mirror had near zero
effect at the object plane because the object plane is centered at the focus
of the 75 mm mirror.
The head body was constructed in three sections to give access to the
various optical element cells and for machinability. '0' ring seals are used
for the water paths, and sealing is positive up through 100 p.s.i. The head
body is covered by 1 cm of fiberglass insulation and by a protective stainless
steel jacket with openings only for optical viewing and connections. The photo-
detector assembly loads from the rear and can be removed without splitting the
head body.
Initially, purge air was provided only for the 75 mm mirror. This element
has a small aperture and is easily purged. Contamination on its surface, however,
is more critical than on the window. We eventually purged the window, but did
not do so initially in order to determine its potential exposure and rate of
contamination. The probe head was also purged to prevent contamination of the
folding mirror surfaces.
3.3 FPSSS Support Beam and Port Bearing
The FPSSS is designed to permit installation on a stack port flange and
can penetrate into the stack to considerable length, sampling anywhere across
the inside diameter of the stack. This was accomplished by using a section
tubular support beam fed through a lateral bearing mounted to a stack port
flange. A typical installation is shown in Figure 6 and 7. All water'lines
and electrical cables are continuous through all boom sections. To add a
section one simply screws it into the preceding one. All boom sections are 4
feet in length. The boom sections are constructed of thin-walled stainless steel
tubing, 2.5" O.D. and 0.09" thickness. Approximately 0.5" of fiberglass in-
sulation and an inner 1.125" I.D. PVC sleeve allow sufficient room to slide the
lines and cables when joining and separating sections. A thin, nylon-braided
19
-------
abrasion jacket bundles all of the water, air; and electrical lines together
to allow the boom sections to slide freely. Each boom section adds about
6 pounds of suspended weight, although the final section is slightly heavier
due to the larger wall thickness required to support the increasing side loads.
The port bearing design is of a clamshell nature and is split to allow
passage of the larger probe head (see Figures 2 and 7). Once the head is by
the bearing, it is closed, forming a snug fit to the 2.5" O.D. stainless steel
walls of the boom sections. Two lateral bearings having four and six sets of
concave stainless steel rollers were constructed. The larger bearing is used
for configurations where the boom is extended beyond 12 feet (3 sections).
3.4 Thermal Control System
The thermal control system (TCS) shown in Figure 2 and diagrammatically
in Figure 8 was rather simply constructed using a small centrifugal pump, an
automotive-type heater radiator, an instrument blower; and a storage reservoir
all packaged in a single housing. The purge air pump is also housed in the
thermal control system. The water pump has a throughput of 1 gpm with 30 feet
of line and a corresponding pressure drop of 7 p.s.i. Quick disconnects are
used on all water and air lines.
3.5 FPSSS Electronics Console
The FPSSS electronics console shown in Figure 3 contains all electronics
with the exception of the laser, photodetectors, and preamplifiers in the probe
head. The mechanical envelope was constructed from aluminum extrusions for
ruggedness. The packaging is very compact yet dense, resembling a heavy suit-
case in handling. Various scope carts adapt to the unit easily.
The probe head generates three photo signals (two scattering and one ref-
erence) while accepting two control signals for switching size range gains.
(Programmable preamplifiers are used to switch gains and thus size ranges)-
The scattering signals are ac-coupled, requiring baseline restoration by sig-
nal conditioning circuitry. The scattering signal pulse train is pulse height
analyzed and processed to establish sample volume validity. The widths of the
scattering pulses are related to the velocity of the particles traveling through
the laser beam; a selected range of pulse amplitudes is processed to determine
flow velocity.
20
-------
RESERVOIR
FAN
HEAT EXCHANGER
VACUUM PUMP
WATER PUMP
CUSHION WEATHER STRIP
JSPA FPR FPSSS heat exchanger assembly.
-------
The above probe-related information is input to memory through direct
memory access (DMA) without computer control. Data entry from front panel
switches and the time-of-day clock also is input to memory via DMA. Informa-
tion as to sample duration is entered automatically through a data port under
program control; size range information is passed to the probe head and pulse
height analyzer via a complementary type of port. Program storage resides in
a 16 K ROM while data storage consists of 4 K of RAM. The RAM data storage also
serves to provide CRT refresh memory. The CRT gives a 4" display of all size
range data simultaneously, individually, or in any paired grouping. A six digit
numeric display provides close inspection of specific data. The printer-plotter
accesses information for hard copy generation as called for in the program plot-
ting sequence selected. An important feature of the system is the ability to
present time-line (strip-chart) data such as mass loading and number density
for a selected time period and plot out size and mass distributions from up to
8 sequential time segments within the selected period. A statistical analysis
package also lists important analysis summaries.
The Z-80 CPU is, of course, responsible for the exercising of all instruc-
tions and controls all operations intended to be under program control. Power
supplies for all probe electronics, data acquisition, displays, and the laser are
also contained in the electronics console. The printer-plotter is an Axiom
unit which has self-contained power supplies. Also residing in the electronics
console are a temperature sensor and thermal limit sensing circuit with an audio
alarm.
Maintenance consists primarily of cleaning and periodically checking probe
calibration with monodispersed spheres. Most problems arise from optical con-
tamination when the probe is operated with "sticky" particulates, or less fre-
quently, from mirror misalignment after physical abuse or accident. The laser
has the lowest MTBF (mean time between failure), with about a two year expected
lifetime.
3.6 Operation and Use
The FPSSS is rather easy to install and operate; results of experiments
can be interpreted readily too. The power requirements at the measurement
site are only 7 A of 115 VAC, 60 Hz. The operator screws together as many
22
-------
boom sections as required by probe insertion depth. He then makes electrical
power and signal interconnections between the probe and DAS and connects the
air and water lines between the probe and the TCS. The lateral bearing is
mounted at the stack port of interest and the probe is inserted through the
bearing into the stack, observing the orientation of the probe with respect
to the airstream. The DAS is powered, and associated data are entered to re-
flect current conditions such as time of day, stack airstream velocity (although
the velocity may be measured by the probe if so required), particulate density,
and stack diameter (used only to compute opacity). Finally, the type of out-
put desired such as time plots or spectral plots is selected for output data
logging.
The probe reports only properly sized particles to the data processing
circuitry, but it also transmits an activity signal which reflects the per-
centage of time any particle is detected by the probe, valid or not. Activity
is useful when correcting for coincidence error and is consequently monitored
for use by the operator. Automatic correction for coincidence loss is performed
by the FPSSS DAS.
Built into the probe head photodetector module are gain switchable pre-
amplifiers for the masked and unmasked detectors which are under program control.
Four sets of gains are provided, enabling the probe to operate over all four
particle size ranges, one range at a time. The FPSSS ordinarily uses one pair
of ranges to cover 0.4 to 5.65 ym and the other pair of ranges to cover 0.5 to
11, ym. The operator chooses the pair of ranges to be used at the DAS.
In each range, each properly sized, or valid, particle is assigned to one
of fifteen size classes, depending on the amount of light scattered. Under
ideal conditions with monodispersed spheres the resolution attainable approaches
the resolution provided by the system. Under the actual stack conditions, with
high temperatures and nonspherical and inhomogeneous particles, resolution of
this instrument is surely degraded but overall accuracy is sufficient for most
analyses required.
Data processing can employ any or all of the devices, whether the micro-
computer, printer-plotter, six digit numeric display, or the cathode ray tube
to display real-time data. Number density and mas's spectra may be printed or
23
-------
plotted, and parameters such as mass median diameter and opacity may be printed
or plotted versus time. For easy intercomparison against aerodynamic devices
such as impactors, optical diameters may be automatically transformed to
aerodynamic diameters.
The printer-plotter is the primary data output mechanism; the variety of
processed output options are discussed in detail in Section 6. The remainder
of this section will be devoted to a discussion of samples of data outputs.
During a data accumulation cycle the operator normally observes the raw
size spectra on the CRT. At the end of the cycle a header is printed out of
various housekeeping data, followed by program-selected processed data.
Figure 9 is a sample of numerically listed data consisting of the header
data, summary of size counts, and processed statistical information. Figure 10
is a sample of normalized size distributions from the two groups of paired size
ranges. Figure 11 gives the mass distributions for the data given in Figure 10.
Figure 12 is a plot of number density and mass loading versus time. It is a
sample of the time series (strip-chart) capabilities of the FPSSS. Mass
loading, number density (concentration), opacity, or any pair of these parameters
may be followed at selected plotting speeds down to 1 second per sample.
These data samples are obviously photostats of the actual FPSSS printer-
plotter output which is an electroconductive metalized paper. Because of
difficulties in reproduction, we have selected to redraft all other data
samples in this report. However, the presentations resemble those of the
actual output in nearly every detail.
24
-------
VEAR - DH ¥ 79-257 T ffl'E (hP M'
OPTICAL.. LARGE '
MAX f IME "•'.H-M •£?"~g0":'ftZ:"?
MAX POP CDE> " 7~a "
li{fMP?H3^
• STACTOXA. 10 :31 -11
i.ese
"
i 450.
1:350
~Z, 300
3 .SB0
4 .10.0
sis©
5 . 300
6.580
7.100
7 .700
S.300
3 . 300
'3 . 500
la.100
10.700
..
.la0ME 02
70QC1E 0i
.6003E"" 01
Figure 9. FPSSS numerical testing output samples. Above,
at the start of a data sample, a listing of housekeeping data
is printed with the time. The second listing includes the
run end time and the total number of particles measured,
average number density, mass loading, mass median diameter,
and opacity. Below, a listing of the total number of particles
counted in each of 30 size classes prior to graphics presenta-
tion. (Data from Valmont power plant, Boulder, Colorado.)
25
-------
Figure 10. FPSSS size distribution graphics samples. Left,
size distribution generated using ranges 3 and 4. Right, size
distribution generated using ranges 1 and 2. (Data from Valmont
power plant, Boulder, Colorado.)
26
-------
Figure 11. FPSSS mass distribution samples corresponding
to data in Figure 10. (Data from Valmont power plant,
Boulder, Colorado.)
27
-------
i;"t_' ri
S3 ' ^.4
58 '4C.
58 -Si
S3 ••Z7
51 03
59 -14
S3 ^28
53 '26
S3 -31
53 '37
S3 '43
S3 :4'3
o_3 • i= •-'
80 -Ml
88 '07
8 8 ' 12
88 '13
38 '24
89 '30
8 y : 3 Q
RPi - 4Z
RF1 ' 4K
88 '54
88 --S3
01 '0S
01 '11
91 '17
8'i '23
-i '2S
01 '34
01 '48
01 '46
01 '52
81 '53
82 '04
«Z '18
82 '13
02'2Z
RZ '23
82 '34
82 '4W
n "4 ' 4 b
82 'SZ
@Z 'S3
03 '84
03 -18
03 -16
fl 3 : •? •?
ti^-ll
83 '33
Pi:-: -45
03 '51
85 '43
Figure 12. Sample of FPSSS time series plot. Time is in
hours and minutes. (Data from Valmont power plant, Boulder,
Colorado.)
28
-------
4. EXAMINATION OF THE OPTICAL PROPERTIES OF PARTICLES AND THEIR
IMPLICATION FOR STACK MEASUREMENTS
In this section we review the light scattering (Mie) theory of particles.
The FPSSS design range of 0.5 to 5 ym covers a difficult region of particle size
(^ A and slightly above) for light scattering sizing techniques. Our intent is
to present those important underlying aspects that bear upon the design of all
light scattering optical particle size spectrometers. The results can easily
be extended to instruments using the extinction effect as well. We will begin
by trying to bridge the gaps in our common knowledge of scalar theory (ray
trace), physical optical theory, and wave theory (Mie results). We will pro-
vide examples of the importance of such results in the design of optical par-
ticle size spectrometers and discuss the necessary tradeoffs that must be made
in such designs. The final result is the optimum FPSSS optical system design.
The response of optical counters is invariably calculated utilizing Mie
theory. Even though one seldom deals with the ideal boundary conditions re-
quired for the exact solution to Maxwell's equations (infinite plane wave,
coherent illumination, spherical particles, etc.), the theory gives a good rep-
resentation of what to expect. The response for nonspherical particles is
only approximate in most cases. For instance, for finite cylinders the solu-
tions are only tractable for low aspect ratios (less than three) and sizes
near the wavelength. Furthermore, advanced numerical methods deal with par-
ticles of random morphology with only approximate results.
In general, Mie theory can give us a theoretical calibration curve that
later invariably must be tailored by empirical results using physical particle
standards. This is not simply due to the deviations from proper boundary
conditions cited above. It also comes from the difficulty in modeling optical
transmission and detector response as a function of collecting angle, wave-
length, and state of polarization.* In actual practice, although one might
use ideal laser sources and measured optical parameters, it is difficult to
*To our knowledge, the optical transmission characteristics are always
neglected in computed Mie results. Probably the best analysis yet provided is
that of Cook and Kerker (1975). It adequately accounted for source, spectral,
and detector responses, but still neglected optical transmission functions.
29
-------
directly use Mie-derived calibrations without some adjustments from empirical
results using monodispersed spheres.
In trying to gain insight into the optical behavior of particles, we have
found it helpful to break down the total scattering into reflected, detracted,
diffracted, and absorbed wave group components. This method can be qualita-
tively used regardless of particle morphology and size; it provides a physical
interpretation of Mie computational results. The primary scattering wave
group components are shown in Figure 13 for a particle of a few mircons' size.
As illustrated, most of the light is scattered in the forward direction, and
instruments designed to optimize collection in the forward direction obviously
provide the greatest sensitivity. This simple fact is a fundamental premise
in most optical system designs for light scattering particle counters.
The shape of the diffracted wave group is the only one that is size-
2
dependent, although the amplitude of all wave groups varies as '^ irr , or cross
section. The total amount of reflected light is nearly isotropic and very small
(< 10% of total) but proportionately greater for backscattering angles. In gen-
eral, light scattering problems are dominated by refracted and diffracted light
components. Diffracted light is always present in an amount dependent upon
intercepted cross section and equals 50% of all the light scattered. The cen-
tral diffraction peak, or lobe, is called the Airy disk and contains 85% of the
total diffracted light. The angular distribution of refracted light is a
function of the real index of refraction, n ; its overall contribution is
re
diminished by absorption, accounted for by an imaginary index value, n . The
im
presence of absorption thus generates a complex refractive index, n = n - in. ,
re im
The scattering collecting solid angle requirements for selected emphasis on re-
fracted or diffracted light are thus different and certain instruments are in-
tentionally designed to maximize the collection of one as opposed to the other.
The solid angle is generally limited by a pair of inner and outer collecting
angles which define a conic. In principle, strong forward scattering optical
system designs can reduce the inner collecting angle to essentially 0° and the
collecting geometry can be simply expressed in terms of the outer collecting
angle. The collecting angles required to collect 89% of the light diffracted
(90% of Airy disk) are shown in Figure 14. It is readily apparent that coverage
30
-------
UJ
Ill
z
a
i-
UJ
S
o
z
PARTICLE
O
DIFFRACTED WAVEGROUP
REFRACTED WAVEGROUP
ABSORBED WAVEGROUP
REFLECTED WAVEGROUP
Figure 13. Wavegroups contributing to light scattering.
-------
4-
UJ
3-
2-
1 -
0
T
10
l
20
30 40
COLLECTING ANGLE (DEGREES)
I
50
Figure 14. Required light collecting angles to collect
89% of the light refracted by particles in the 0.5 to 5 micron
range.
-------
(by diffracted light) of a 0.5 to 5 ym size range requires a reasonably large
outer collecting angle for small particles and a small inner collecting angle
for larger particles. For most refractive indices, refracted light accounts
for up to 45% of the remaining scattered light, its exact amount determined by
the complex index. Figure 15 shows the relationship between collecting angles
and refractive index in order to collect all refracted light excluding that
involving internal reflections. It is important to bear in mind that absorption
occurs from the attenuation of refracted light and that size affects only its
magnitude and not angular distribution.
Note that the refracted and diffracted wave groups are strongly forward
scattered. They dominate the scattering characteristics of most particles.
Since they normally contribute greater than 90% of the scattered light, the
highest detectively of particles is obtained by optical systems oriented to
collect them.
Figures 16a, b, and c illustrate the total scattering phase functions for
particles of 0.6, 2, and 5 ]im diameter computed from Mie theory. The pronounced
forward scattering is obvious in all cases and for the index and size used con-
tains refracted and diffracted components in similar quantities. If one intro-
duces varying amounts of absorption, the amplitude of the refracted wave is re-
duced, with an at first expected reduction in the total amount of forward scat-
tering. However, at this point one must be careful not to oversimplify the Mie
results. For coherent radiation the phases as well as the amplitudes of the
scattered waves must be considered. Since the refracted wave suffers a phase
lag proportional to the optical path length it traverses, while the diffracted
wave's phase relationship is defined as a function of scattering angle and size,
they may interfere to produce a resonant combined amplitude greater than less
than the sums of their independent intensities—i.e., coherence-induced resonance
or destruction interference. The effect is most pronounced when the particle
size is chosen so that the central maximum of the diffraction pattern (Airy disk)
is of comparable size to the refracted wave cone and collecting solid angle is
sized to just collect both scattered contributions. The phase change from the
center to the edge of the Airy disk is ir/4 (see Borne and Wolfe, 1972). Because
there is just such a size for any fixed collecting solid angle, it is impossible
33
-------
2.0
1.8 -
x
UJ
Q 1.8 -
LU
oc
u.
1.4 _
1.2 -
1.0
20
40
60
80
100
DEGREES COLLECTING ANGLE
Figure 15. Required collecting angles in order
to collect 80% of all refracted light involving
single pass transmission through the particles.
The complex index is assumed to be zero.
34
-------
1U
10
10 -
<2 10 -
UJ
o
S 1
UJ
10
.-1 -
10
-2
I
40
I
80
T
120
160
200
SCATTERING ANGLE (DEGREES)
Figure 16a. In the above diagram, the S-^ and 82 polarization
states and average intensities are shown for monochromatic
light of X = 632.8 nm. (D = 0. 6
35
-------
10
40 80 120 160
SCATTERING ANGLE (DEGREES)
200
Figure 16b. In the above diagram, the S, and 82 polarization
states and average intensities are shown for monochromatic
light of X = 632.8 nm. (D = 2
36
-------
o
<
£
10
106H
105-
I 104H
10 -
1 -
10
,-1
40 80 120 160
SCATTERING ANGLE (DEGREES)
200
Figure 16c. In the above diagram, only the average intensity
is shown for monochromatic light of X = 632.8 run. (D = 5
37
-------
to avoid the effects. The general result is a smaller or greater scattering
response than expected; if pronounced it can lead to ambiguous scattering-size
relationships (i.e., a larger particle producing a smaller signal).*
Regardless of whether the refracted and diffracted waves overlap and inter-
fere, a similar result occurs if the inner scattering angle cannot be maintained
close to 0°, which, while easy to do with lasers, is nearly impossible to approach
(within 10°) without them. This limitation results from the fact that as par-
ticles increase in size (e.g., from 1 to 5 ym) the diffracted light is suppressed
to smaller and smaller angles and eventually cannot be separated from the trans-
mitted light of the source. As previously stated, the refracted light is main-
tained in an angular cone that does not change with size but is generally smaller
in contribution than the diffracted light. Under such conditions there is the
probability that a smaller particle, whose strong forward scattering contribution
(dominated by diffraction) is well matched to the collecting optics solid angle,
may produce a larger collected response than a larger one whose diffracted light
is lost inside the apodimized region of the collecting optics. It occurs whether
coherent or incoherent light is used. Thus, ambiguities in sizing are the ex-
pected rule in certain size ranges. In order to select the most appropriate
collecting geometry for the FPSSS, various existing designs were reviewed and
a series of Mie computations performed.
In Figures 17 and 18 two responses from Cooke and Kerker (1975) show the
optical oscillating response in two optical counters using white light. Figure
19 shows the similar results for the laser-fed PMS CSASP-100. If one chooses
to avoid the forward region, a smoothed response is obtained, as illustrated in
Figure 20 for a white light counter and in Figure 21 for a laser-fed PMS LAS-200.
These are both essentially refracted light instruments in the 0.5 to 5 ym range.**
*Coherence-induced resonance effects are generally observed to be damped be-
cause of lack of fulfillment of plane wave conditions. Multimode lasers are
sufficiently lacking in spatial coherence to reduce some of the effects (Knollen-
berg, 1976b) . It is a primary reason for the use of high-order multimode lasers
in CSASP instruments.
**The above result may appear surprising to some since it is often claimed
that lasers are not as desirable as white light sources for particle sizing
purposes because of the coherence-induced resonant interference. However, the
overall problem is obviously dominated by collecting optics consideration.
38
-------
Figure 17. Response of the Royco 218 particle
counter for particles with refractive indices
1.33, 1.70, 1.90, and 2.20. The curve for 2.50
is very close to that for 2.20 (from Cooke and
Kerker 1975).
39
-------
10
0.1
1.70
0.1
0.5 1.0
DIAMETER
10
Figure 18. Response of the Royco 245 particle
counter for particles with refractive indices
1.33, 1.45, 1.70, 2.20 and 2.50. The curve for
1.90 is very close to that for 2.20 (from Cooke
and Kerker 1975).
40
-------
CM
E
o
(0
CO
CO
o
oc
o
10
0.1 1.0
DIAMETER (/Jim)
Figure 19. Theoretical response for PMS CSASP-100 instrument
using 4° to 22° collecting geometry for refractive indices
1.30, 1.40, 1.50, 1.60, 1.70, 1.80, 1.90 and 2.00.
41
-------
100
DIAMETER
Figure 20. Response of the Climet particle sensor for
particles with refractive indices 1.33, 1.45, 1.70 and
1.90. The curves for 1.90, 2.20, and 2.50 are very
close to each other and above radius 0.45 ym are very
close to 1.70.
42
-------
u
"
g
10
-7
10
-------
One might invoke that the large collecting angle (nonforward) is most de-
sirable because of the monotonic calibration curve. However, if one considers
the effects of absorption this is not the case. Again from Cooke and Kerker we
can compare the instruments in Figures 22 and 23 with absorbing particles shown
in Figures 17 and 20. Clearly, unless one knows he is dealing with transparent
spheres, the Royco instrument has less total uncertainty. The Climet, like the
LAS-200 (being a refracted light instrument) loses signal when the refracted
light is attenuated by absorption. Furthermore, if one considers unknown shape,
the larger collecting angles of the Climet or the PMS LAS-200 are subject to
scintillations (direct reflections from crystal planes) and without the morphology
independent diffracted light contribution may give quite spurious results.
In our experience and from this current analysis, it is clear that a strong
forward scattering system would generate the least uncertainty in sizing, with
the exception of particles that are nearly spherical and rather transparent. A
laser-fed optical system provides the opportunity to collect the strongest for-
ward scattered light. The particles in a stack are often spherical but seldom
transparent. Ordinarily, they are translucent, made up of possibly several
chemical species of differing refractive index. Most are at least slightly ab-
sorbing at visible wavelengths. Considering these facts, we selected a strong
forward scattering laser-fed optical system. Finally, we can also note that the
phase integrity of the refracted light is randomized in translucent particles, re-
ducing potential interference and coherence-induced resonance.
Also, it may not be apparent, but a range of collecting angles from 35 to 120°
cannot easily be used for -in situ measurements unless another means is used simul-
taneously to define particle position. Reflecting optics are required for the
large collecting angles of the Climet and the LAS-200 instruments, and they cannot
be easily manufactured with the tolerance demanded by the imaging required for
sample volume definition, nor is a particle in situ if totally enwrapped in optics.
If we truly desire in situ measurements we should also point out that back-
ground light is nearly impossible to deal with unless monochromatic sources
are used to enable filtering. If a monochromatic situation is desired, then
a laser is even more desirable. As it turns out, filtering is required less
often when using laser sources because of their higher energy density permitting
the use of photodiodes rather than the more sensitive, but also quite temperature-
44
-------
o
So
Uj CO
> w
= Q
UJ
oc
1.54 - O.OSi
.1 0.5 1
DIAMETER
10
Figure 22. Response of the Royco 218 particle counter
for particles with refractive indices 1.54, 1.54-0.051,
and 1.54-l.Oi. The curve for 1.54-0.51 is very close to
that for 1.54-l.Oi.
45
-------
"
ga
U- 2
Eg
LU
> O
P o
< oc
J O
1.54 - O.OSi
0.1
0.5 1
DIAMETER
5 10
Figure 23. Response of the Climet particle sensor for
particles with refractive indices 1.54, 1.54-0.051,
1.54-0.51, and 1.54-1.01.
46
-------
sensitive, photomultiplier tubes. The reader is referred to Knollenberg (1975,
1976b) for detailed discussions on the advantages of laser light sources over
white light sources and comparative signal-to-noise analysis of solid state
photodiodes and photomultiplier tubes.
Thus far we have discussed many aspects of particle optics, possible effects
of differing light sources, and factors affecting design. What can we conclude
with regard to the possible FPSSS optical design? First, that we want a laser
light source. Second, that we need a collecting optical design that permits
the collection of light as close as possible to the optical axis. If the par-
ticles are not nonabsorbing spheres and are to be measured -in situ, a forward
scattering imaging optical system is required and the best light source is a
laser. The task remaining is thus one of optimizing a forward scattering optical
system for the FPSSS.
There are two standard instruments manufactured by PMS that use strong for-
ward scattering; the Model ASAS-100 and the Model CSAS-100. Both happen to
have 4-22° collecting angle geometries. The computed response for 4-22° col-
lecting angles is given in Figures 24 and 25. Figure 24 shows the range of
responses for real indices from 1.33 to 2.5 in the absence of absorption. Con-
siderable uncertainty exists at sizes from 1 to 2 microns and indices between
1.33 and 1.7 due to strong resonance. Furthermore, above 6 microns there is a
systematic difference in sizing; the signal is greater the lower the real re-
fractive index. Figure 25 exhibits the sensitivity to dispersion of imaginary
index with a real index of 1.5. The primary effect of strong absorption is
in the smoothing of the oscillations—the effect being more noticeable the larger the
size or the greater the amount of absorption. For strong absorbers a systematic
oversizing begins to show up at diameters greater than 4 microns.
These results have been known for some time. In many cases we have found
a reduction in the sensitivity to changes in both real and imaginary index com-
ponents by reducing the inner collecting angle. By selectively manipulating the
inner and outer collecting angles, we determined that the least sensitivity to
changes in complex refractive index is achieved with a range of collecting angles
of approximately 2-11°. Figures 26 and 27 show the computed sensitivity to dis-
persions in real and imaginary components, respectively, for 2-11°. Comparison
47
-------
SENSITIVITY OF 4°- 22° COLLECTING ANGLES
TO REAL INDEX DISPERSION
10
ILj
(/>
z
o
a
w
UJ
EC
a
z
E
LU
o
W
HI
UJ
cc
10-
10
0.1
n 1.33 - .OOi
n 1.7 - .OOi
n 2.5 - .OOi
I I I
1.0 10
DIAMETER (ptm)
Figure 24. Computations above for A = 632.
100
run.
48
-------
SENSITIVITY OF 4°- 22° COLLECTING ANGLES
TO COMPLEX INDEX DISPERSION
10 -
HI
V)
z
o
Q.
V)
HI
c
a
z
E 3
HI 10 •
o
V)
HI
ui
oc
10-
10-
0.1
n 1.5-.OOS
n 1.5 - .011
n 1.5-.1i
1—r
10
100
DIAMETER (fJ.m)
Figure 25, Computations above for X = 632.8 nm.
49
-------
SENSITIVITY OF 2°-11° COLLECTING ANGLES
TO REAL INDEX DISPERSION
10
104-4
UJ
CO
z
o
Qu
-------
SENSITIVITY OF 2°- 11° COLLECTING ANGLES
TO COMPLEX INDEX DISPERSION
10
UJ
V)
z
o
a.
V)
UJ
tr
o
z
E
UJ
o
V)
UJ
UJ
cc
10-
102H
10-
n 1.5-.OOI
n 1.5-.01i
n 1.5 -.1i
0.1 1.0 10
DIAMETER (/J.m)
Figure 27. Computations above for X
100
= 632.8
51
-------
with Figures 24 and 25 shows improvement in terms of the dampening of the oscil-
latory behavior of the response functions and the elimination of systematic
errors in the responses attributable to variations in real or imaginary com-
ponents. A further advantage of the 2-11° collecting angle range for particle
sizes from 0.5 to 10 microns is that small errors in either inner or outer col-
lecting angle have negligible effect on the scattering response. Figure 28
shows the effect of varying either collecting angle by 1°. It is clear that
changes on the order of 1° should result in nearly negligible difference from
the 2 to 11° collecting angles selected.
Of course, we are trying to optimize for a range of 0.5 to 10 microns.
Were one to concentrate on the submicron region, one might opt for an outer
collecting angle of 22°. Likewise, if the emphasis were on sizes as large as
50 microns, the optimal and the outer collecting angles would be smaller still.
In any case, utilizing a best fit calibration curve through the center of the
envelope of nested curves in Figure 26 results in a tolerance of 0.3 microns
at 1 micron diameter; 0.4 microns at 2 microns diameter, reducing to less than
10% at sizes greater than 6 microns diameter. The average error is, of course,
much smaller. Verifications of these theoretical results are deferred to
Section 7 -
52
-------
SENSITIVITY OF 2 -11° COLLECTING ANGLES
TO 1° ERRORS
Ul
W
z
o
a.
v>
ui
cc
o
cc
HI
<
o
(A
UJ
>
UJ
cc
2 -11
2 -12°
1-10° and 2 -10°
10
DIAMETER
Figure 28. Computations above for X = 632.8 nm.
100
53
-------
5. RESULTS OF AERODYNAMIC TESTS AND THERMAL MODELING
Aerodynamic Tests
As mentioned previously, aerodynamics play an important role in the in
situ measurement of particle size. To provide the least disturbance to a field
of particles their trajectories must not be strongly influenced by the presence
of the measuring device. Since a particle's trajectory is basically controlled
by the surrounding pressure, which controls the streamlines of the carrier of gas,
modeling or measurement of the pressure field at the desired point of measurement
is quite useful. From experience in measurements at high number densities we
anticipated that our greatest success would come from pursuing the basic CSAS
design. This instrument concept requires near proximity to the particle field
to be viable. The primary collecting optics must be only a few inches away
from the particles being sampled. More importantly; the collecting angles re-
quired are fairly well fixed. Since the diameter of the collecting optics would
increase in proportion to the distance removed from the particle, any advantage
of maximizing the working distance is cancelled by the increasing size of the
collecting optics because the aerodynamic disturbances scale in an identical
manner, i.e., the degree of disturbance is directly proportional to the size
of the intruding object and inversely proportional to the distance from it.
We realized that some disturbance to the particles would probably exist
and looked at this problem in the early design stages. In fact, our pre-proposal
experimentation included wind tunnel testing of a mock-up probe model to define
the magnitude of the problem. That and early tests during the contract period
centered around a probe construction utilizing an annular sample ring construction
to force a balanced flow around the sample point. This early FPSSS probe con-
figuration is shown in Figure 29; it could accommodate the optical design sug-
gested in the previous section.
Wind tunnel tests were conducted on the configuration using test models
employing 4" by 6" diameter sample rings. The resulting pitot pressure profiles
are given in Figure 30 and 31 as measured in our wind tunnel (see Appendix I
for a description of the PMS wind tunnel test facility). These figures indicate
essentially identical flow (pressure) fields for the two tests. The tests also
showed that there is nothing to be gained by the larger structure. We were
54
-------
WATER JACKET
INLET
WATER INSULATION
ELECTRICAL
WIRING
FEEDTHROUGH
EXTENSION TUBE
BOOM HANDLE
Figure 29. The above diagram depicts the first conceptual FPSSS probe head, models of which
underwent initial wind tunnel testing. Both 4" and 6" diameter rings were tested.
-------
TEST WITH 4 INCH ANNULAR RING PROBE MOCK-UP
2.5-
1.0
A FREE STREAM VELOCITY
1 1 1—
1.2 1.6 2.0 2.4 2.8
DISTANCE FROM BOTTOM (INCHES)
O.O O.4 O.8
3.2 3.6 4.O
Figure 30. Probe mock-up mounted in vertical orientation.
-------
TEST WITH 6 INCH ANNULAR RING PROBE MOCK-UP
A FREE STREAM VELOCITY
1.0
0.4 1.2 2.0 2.8 3.6 4.4 5.2 6.0 6.8 7.6
DISTANCE FROM BOTTOM (INCHES)
Figure 31. Probe mock-up mounted in vertical orientation.
-------
somewhat surprised by the magnitude of the reverse eddy as one approached the
probe body—roughly one half that of the free stream flow. The best region for
sampling was found to be closer to the external mirror than the objective. One
notable difference between the two envelopes was the observation that the measured
velocity at the sample ring center was less than free stream for the 6" ring
but greater than free stream for the 4" ring. We attributed this effect to
the fact that the ring thicknesses were the same (3/8"), forcing a larger re-
lative volume through the smaller ring's I.D. Other than the thickness of the
ring, exact similitude was followed.
The above aerodynamic results were satisfactory as the basis for a proto-
type design. However, other factors required the consideration of alternatives.
Most important was the observation that nearly all stack installations had
standardized 4" series pipe ports. The maximum diameter circular ring that could
be inserted along the lines of our conceptual model was 3". A rough scaling of
our tunnel results indicated that it might be risky> and iterations with our
thermal model (described below) indicated added areas of uncertainty. We thus
considered an alternative model which became our adopted configuration.
The final FPSSS probe head configuration is shown in Figure 32. It strongly
resembles the LCPS design—particularly with regard to the pair of standoffs
to support the external beam folding mirror. The cantilever mirror support de-
sign allows particles to be sampled before encountering the disturbance created
by the standoff support rods. Pressure field mapping was conducted axially
along the optical axis, transversely across the object plane, and transversely
behind the pair of standoffs as indicated in Figure 32. The results shown in
Figures 33-35 were better than expected. In particular, the axial scan of
Figure 34 showed the surprising loss of the reverse eddy near the window. In
fact, we were not able to sample within 1/4" of the window; a reverse eddy
likely still exists within the boundary layer. However, only evidence for
greatly reduced flow was found behind the shield. A slight positive excess
is noted ahead of the shield. We concluded that the improved results were largely
due to the loss of the sample ring and the addition of the aerodynamic shields.
Further work was not justified in attempting to investigate further alterations
of the design. Additional tunnel testing was conducted to tailor the aerodynamic
58
-------
WIND TUNNEL SCAN LOCATIONS
AIRFLOW
A AXIAL SCAN ALONG OPTICAL AXIS
B TRANSVERSE SCAN ACROSS OBJECT PLANE
C TRANSVERSE SCAN BEHIND STANDOFFS BEHIND OBJECT PLANE
Figure 32. FPSSS probe head configurations during wind tunnel
tests. Initial mapping of the pressure field used a miniature
pitot-static system on this head configuration, results of
which are presented in the following figures. More recently,
they have been verified with a hot-wire anatnometer.
-------
TEST WITH BREADBOARD FPSSS PROBE HEAD
SCAN ALONG OPTICAL AXIS FROM WINDOW TO MIRROR
INSIDE
WINDOW
SHELD
O
M
X
O
-• •-
A SCAN
OBJECT PLANE
FREESTREAM VELOCfTY
30 m sec
-1
MIRROR
SHIELD
EDGE
0.4 O.6 0.8 1.O
1.2 1.4
INCHES
1.6
1.8 2.O 2.2 2.4
Figure 33
-------
TEST WITH BREADBOARD FPSSS PROBE HEAD
TRANSVERSE SCAN ACROSS SAMPLE PLANE
8-
7-
6-
5-
O
N
I
oj E 4H
a.
3-
2-
1-
BSCAN
FREESTREAM VELOCITY
30 m sec"1
0.2 0.4 0.6 0.8
1.0 1.2 1.4
INCHES
Figure 34
1.6 1.8 2.0 2.2 2.4
-------
TEST WITH BREADBOARD FPSSS PROBE HEAD
TRANSVERSE SCAN ACROSS SAMPLE PLANE BEHIND STANDOFFS
C SCAN
STANDOFF POSfTIONS
FREESTREAM VELOCfTY
2.2
2.4
Figure 35
-------
mirror and window shields as well as to investigate the effects due to mis-
alignment of the probe to the flow. Water injection was used to study the
shield designs. (Their primary purpose is to prevent the soiling of optical
components.) No noticeable deleterious effects were discernible in the flow
field at misaligned test angles up to 20° (axially).
Thermal Modeling
All of the FPSSS probe designs required an active cooling system. This is
necessitated by the proximity of the laser and photodetector circuitry to the
high temperature environment. Computations showed that although insulating
materials were capable of maintaining operable temperatures in an unpowered
probe for a couple of hours with a kilogram of phase change material the head
load from the laser itself was sufficient to demand active cooling. These in-
ternal heat sources are given in Table I.
Obviously, if we are to have electronics and optics (aligned) operate
continuously in a 250°C environment we definitely need to cool them. Thus, as
a first step Ball Brothers Research Corporation (BBRC) provided a first-order
analysis of the potential thermal load. For computational purposes the FPSSS
head body envelope considered was essentially that of Figure 36 which is a
water cooled, flat walled, insulated cylindrical section of 30 cm length. Our
goal was to calculate the thermal resistance (R) of this envelope, then compute
from this calculation a temperature rise for the water for an estimated reasonable
water flow.
We used the following input values:
Water flow: 2 gpm
— -2 -1
h = heat transfer coefficient 284 Watts (W) m °C
for stack gas:
K - thermal conductivity of 0.052 W m °C
fiberglass insulation:
The Thermal insulation resistance, R. = -z~, must be calculated as a parallel
1 i
network of conductance terms for the flat side walls and cylindrical top and
bottom sections:
63
-------
TABLE I
FPSSS COMPONENT POWER DISSIPATIONS
NODE
NO.
11
23
50
DESCRIPTION
PC Board
Laser
Window Heater
POWER
(WATTS)
1.5 to 2.0
12 to 15
1 to 5
-------
0.5 cm
HEAD
BODY
STAINLESS
STEEL JACKET
T
3.5 cm
t
5 cm
3.5 cm
JL
INSULATION
Figure 36. Simplified model of the FPSSS head
cross section used for thermal analysis.
65
-------
= 2ITkL k 2 HL
i L D /D. X
n o i
_ 2 11(0.052) (0.40) (0.052) (0.04) _ , 0 -1
~ + 0^05 ll264 w c
Then,
R. = •- = 0.7912 °C
i ^ .
i
The thermal resistance between the outside wall and the stack gas can be
calculated by
sg hQ (HDL 4- 2 HL) (284) (0.1281)
= 0.02751 °C
The total thermal resistance is the series sum of the insulation resistance
and that between the outside probe wall and the stack gas :
I R = R. + R
i sg
= 0.08187 °C w"1
C = — - = 1.2214 W
L R
We assumed a temperature difference between water and stack gas
= 250 - 45 = 205°C
And thus our thermal load is
q = CAT
= (1.2214) (205°C) = 250.4 W
The temperature rise of the water is
AT = 2504
w Cxi gpm (1) (126.2 grams/sec) (4.18) >
Since the temperature rise computed is small, additional iterations change
the resultant only slightly. A low temperature rise is important in that it lowers
the thermal gradient in the envelope and reduces the "hot dog" type bending that
66
-------
could introduce optical misalignment. The small amount of water required is
an advantage in minimizing total system packaging complexity. In fact, actual
tests revealed that halving the water supply and doubling the temperature were
tradeoffs. In any case, the ability to cool an instrument assembly of the type
we envisioned in the stack environment was found to be a straightforward pro-
cess.
The thermal numerical model of the FPSSS was designed to be flexible and
allowed an iterative approach to the final design. Essentially what was modeled
here is a thermal control system (TCS). The TCS is designed to protect the
sensor head from thermal distortions that could degrade optical performance, to
cool the electronics, laser, and internal optics, and to prevent condensation
from forming on the external optics and window. It consists primarily of fiber-
glass insulation, coolant loop, and a window heater. (The window heater was
later deleted when purge air was added.) The following description gives the
important thermal properties of the TCS.
Both the sensor head and the boom are thermally protected from the stack gas
environment by a one—half inch thick layer of fiberglass insulation. The insula-
tion on the head is sandwiched between the head itself and a thin metal jacket
that protects it from being damaged by impact and/or abrasion. The boom in-
sulation is located inside the 2ij-inch diameter tube that forms the jacket for
the boom. The coolant loop consists of two parallel flow passages (^-inch in
diameter) through which one to two gpm of water are circulated. Water is pumped
through two ^-inch I.D. tubes in the boom to the sensor head where it flows first
along the bottom of the head (from the rear optics housing to the front) and
then back out through the top. On leaving the head it flows back through the
boom to a heat exchanger where it mixes and rejects (to the ambient air) the
heat that it has picked up in the boom and sensor head. Included in the model
was a 2-watt heater mounted on the periphery of the window on the front of the
sensor head used to prevent condensation from forming on the window. It was en-
visioned that the heater would keep the window temperature above that of the
stack gas but it was later replaced with a purge air system. The external
optics, which are connected to the sensor head by low conductance standoffs,
were not modeled with heating elements, because their temperature was expected
67
-------
to be above the dew point of the stack gas.
The thermal performance requirements on which the FPSSS thermal control
system (TCS) is based are as follows:
a. The sensor head (optics housing) shall not exceed 70°C (158°F).
b. None of the electronics shall exceed 85°C (185°F).
c. Temperature gradients in the sensor head (either side-to-side or top-to-
bottom) shall be no greater than 1°C (1.8°F).
d. The window heater shall maintain the outer surface of the window
approximately 10°C (18°F) above the environment.
It was assumed that the FPSSS would not be subject to stack gas temperatures
in excess of 250°C (482°F) and the ambient air temperature would not exceed
35°C (95°F).
The FPSSS thermal analyses were performed with the SSFLD computer program,
a version of BBRC's steady-state thermal analyzer program known as SSI that
has the capability of modeling a fluid loop. The input data describe the thermal
network of conductive, radiative, and directive conductive couplings between
nodes that represent the parts of the hardware for which temperatures are cal-
culated. Boundary nodes are defined with specified temperatures and become sources
or sinks. Thermal heat dissipation is applied to each node that has joule or
other internal heat generation. The directed conductive couplings permit heat
conduction in one direction only and are used to mathematically establish the
direction of the fluid flow in the stationary solid nodes.
The program forms energy balance equations for each floating node (not
4
specified temperature), and the resulting nonlinear equations in T and T are
solved by iteration for the temperatures (T). After convergence, the energy
balance is computed and tabulated for each node. The heat flow through each
coupling to every other node and the nodal dissipation are tabulated along with
the residual unbalance at convergence.
The FPSSS thermal mathematical model consists of 103 nodes, of which 100 are
mass nodes, 1 (the window heater) is a massless node, and 2 (the ambient air
and stack gas) are specified nodes. The three types of nodes are defined as
follows:
a. Mass - a node that has a finite thermal capacitance; its temperature
68
-------
is calculated in the network solution.
b. Massless - a node that has no thermal capacitance; its temperature
is calculated in the network solution.
c. Specified - a node, sometimes referred to as a boundary node, that has
an infinite thermal capacitance; its temperature is specified in the
network solution.
Figure 37a and 37b show the nodal breakdown and Table II lists the nodes in
numerical order and gives a complete description (identification number, material,
and thermal capacitance) of each. See Table II for the therrno-physical properties
of the various materials used in the construction of the FPSSS and Table III for
the computed radiation conductances.
The steady-state thermal analyses discussed here were performed to determine
the equilibrium temperatures that the FPSSS would attain if it were exposed to a
250°C stack gas and rejecting heat to 35°C ambient air. Three different coolant
flow conditions were considered. They were:
balanced flow of 1 gpm in each loop;
balanced flow of 0.5 gpm in each loop;
unbalanced flow with 1 gpm in one loop and 0.5 gpm in the other.
These flow rates bracket the range of flow that can be easily obtained using avail-
able circulating pumps. The unbalanced flow condition was analyzed to determine
if the sensor head would "hot dog" as a result of side-to-side and top-to-bottom
temperature gradients.
The steady-state temperatures associated with each of the above flow con-
ditions are presented in Table IV. Table V presents a few of the most signifi-
cant findings from the analyses including the coolant temperature rise through
the FPSSS, the amount of heat rejected, and certain component temperatures. For
a description of the nodes listed in these tables, refer to Table II.
The results indicate that the FPSSS TCS would do the job for which it was
designed, i.e., cool the internal optics and laser, minimize thermal distortion
in the sensor head, and prevent condensation from forming on the window. In
actual practice, we found that the heater was not required to prevent condensation
from forming on the window. However, the purge air was necessary to preclude
ash deposition. We reran computations without the heater source, resulting in
slight changes in the temperatures associated with the flow conditions shown in
69
-------
FPSSS SENSOR HEAD NODAL LAYOUT
LOOKING TOWARD FRONT END
-g
O
Figure 37a
-------
NODAL DIAGRAM OF FPSSS COOLANT LOOP
BOOM (Four 5 ft sections)
SENSOR HEAD
HEAT
EXCHANGER
438 428 418 408 104(114 124 134 144 154
LLJ , !.. ILLJL-L-LJ-,
"338 328 318 308 108! 118 128 138 148 158
337 327 317 3O7 107 ! 117 127 137 147 157
TTIi i I I
437 427 417 407 103J113 123 133 143 153
-,-.—. i
7O 1309 436 426 416 406 102! 112 122 132 142 152
I —I _ * * • •!• • • * •
L-L I. M U... I. JJ.
336 326 316 306 1O6! 116 126 136 146 156
335 325 315 305 105 j 115 125 135 145 155
TTTTTTTTTT
435 425 415 405 405! 111 121 131 141 151
-— TUBES
COOLANT
TUBES
TUBES
COOLANT
TUBES
Figure 37b
-------
TABLE I I
FPSSS NODAL DESCRIPTIONS
NODE
NO.
1 1
12
21
22
23
31
32
33
41
42
43-44
45
46
50
70
99
101-104
105-108
111-114
115-118
121-124
125-128
131-134
135-138
DESCRIPTION
PC Board
Reference Detector
Beamsplitter & Photo Detectors
Nikon Microscope Objective
Laser
Lens
Mi rror Mount Insert
Mi rror Mount
Window
Window Frame
Outer Mirror Standoffs
Outer Mi rror Mount
Outer Mi rror
Window Heater
Amb ient Air
Stack Gas
Rear Optics Housing
Coolant in Rear Optics Hsg.
Main Optics Housing, Rear Half
Coolant in Rear Half of Main
Optic Hsg.
Main Optics Housing, Front
Half
Coolant in Front Half of Main
Optics Hsg.
Front Optics Housing, Rear
Sect ion
Coolant in Rear Section of
Front Optics Hsg.
MATERIAL
Compos! te
Compos i te'
Compos i te
Compos! te
Compos i te
Compos i te
Al 6061
Al 6061
Glass
Al 6061
SS 304
ss 303
Al 606]
Water
Al 6061
Water
Al 6061
Water
A! 6061
Water
CAPACITANCE
(BTU/°F)
0.05
0.001
0.001
0.016
0.375
0.038
0.017
0.007
0.01
0.01
0.015 (ea)
0.017
0.027(ea) lower/
0.045(ea) upper
0.007 (ea)
0.046(ea) lower/
0.063(ea) upper
0.008 (ea)
0.035(ea) lower/
0.059(ea) upper
0.006 (ea)
0.003(ea) lower/
O.OlO(ea) upper
0.003 (ea)
72
-------
TABLE I I (continued)
NODE
NO.
141-144
145-148
151-154
155-158
201
202
203
301
302
305-308
309
310
315-318
325-328
335-338
405-408
415-418
425-428
435-438
DESCRIPTION
Front Optics Housing, Center
Section
Coolant in Center Section of
Front Optics Hsg.
Front Optics Housing, Front
Section
Coolant in Front Section of
Front Optics Hsg.
Optics Housing Cover, Rear
End
Optics Housing Cover,
Cylindrical Section
Optics Housing Cover, Front
End
Extension Adapter & Coupling
Boom Jacket
Coolant in Boom Section
Nearest Optics Hsg.
Heat Exchanger in 5/8 Coolant
Loop
Heat Exchanger in 6/7 Coolant
Loop
Coolant in Boom (Midsection)
Coolant in Boom (Midsection)
Coolant in Boom Section
Nearest Heat Exchanger
Boom Tube Section Nearest
Optics Hsg.
Boom Tube (Midsection)
Boom Tube (Mi dsect ior>)
Boom Tube Section Nearest
Heat Exchanger
MATERIAL
Al 6061
Water
Al 6061
Water
SS 304
SS 304
SS 304
ss 303
SS 321
Water
Water
Water
Water
Water
Water
SS 321
SS 321
SS 321
SS 321
CAPACITANCE
(BTU/°F)
O.OlO(ea) lower/
0.045(ea) upper
0.003 (ea)
0.003(ea) lower/
0.017(ea) upper
0.001 (ea)
0.009
0.02
0.009
0.02
0.02
0.235 (ea)
1.080*
1.080*
0.235 (ea)
0.235 (ea)
0.235 (ea)
1.065 (ea)
1.065 (ea)
1.065 (ea)
1.065 (ea)
^Combined into a single heat exchanger (Node No. 309) in final analyses
73
-------
TABLE III
FPSSS RADIATION CONDUCTANCES
NODE B(N.M)
N M FT**2
11 101 0.130E-01
23 12 0.300E-02
23 113 0.210E-01
41 50 0.850E-02
NODE B(N.M)
N M FT**2
11 102 0.130E-01
23 103 0.800E-02
23 114 0.210E-01
42 50 0.660E-00
NODE B(N.M)
N M FT**2
11 103 0.130E-01
23 104 0.800E-02
23 123 0.120E-01
NODE B(N.M)
N M FT**2
11 104 0.130E-01
0 0 O.OOOE-00
23 124 0.120E-01
-------
TABLE IV
FPSSS TEMPERATURES FOR BALANCED 1 GPM FLOW
NODE TEMP
NO. (°F)
11 175.867
31 129.810
43 460.971
101 129.624
106 128.322
113 130.541
118 128.992
125 128.541
132 128.922
137 128.771
144 130.151
151 130.010
156 128.644
203 475.986
307 129.537
317 130.085
328 130.632
405 128.368
416 127.817
427 131.011
438 131.557
NODE TEMP
NO. (°P)
12 131.568
32 130.601
44 460.971
102 129.624
107 129.168
114 130.541
121 129.501
126 128.541
133 129.606
138 128.771
145 128.616
152 130.010
157 128.673
301 248.644
308 129.537
318 130.085
335 126.328
406 128.367
417 130.468
428 131.000
NODE TEMP
NO. (°F)
21 130.047
33 161.052
45 479.584
103 131.200
108 129.168
116 128.481
122 129.501
127 128.850
134 129.605
141 129.469
146 128.615
153 131.052
158 128.673
302 463.298
309 128.751
325 126.882
336 126.328
407 129.917
418 130.485
435 126.712
NODE TEMP
NO. (°F)
22 130.047
41 529.505
46 479.804
104 131.200
111 129.553
116 128.441
123 130.278
128 128.850
135 128.589
142 129.469
147 128.722
154 131.052
201 459.438
305 127.986
315 127.434
326 126.882
337 131.179
408 129.917
425 127.265
436 126.712
NODE TEMP
NO. (°F)
23 148.201
42 540.253
50 543.823
105 128.322
112 129.553
117 128.992
124 130.278
131 128.922
136 128.589
143 130.151
148 128.722
155 .128.644
202 449.726
306 127.986
316 127.434
327 130.632
338 131.179
415 127.817
426 127.265
437 131.557
-------
TABLE IV (continued)
FPSSS TEMPERATURES FOR BALANCED 2 GPM FLOW
NODE TEMP
NO. (?F)
11 160.905
31 113.154
43 459.952
101 112.983
106 112.388
113 113.597
118 112.729
125 112.489
132 112.639
137 112.608
144 113.355
151 113.271
156 112.539
203 474.970
307 113.025
317 113.312
328 113.598
405 112.417
416 112.129
427 113.797
438 114.083
NODE TEMP
NO. (°F)
12 104.325
32 113.696
44 459.952
102 112.987
107 112.828
114 113.597
121 112.988
126 112.489
133 112.959
138 112.608
145 112.528
152 113.278
157 112.556
301 237.198
308 113.025
318 113.312
333 111.353
406 112.417
417 113.511
428 113.797
NODE TEMP
NO. (°F)
21 113.275
33 114.038
45 479.467
103 113.981
108 112.828
115 112.483
122 112.986
127 112.689
134 112.855
141 112.954
146 112.524
153 114.039
158 112.556
302 462.433
309 112.618
325 111.641
336 111.353
407 113.224
418 113.511
435 111.553
NODE TEMP
NO. (°F)
22 113.275
41 529.215
46 479.697
104 113.981
111 112.953
116 112.443
123 113.450
128 112.649
135 112.512
142 112.954
147 112.583
154 114.039
201 458.351
305 112.217
315 111.929
326 111.641
337 113.885
408 113.224
425 111.841
436 111.553
NODE TEMP
NO. (°F)
23 131.300
42 539.893
50 543.468
105 112.388
112 112.953
117 112.729
124 113.450
131 112.639
136 112.512
143 113.355
148 112.583
155 112.539
202 448.182
306 112.217
316 111.929
327 113.598
338 113.885
415 112.129
426 111.841
437 114.083
-------
TABLE IV (continued)
FPSSS TEMPERATURES FOR UNBALANCED 1 & 2 GPM FLOW
NODE TEMP
NO. (°F)
11 166.206
31 119.000
43 460.283
101 119.455
106 117.467
113 119.032.
118 118.987
125 118.712
132 117.845
137 117.794
144 119.583
151 119.815
156 117.697
203 475.329
307 118.279
317 118.562
328 120.609
405 118.673
416 117.201
427 119.041
438 121.560
NODE TEMP
NO. (°F)
12. 120.460
32 119.637
44 460.338
102 118.278
107 118.079
114 120.136
121 119.113
126 117.623
133 118.349
138 118.841
145 118.747
152 118.586
157 117.718
301 241.265
308 119.482
318 120.046
335 116.577
406 117.484
417 118.759
428 120.999
NODE TEMP
NO. (°F)
21 119.176
33 120.033
45 479.508
103 119,450
108 119.124
116 118.669
122 118.364
127 117.851
134 119.289
141 119.217
146 117,677
153 119.572
158 118;786
302 462.741
309 118.144
325 117.146
336 116.435
407 118.476
418 120.463
435 116.972
NODE TEMP
NO. (°F)
22 119.176
41 529.317
46 479.735
104 120.765
111 119.163
116 117.548
123 118.925
128 118.888
135 118,733
142 118.303
147 117.757
154 120.495
201 458.737
305 118.280
315 117.714
326 116.719
337 119.127
408 119.873
425 117.540
436 116.633
NODE TEMP
NO. (°F)
23 137.271
42 540.020
50 543.594
105 118.603
112 118.374
117 117.954
124 119.805
131 118.909
136 117.659
143 118.898
148 118.816
155 118.765
202 '448.726
306 117.287
316 117.004
327 118.845
338 121.171
415 118.107
426 116.917
437 119.323
-------
TABLE V
FPSSS THERMAL ANALYSIS RESULTS
(250°C Stack Gas/35°C Ambient Air)
Flow Condition
Balanced 2 GPM Flow
Balanced 1 GPM Flow
Unbalanced 1 & 2 GPM
Flow
Heat
Rejected
(BTU/HR)
2718
2603
2678
Coolant Loop
Temperature Rise(°F)
5/8 Loop
2.8
5.1*
5.2
6/7 Loop
2.8
5.14
3.0
Component Temperatures (UF)
PC Board
w/2 Watts
161
176
166
Laser
2/15 Watts
131
1U8
137
Window
2/15 -Watt HTR
529
530
529
Sensor Head
Side/Side AT^
0
0
1.3
TOP/BTM ATmx
1.0
1.6
1.3
00
-------
Table IV but no noticeable change in the parameters of Table V. Also, the model
was run for an elevated 500°C temperature. The result was predictable: near
doubling of the thermal loading and temperature elevations for the nodes listed.
The above information can be used to adequately specify the TCS requirements,
but it does not solve the problem of alignment. Rather, it simply provides the
inputs necessary to compute thermally induced alignment error. To test these
results an instrument "brassboard" was constructed from which computed distortions
(using temperature data) could be set and the misalignment measured.
The brassboard optical system (shown in Figure 38) used an F/1.8 55 mm objec-
tive lens operating at approximately 3X magnification. The secondary objective
operates at 6.7X, providing total magnification of 20X. The probe optical system
was laid out on a flat aluminum plate with adjustable mounts provided on critical
optical elements. These adjustments were not used in the final design; they were
provided in order to verify our calculations of alignment sensitivity to certain
distortions or displacements due to thermal changes in the structure supporting
the optical elements. For instance, axial displacement of the condensing mirror
will laterally displace the condensed laser beam at the object plane. Obviously,
the laser beam must be of sufficient width to allow for such displacements. By
theoretical considerations one finds that the error is:
Beam Displacement = R [Sin (1/2 9 - 9 ) - Sin 1/2,9]
m r
,. , a o. -1 [Sin 1/2 9 x (Axial Displacement]
where R = mirror radius and 8 = Sin c
m r R
m
For an axial displacement of 1 mm and 9 = 30°, the error is 0.19 mm or 190 microns.
The brassboard actually revealed a much lower sensitivity of 22 microns of lateral
displacement for each millimeter of axial mirror displacement. Of course, the
angle (9) on the brassboard is less than 10°. An axial displacement of 1 mm
would only occur if the head were constructed of aluminum and heated to 500°C.
At 250°C the error is only 11 microns which is quite small but for which our
design must account.
The displacements and distortions at all other optical mount points were
also measured. The x-y adjustments of the laser produce lateral displacement
errors of approximately 2.36 and 2.71 mm/mm, respectively. Although this is a
more sensitive region to misalignment it is not affected by thermal changes, and,
in any case, the laser was intended to be in the thermostated probe .region.
79
-------
CO
o
Figure 38. Brassboard
-------
The external mirror is the critical optical element since it is exposed to full
thermal extremes.
81
-------
6. EXPANDED DESCRIPTION OF FPSSS SYSTEMS
Our primary purpose in giving an overview of the FPSSS at the outset of
this report (see Section 3) was to offer the reader an early understanding of
the instrument to prepare him for the sections dealing with theoretical optical
and thermal modeling. These sections are more readily understood with the FPSSS
firmly in mind. In this section we discuss certain systems within the FPSSS in
detail, emphasizing the theory and functioning of the probe optics and elec-
tronics as well as the DAS architecture and computational software. First, a
detailed description of the electro-optical operation of the FPSSS probe optics.
6.1 The FPSSS Electro-Optical System
In a device with the anticipated performance of the FPSSS, a laser is a
required light source. It provides the greatest amount of diffraction-limited,
usable light energy expressed in both focusing capability and effective collima-
tion of all output energy. A He-Ne laser was the most practical, but the ques-
tions of size and model configuration hinged on other requirements besides
practicality.
As a result of theoretical scattering arguments (given in Section 3), we
desired to collect as close to the axis as possible in order to develop scat-
tering responses which would not be sensitive to refractive index changes. The
results of Section 4 showed that an inner collecting angle of 2° was adequate.
This inner collecting angle automatically specifies the permissible divergence
of the laser beam. The sample volume itself requires that the laser beam be
about 100 ym in dimension, a size requiring focusing since no laser produces a
beam that small on exit. For a TEM mode laser the focused spot size is given
approximately as fX/D, where f is the focal distance of the element used, X the
wavelength, and D the beam diameter at the focusing element. With a typical
laser beam of 1 mm dia. and X = 0.6328 ym, a focal distance of 150 mm is required
to generate a beam of 100 ym dimension. This, of course, is an impractically
large distance over which to maintain alignment. To reduce the focal distance
one generally must first expand the beam diameter with condensing optics before
focusing with a shorter focal length element. The TEM mode would allow for
collection down to a few tenths of a degree in the FPSSS with an illuminated
sample volume 100 ym wide. However, we previously showed that there was no
need for such a small inner collecting angle. This fundamentally allows one to
82
-------
consider multimode lasers that have higher rates of convergence or divergence.
Multimode lasers also have much higher powers, and, if of high order, a fairly
flat energy density profile. The governing equation for the focused spot size
for multimode lasers is CfX/D, where C is a constant which is closely approximated
by the degree of mode degeneracy ±1. For a fourth-order degeneracy and a beam
width of 1.5 mm (multimode lasers intrinsically have larger beams), a focused
beam of less than 100 ym requires a focal distance of one third that for the
TEM case. This type of multimode laser typically generates 2.5 times the
power of its TEM counterpart. It has the same input power for the same
length of discharge.
Thus, we selected a high-order multimode laser for the FPSSS. It gives a
greater amount of power than others over the same sample volume without conden-
sing optics. The optical system briefly described in Section 3.1 indicated an
output power of 5 mW. The device selected was a Coherent Radiation Model 5 T.
This tube has an aluminum jacket extrusion (which also serves as its cathode),
affording good thermal contact for heat dissipation. The laser output is steered
to a 75 mm focusing (condensing) mirror by the use of a pair of plane mirrors
(see Figure 4). Because the plane mirrors are buried in the probe head thermo-
stat, their alignment is not subject to thermal effects. The condensing mirror
focuses the directed laser beam down to approximately 100 ym at its 37.5 mm
focal point; it is the only element with power that is exposed to the ambient
high temperatures. The window is also exposed to high temperatures but it is a
piano-piano element. It is fitted with a beam dump spot or inner optical stop
on the optical axis which dissipates the laser beam.
The light collecting optics include a primary objective of 50 mm f.l.,
operating at a magnification of 2X, and a secondary microscope objective opera-
ting at 10X. The F/1.8 lens operating at these conjugates provides a maximum
aperture to permit an outer collecting scattering angle of 11°. Its performance
is quite good on axis although it has noticeable distortion in wide field appli-
cations. Because our interest was in objects ±50 ym from the optical axis it
proved more than adequate. The secondary objective is a standard 20X Nikon
microscope objective. Reducing its magnification by a factor of 2 results in
almost no change in its important primary conjugate (front object distance) and
83
-------
thus near optimum performance. Its back focal distance includes a 50 mm air
path and a 7 mm glass path through a beam splitter. The beam splitter has a 2:1
split of reflected and transmitted light.
The optical signals from particle images are simultaneously produced on
the two exit faces of the beam splitter and are analyzed and compared in magni-
tude by the probe electronics after they are detected by photodiodes and am-
plified by pre-amps housed in the probe head. The method by which these signals
can be processed to extract both size and positional information is a subject
of considerable interest and requires detailed explanation. Before elaborating
on this technique we should mention that all optical surfaces have coatings to
enhance optical performance. The mirrors have dielectric coatings delivering
better than 99% reflectivity. The window and beam splitter faces have dielectric
anti-reflection (AR) coatings with less than 1% loss. The photodiodes have a
dielectric AR coating on their windows as well as their substrates, effecting
less than 3% total loss. The objectives have broadband AR coatings and lose
15% each. Therefore, they account for nearly all of the approximately 35%
loss in the entire optical system.
The collecting optics essentially comprise a high resolution (although low
magnification) imaging system. It is a dark field situation; the particle
images resemble a moving star field. The beam splitter is aligned such that
all observed particle trajectories cross the photodiodes. Each photodiodes'
active area is 2500 ym in diameter so all particles at the object plane and in
the laser beem are observable (100 ym at 20X. = 2000 ym). However, the photodiode
observing the reflected image from the beam splitter is obscured along its di-
ameter by a mask 0.78 mm wide. Thus, a particle passing through the center of
the laser beam (and thus its image behind the mask) is only observed by the
photodetector with the clear aperture. This photodetector is used to provide
particle size information. The masked photodetector only receives signals for
particles off axis laterally but in focus when displaced axially, or from large
unfocused images that spill out behind the narrow mask. A sample volume of
prescribed dimension can be established by comparing the signals from the pair
of photodetectors and defining accept/reject criteria. Typically, we have re-
quired unmasked to masked signal ratios of 4:1 for acceptance. The sample volume
84
-------
is a direct function of this acceptance signal ratio. To derive the acceptance
signal ratio we actually provide a higher optical (through beam splitter reflec-
tance) or electronic gain to the masked detector and then compare for equality
of amplitude and refer to a gain ratio rather than signal ratio. The following
discussion further details the actual situation in the FPSSS.
The relative size of the FPSSS sample volume cross section with respect
to the laser beam is depicted in Figure 39. The sample volume includes only
the region near the center of the laser beam. The sample volume cross section
is noticeably diamond shaped. The diamond shaped sample cross section results
from the accept/reject criteria used in comparing signals at the masked and
unmasked detectors. The center of this diamond cross section concides with
the object plane of the collecting optics. The points of this diamond cross
section define the limiting depth-of-field of the FPSSS. Both the width of
the cross section and the depth-of-field vary inversely with the magnification
used in the collecting optics.
Shown in Figure 40 are the size and positions of images formed by particles
at various positions in the illuminated volume. It is important to understand
that for small particles (essentially point objects) the image size is only a
function of its axial displacement from the object plane.* The image size is
linearly related to the numerical aperture of the collecting optics and is given
approximately by:
Image size = N.A. x displacement from object plane.
It is apparent that only images that are near the object plane and the center of
the sample volume form images with light concentrated on the opaque mask, as
illustrated in Figure 40. Thus, such images transmit little signals through the
masked aperture. Particles whose pulse amplitudes, as seen by the masked aper-
ture detector; are greater than, those seen by the signal aperture detector are
rejected. This defines the diamond shaped sample cross section and image re-
*Larger particles will generate image sizes that are size-dependent. At
a 2X gain ratio there is never any sample volume size dependency. At higher
gain ratios there is a real effect but it is not as large as what one might first
suppose because of the localization of the light scattering energy in the center
of the images of large particles.
85
-------
RELATIVE SIZE AND POSITION OF
SAMPLE VOLUME CROSS SECTION
OBJECT PLANE
1/E2 INTENSITY
(BEAM EDGE)
oo
o\
SAMPLE
VOLUME
WIDTH
MULTIMODE
LASER
BEAM
SAMPLE VOLUME
CROSS SECTION
MAXIMUM VARIATION
IN INTENSITY OVER
SAMPLE VOLUME
HIGH ORDER
MULTIMODE
INTENSITY
DISTRIBUTION
Figure 39. In the FPSSS, the beam width at the object
plane is approximately 100 microns wide.
-------
IMAGE SIZES AND POSITIONS ON MASKED
APERTURE DETECTOR
"
0.78 mm
k- 2.50 mm
MASKED APERTURE
I I )\ APERTURE AT
NIX \ PRISM FACE
APERTURE AT
PRISM FACE
SIGNAL APERTURE
oo
IMAGE
PLANE
TO OBJECT PLANE
50% BEAM SPLITTER
IMAGE PLANE
Particle Image at end of
Depth-Of-Field at center
of Sample Volume Width
Particle Image within
Depth-Of-Field and within
Sample Volume Width
Particle Image near
Object Plane and center
of Sample Volume Width
Figure 40
Particle Image at end of
Depth-Of-Field but
outside Sample Volume
Width
Particle Image within
Depth-Of-Field but
outside Sample Volume
Width
Particle Image near
Object Plane but
outside Sample Volume
Width
-------
lationship shown in Figure 41. We use a gain ratio of masked aperture detector
to signal aperture detector of 4X for the best noise immunity. The sample cross
section area is the product of the sample volume width and one-half the depth-
of-field.
One final aspect of the sample volume needs explanation. As we have defined
it, the sample volume is only the region of accepted measurements. The total
view volume includes a much larger volume where both accepted and rejected
particles are viewed. This view volume is as much as ten times larger than the
sample volume and must be used for coincidence error estimates. The ratio of
sample volume to view volume is a function of the inner numerical aperture (N.A.)
of the collecting solid angle (2°, N.A. - 0.03). This presents an obvious trade-
off between a design to collect very strong forward scattering and one that
optimizes measurements at the highest number densities.
Because of the near proximity of the laser beam and the inner collecting
angle, we had to map the total view volume experimentally. In general, the view
—f\ ^
volume is '\> 5 x 10 cm . It is this volume that establishes coincidence prob-
ability but not coincidence error. Coincidence errors in such a system are
dominated by particles within the view volume but not the sample volume. Be-
cause of the large 4X gain ratio, any time one of the particles of a coincident
pair is outside the sample volume the pair is most likely rejected. The only
time it would not be rejected is when the particle in the sample volume is
large enough to generate more than four times as much signal; in such a case it
would be better to accept the measurement since the error would be negligible
in size (typically less than 10% since scattering signal is proportional to
2 3
N to N ). In this type of system coincidence generally results in a loss in
counted particles but negligible sizing error. The coincidence loss is corrected
for by monitoring the total activity of the probe electronics (activity is the
percentage of time pulses that are present and reflects and effective duty cycle),
6.2 Electronic Subsystems
With the exception of the photodetector module, all of the FPSSS electronics
are contained within the electronics console. A block diagram of the electronics
console is shown in Figure 42. For a detailed description of actual circuit
operation the reader is referred to the FPSSS manual. For descriptive purposes
-------
IMAGE SIZE AS A FUNCTION OF SAMPLE
VOLUME POSITION
FOR THE CASE OF 4X GAIN RATIO OF
MASKED-TO-SIGNAL APERTURE
OBJECT PLANE
MASKED APERTURE
SLIT WIDTH
T
LIMIT OF DEPTH-OF-FIELD
Figure 41
-------
FPSSS ELECTRONICS CONSOLE BLOCK DIAGRAM
r '
SCATTERING
SIGNALS
LASER
REFERENCED
SIGNAL _
VD
O PROBE
HEAD
SIZE
RANGE
••
r*
L
SIGNAL » SAMPLE
CONDITIONING DEF1NmON
CIRCUrTHV * CIRCUITRY
1
PULSE HEIGHT CONTROL
ANALYZER LOGIC
TRANSIT TIME
CIRCUITRY
1 1 1
SIZE ACCEPT/REJECT TRANSIT TIME
mi I I J
DIRECT MEMORY ACCESS LOGIC
NUMERIC THUMBWHEEL
DISPLAY SWITCHES FRONJ pANEL DATA
t
1
' ^'
DATA PORTS TIMERS | DATA PORT
1 til
2-80 DATA BUS
V ' '
S'2E0RRATNGE Z-80 CPU
1
I
ROM DATA PROCESSING RAM CONTROLLER
PROGRAM DATA
SAMPLE DURATION - LOGIC STORAGE
TIME OF DAY CLOCK i'
" PRINTFR PI OTTCU
LASER POWER SUPPLY
15 5
CRT SWEEP AND
CONTROL CIRCUITRY
-
t
CFRONT PANEL
CONTROLS
CRT HV POWER SUPPLY
Figure 42
-------
here, we can only give a walk-through of signal processing and systems archi-
tecture.
The scattering photodetectors and pre-amplifiers are housed on a single
photodetector module circuit board inside the FPSSS probe head. A separate
photodiode with pre-amplifier attaches to the rear of the laser as a reference.
A block diagram of the "probe" electronics is shown in Figure 43a. The photo-
detector module is identified by the dashed section.
The three-stage amplifier sections have a minimum bandwidth of 300 KHz.
The second stage is a programmable amplifier used to gain-switch between size
ranges. The two diodes with their pre-amps provide the signals that make the
FPSSS capable of distinguishing whether a particle is within the proper sample
volume and rejecting particles that are not. The reference detector provides
a reference signal proportional to the power output of the laser. Changes in
laser output are thus cancelled when the signal is applied as the reference
input to the pulse height analyzer.
The remainder of the FPSSS electronics is contained in the electronics
console. The two signal pulse streams are first properly leveled with baseline
restoration and DC offset circuitry. Dual peak readers then momentarily store
the pulse heights from both the signal and the masked aperture detectors. The
peak readers are used to provide inputs to a comparator (sample volume validation
comparator) to determine which of the two pulse heights from the pre-amps is
the greatest. This measurement determines when particles are in the accepted
sample volume.
A pulse height analyzer sizes the signal pulses of all particles passing
through the beam. The pulse height analyzer has a set of 16 voltage comparators
and latches. The reference voltage to the comparators provided by the reference
detector is divided by resistive dividers. A separate group of resistive dividers
is associated with each size range. Because its reference voltage is derived
from the source of illumination, the entire system has an effective automatic
gain control (AGC). The output of the pulse height analyzer is encoded in 4
binary weighted address lines that access memory via DMA (see Figure 43a).
A separate set of voltage windows selects pulses of a narrow range of
amplitudes within the transit time analyzer. These same pulses are also delayed
91
-------
FPSSS PROBE ELECTRONICS BLOCK DIAGRAM
VO
K>
MASKED (!)
APERTURE
DETECTOR
SIGNAL
APERTURE
DETECTOR
BIAS
L.
1 ,
1 ,
DC
%
o
SELINE RES1
<
ffl
SAMPLE
VOLUME
VALIDATION
COMPARATOR
DELAY
LINE
TRANSIT
TIME
ANALYZER
PULSE
HEIGHT
ANALYZER
MEAN VELOCITY
ACCEPT
REJECT
LOGIC
ACTIVITY
GATED STROBE
BIT
LINES
TO DAS
via DMA
Figure 43a
-------
through a delay line. The transit time analyzer measures the half-width of the
delayed pulses that are found to be in the sample volume. The transit times
are averaged over a great number of particles to provide a true statistical
velocity average. (Note that velocity is actually the reciprocal of transit
time.)
The control (accept/reject) logic provides the proper delay between an
event and the strobe and reset pulses. This logic also provides gating for
valid and invalid strobes. Circuitry within this section also computes the
percentage of time that the pulse processing electronics are busy. This
activity signal (A) is directly related to the coincidence loss correction as
follows:
Nm= N
T meas. ' 1-.006A ,
where N is total number and N is the measured number. Activity, in per-
T meas. j > r
cent, is in the form of a 0-100 Hz frequency.
The pulse height analyzer is configured to provide four size ranges of 15
channels each, as follows:
Range 1 0.4 - 1.15 ym (0.05 ym intervals)
Range 2 1.15 - 5.65 ym (0.3 ym intervals)
Range 3 0.5 - 2.0 ym (0.1 ym intervals)
Range 4 2.0 - 11.0 ym (0.6 ym intervals)
All size intervals are linear; the maximum resolution in range 3 is 0.03 ym.
As mentioned in Section 3, ordinarily ranges 1 and 2 or 3 and 4 are paired for
routine data gathering. The ability to switch ranges has been found to be par-
ticularly important in field measurements. As previously noted, if you had un-
limited size sensitivity and a wide dynamic range (e.g., that produced by
logarithmic amplifiers) you would inevitably find a small enough size to satu-
rate the instrument through coincidence error. The ability to range-switch
with desired levels of minimum size sensitivity provides a method of approaching
conditions limited by coincidence error without jeopardizing measurements at
larger sizes which in themselves would present no coincidence error problems.
Logarithmic amplifiers also pose other difficulties compared to range-switching
linear amplifiers. The primary difficulty in the FPSSS arises from the necessity
for excellent matching of the two signal trains. Due to lack of control over
the necessary amplifier parameters as well as variable baselines (caused by
93
-------
changes in background light levels at each photodetector), logarithmic amplifi-
cation suffers further.
The maximum particle number density that can be measured by the probe
electronics depends to a certain degree on the size distribution and the par-
ticular size range in question. Although the electronics do not process particle
sizes smaller than the lower limit of the size range in use, they do process
particles that are larger than the upper limit. Thus, it is possible to have a
fairly low apparent number density on a certain size range yet very high
activity. Correction through the relationship given allows retrieval of the
proper number density. For the typical measurement environment presented by
4 -3
stack work, the FPSSS can perform at number densities above 2 x 10 cm . We
4 -3
observed measured number densities as high as 5 x 10 cm in field testing at
precipitator inlets (see Section 8).
The electronic circuits involved in the FPSSS size measurements are con-
servatively designed. Components within the probe head are MIL STD parts and
the remainder are largely commercial grade. Circuit stability involving such
parameters as amplifier offsets, voltage dividers, and bias levels generally
exceeds requirements. The limitations on sizing performance are essentially
determined by the optics and laser light source. The mode stability of the laser
probably limits overall sizing accuracy and sample volume definition to 90-95%
(5-10% error).
The FPSSS data processor is a Z-80 based microcomputer. Up to 16K bytes
of ROM and 4K bytes of RAM are available. All I/O except for DMA activity
(probe and printer) is implemented with Z-80 PIO circuits. The time-of-day
clock and other timing functions are implemented with a Z-80 counter-timer
chip, or CTC. Particle accumulations and plotting-printing raster drive are
implemented with a single DMA circuit. Activity is started by generating an
interrupt with each positive transition of the activity line and driving in-
ternal counters. The CRT is refreshed from a separate 256 X 8 bit dual port
memory; this memory appears as write-only memory in parallel with a 256 byte
segment of RAM. The DAS is shown expanded in Figure 43B.
Figures 43a and 43b show the logical interface between "probe" circuitry
and "data processing" circuity. Two lines from the processor define the
94
-------
FPSSS DATA ACQUISITION SYSTEM BLOCK DIAGRAM
NEW PARTICLE SIZE AND VELOCITY
FPSSS PROBE .
ELECTRONICS
CE.T. SLCT^
(^MP.SLCT
ELAPSED
TIME
LOGIC
MAX POP
LOGIC
INT
^
1
2-80 CPU
ROM DATA
PROCESSING
PROGRAM
( TICKING )
(FOCUS, CENTER)
AXIOM
PRWTER/PLOTTER
PAPER ADVANCE)
( PRESET HOURS ) ( PRESET MHUTES )
Figure 43b
-------
range in use. Four lines and a strobe mark the detection of a valid particle
and the bin into which its size falls. The 0-100 Hz activity line encodes
0-100% activity. A single status signal, WARNING, is asserted whenever the
laser power output goes below a certain level. Finally, 9 lines encode the
average particle velocity; the processor asserts an inhibit signal to prevent
the probe from changing the velocity lines while they are being read.
The firmware resides in 8 integrated circuits which are 2K X 8 bit UV
erasable field-programmable, read-only memories.* (Consecutive memory chips
represent consecutive ascending segments of address space.) RAM is implemented
with 4K X 1 bit dynamic devices. The Z-80's built-in refresh feature is used
to refresh the RAM.
Z-80 code is written entirely in Z-80 assembly language. The program is
currently written in 14 relocatable segments and linked with a relocatable
loader to form a contiguous memory image. A MOSTEK SDB-80 with extended memory
is required to assemble and link the program. In addition, the program makes
heavy use of a floating point package that resides in a fixed area of memory.
When the FPSSS is powered up, an RC circuit on the CPU reset line causes it
to be asserted for about 0.2 second. When it terminates, the CPU begins exe-
cuting instructions at memory location 0. The reset also causes initialization
of various other latches in the hardware. The program executes an initiali--^
zation of various other latches in the hardware. The program executes an
initialization phase. Memory is cleared, PIO operating modes established, and
the default parameter set (a set of operating conditions and output programs
initiated if not set before sampling is begun) is installed in RAM. The main
loop consists of two phases: acquiring data and printing data just acquired.
There is a "repeat" mode that is a variation of this basic cycle. When repeat-
ing, the data just acquired are moved into one of seven auxiliary buffers. When
all samples have been acquired, each data set is moved back into the primary data
buffer to be printed.
When acquiring data, first in the odd range and then in the even, the pro-
*The term "firmware" (as opposed to "software") is often used herein to
illustrate memory involatility.
96
-------
gram continually tests the elapsed time counters (interrupt driven), total
population (DMA driven), and the operator pushbutton (interrupt driven) to see
if the end of the range interval has expired. If so, the program moves to the
print phase. The diagnostic modes are another variation on the basic cycle
in which only one specified range is sampled (rather than a pair) and the print
phase is skipped.
A switch allows one to temporarily suspend data accumulation. It is sampled
in the clock interrupt service routine, which in turn sets a bit in a PIO that
disables/enables DMA-driven particle accumulations. The same status bit also
controls accumulation of elapsed time and activity counting.
The print phase begins by examining the most significant output program
switch, which specifies one of sixteen output programs. The ABORT switch allows
one to bypass or suspend printer-plotter operation. It is checked each time a
print line is to be transferred to paper, and the transfer is skipped if the
switch is set. If the ABORT switch is clear, the HOLD switch is checked; if set,
the program waits until it clears, suspending printout. The output program
switches are checked again when the print is complete; if they have changed,
the print is repeated under the new switch setting. In addition, the ABORT
switch is checked at the beginning and end of the print; if set at the beginning
and clear at the end, the program is repeated.
The possible data sampling and output modes are too numerous to mention
and are rather flexible. Some output programs take several minutes to print
out. You can also process and print the same data set with more than one out-
put program. Scale sensitivity can also be changed if initially inadequate.
Most of these capabilities will be evident in the data samples presented in
Sections 7 and 8.
6.3 Basic Data Processing Equations
This section provides an overview of the basic mathematical manipulations
performed by the DAS electronics in the FPSSS. The FPSSS can size particles
in four sets of size ranges numbered 1, 2, 3, and 4. Only one range can be
in use at any one time. In a normal sampling operation, ranges 1 and 2 or 3
and 4 are automatically paired together giving "small" and "large" size ranges
of 0.4-5.65 and 0.5-11.0 ym, respectively. The operator selects either "small"
97
-------
or "large" size range mode from the front panel, which evenly time shares the
sampling time between the respective range pairs. Each range has 15 size
classes or channels. Since two ranges are used alternatively, it is convenient
to consider the probe as having 30 "bins". In the small mode, bins 1-15 corres-
pond to range 1, channels 1-15; bins 17-31 correspond to range 2, channels 1-15.*
In the large mode, bins 1-15 correspond to range 3, channels 1-15; bins 17-31
to range 4, channels 1-15. The bin is represented by the index "i".
Associated with each channel of each range, or each bin, is a minimum,
mean, and maximum (optical) diameter of particles sized into that bin. Every
channel of a given range has the same bin width or size interval. The maximum
diameter of bin i is the same as the minimum diameter of bin i + 1. The mean di-
ameter is the arithmetic average of the minimum and maximum diameters.
If DL. represents the minimum, or lower, diameter of bin i, D. represents
the mean diameter of bin i, DU. represents the maximum, or upper, diameter of
bin i, and BW. represents the bin width of bin i, then
BW1_15 = 0.05 ym
BW17-31 = 0.3 urn
DLi =0.4 urn
_TT •, •, r if in the small mode
DU15 = 1.15 ym
DL17 = 1.15 ym
DU3i = 5.65 ym
and
BWi-is =0.1 ym
BWi7_3i = 0.6 ym
DLi =0.5 ym
DU15 = 2.0 ym if ln the large mode'
DL17 =2.0 ym
DU3i =11.0 ym
Particles are accumulated first in range 1 (or 3 if in the large mode) for
a length of time, then in range 2 (or 4) for another (usually equal, possibly
different) length of time.
*Bins 0 and 16 are reserved for time and. housekeeping.
98
-------
is corresponds to the range 1 time interval .,-.._•, -,-, ,
-13 v & if in the small mode;
ETi?_3i corresponds to the range 2 time interval
ETi-is corresponds to the range 3 time interval ±f ±Q fche
ETi7-3i corresponds to the range 4 time interval
N. = number of particles accumulated in bin i during the interval of
elapsed time ET..
The probe has a fixed sample area, SA, which is normal to the direction of
particle flow. The particles have a velocity, V- with respect to the probe.
Therefore, each range accumulation interval has an associated sample volume.
SV. = ET. • SA • V
The number density, or concentration, of particles in bin i is
N.
C. =
i SV.
i
Similarly, the mass density, M., particles in bin i is
N. • IT/6 • D. • p
M = —
i SV.
i
where p is the density of the particles.
It is convenient to normalize number and mass densities by their bin widths,
to arrive at number density per unit diameter, C./BW., and mass density per unit
diameter, M./BW..
The total number of particles observed during the accumulation interval is
15 31
N = £ N. + L N.
1-1 X i-17 X
Similarly, we have total concentration,
15 31
C = Y, C. + I C.
1-1 " i-17 X
and total mass,
15 31
M = E M. + Z M..
99
-------
To simplify the nomenclature, introduce an index j, and a bin-to-index
assignment b., such that,
bi-is = 1-15
bie_30 = 17-31
Having made this definition, we could restate the definition for total
mass, for example, as
3 0
M = z . v-
,7=1 J
The mass median diameter HMD is that diameter at which half the mass re-
sides at smaller sizes and half at larger sizes. Let:
k
M,
3
M,1 = E M,
k . b .
k-1
M,1 , = £ M,
k-1 . , b .
J=l 3
where k is selected so M,1 < 0.5 • M, M^ > 0.5 • M. Then the median is computed
AC— 1 AC
from a linear interpolation over the interval k-1 to k. That is:
(DU, - DU, ) • (0.5 • M - M,1 ,)
If k-l
MMD = DU, + —
bk-l (M1 - M£_ )
Note that the upper limit of diameter is used here rather than the mean diameter
since it reflects all the mass in the bin rather than only a portion of it.
The opacity (OPA) is a measure of the percent loss of incident collimated
light by particles distributed uniformly throughout a volume of specified
thickness (T) which in our case is the diameter of the stack at the point where
the measurement is made. The light source should ideally be taken as the inte-
gral over a source spectrum typical of opacity meters, but for the purpose of
simplifying computations, it is taken here to be monochromatic at X = 0.55 \m,
near the center of the visible spectrum. The extinction loss occurs due to
scattering.
100
-------
Thus,
GAP = 100 x [l-I/I ]
where I/I = transmission ratio.
by:
From Beer's law I/I = e , where 6 = total extinction coefficient given
15 31
5 = E 6. + I 6. , where
1=1 1 o=17 X
6± = C± • [TT • (D±/2)2] .
The extinction efficiency Q. may be approximated as
* = 2 - 4/p± sin p± + 4/p± (1 - cos p±) ,
2 D . (m-n )
i re
where i = - 7 - and n = the real index of refraction of the particles
A re ^
taken to be the value for glass (1.51). The absorption component is ignored
here for simplicity.
Finally, the FPSSS can display spectral data against either optical (physical)
diameters or aerodynamic diameters. Ignoring the slip correction factor, the
aerodynamic diameter D of a particle of optical diameter D is a function of the
3.
density (mass per unit volume) of the particle:
D = D v/p/p _ .
a a water
101
-------
7. CALIBRATION AND LABORATORY TEST AND SIMULATION OF FLY ASH ENVIRONMENTS
As we noted in Section 4 it is extremely difficult to adequately model an
instrument's response characteristics from theory alone. In particular, the
behavior of photodetectors and electronics in the signal train can modify an
anticipated optical signal in an unexpected way. Thus, the FPSSS was thoroughly
calibrated empirically with an assortment of monodispersed spheres of known
size and refractive index. This provides a primary calibration curve for the
instrument through primary particle standards. A second form of laboratory
evaluation was the intercomparison of the FPSSS with other PMS particle size
spectrometers serving as secondary standards. These intercomparisons are with
particles of random morphology and shape (often ambient aerosol) and provide
broad size spectra. The FPSSS velocimeter calibration involved wind tunnel
tests at uniform flow velocities to establish a relationship between particle
transit time and tunnel velocity.
All of the above calibrations and tests were conducted at PMS. However -
a series of tests was also conducted at the EPA test facilities at IERL. These
included a substantial number of tests in the wind tunnel test facility using
known dispersal rates of collected fly ash and intercomparison with impactors.
Tests of the velocimeter were also performed along with sensitivity checks of
the FPSSS to flow alignment. A second set of measurements was made in a separate
hot flow facility incorporating a precipitator. A final set of measurements was
performed in the flue of an oil-fired combustion unit.
7.1 Calibration and Evaluation of the FPSSS at PMS
The primary calibration curve for the FPSSS was established using moni-
dispersed latex and glass spheres. Through prior experience we have found that
only a few points are generally required, although it is a fairly simple test
and many different sizes can be run quickly. Most of the sizes used are in the
submicron region since the signal becomes a stronger function of size the smaller
the particle. Polystyrene latex (PSL) or polyvinyl toluene (PVT) spheres are
available from 0.01 to several microns and have similar optical properties
(nearly identical refractive indices); they are generally referred to simply
as latex spheres. For the FPSSS, latex spheres were used to cover sizes from
0.312-2.02 urn. Glass beads having modal diameters of 5.5 and 11.5 ym established
the upper end. The glass bead distributions are fairly broad, with standard
102
-------
deviations of several microns as compared to the few hundredths of a micron
generally quoted for latex, but their modal size is sharply defined.
The results of FPSSS sizing with these standard spheres are given in
Figure 44. Theoretical response curves for latex (n = 1.58-.0i) and crown
glass (n - 1.51-.0i) are also shown for reference. Inspection clearly shows
relatively good agreement up to several microns, but a reduction in scattering
signals at larger size. This is an expected result caused by the fact that
large particles scatter more strongly at directions closer to the optical axis
while optical transmission is greater at larger angles. The overall calibration
must thus be adjusted systematically, giving the modified error envelope and
centerline (adopted) calibration shown. This adopted calibration curve is then
normalized to the measured reference signal and the individual gains for each
range established to provide best overall signal-to-noise ratio and coverage
of the size range selected. Each size range is divided into 15 equal width
size classes with their normalized signal windows determined directly from the
primary calibration curve.
The calibration just described is generally only necessary for each type
of instrument, and identical copies of an instrument will calibrate without
modification of the primary curve. However, sizing calibration is only one as-
pect of total calibration. It is also useful to determine whether an instrument
counts the correct number of particles of each size which in effect is a check
on how well the sample volume is defined. For this purpose one can choose to
map the sample volume with a precisely controlled particle source stream or make
comparisons with other secondary standard instruments. Both kinds of tests
were performed for the FPSSS. Figure 45 shows the relative counting rate of the
FPSSS observed by moving a hypodermic needle through which a microstream of
latex particles is injected along the depth-of-field of the FPSSS sample volume.
It clearly defines the diamond shaped sample volume with maximum counting rate
at the object plane and minimum at the ends of the depth-of-field. This type
of experiment is not sufficiently controllable to afford an accurate calibration,
but it is a particularly nice demonstration of the technique's mode of operation
and answers questions of size dependency of the rejection processes used to de-
fine sample volume.
103
-------
10
UJ
(A
o
a
UJ
cc
o
E
UJ
o
en
UJ
UJ
cc
1.51 -.01
1.58 - .Oi
FPSSS CALIBRATION POINTS
10
0.1
100
DIAMETER (/J.m)
Figure 44. In addition to the use of latex and crown
glass spheres, particles of other refractive indices
were generated with a vibrating orifice generator; how-
ever, these are difficult to use as primary calibration
points.
104
-------
QC
Ul
m
5
800 -
600 -
400 -
200 -
-400
-200
200
400
DISTANCE (/im)
Figure 45. The above sample volume map was constructed using a 28 gauge
hypodermic needle, which generates an aerosol stream approximately 75 microns
wide. The needle was moved at 25 um distance intervals starting outside the
depth-of-field and continuing until beyond the depth-of-field.
-------
In any multirange device such as the FPSSS, a certain amount of overlap
exists between size ranges. In the FPSSS this overlap is nearly 50% and allows
a check to be made on its internal consistency. Figure 46 is a plot of size
distribution data obtained during a sample run using glycol aerosol at high
number density (worst case condition). Two spectral peaks are sharply defined
using both the small and the large size modes. The effect of lower resolution
in the 0.5-11 Vim range is evident in the smoothing of spectral features at
submicron sizes.
Probably the simplest means of evaluating an instrument's performance is
through intercomparison with other instruments that have been sufficiently
calibrated and characterized to serve as laboratory standard instruments. At
PMS one instrument in each of the standard product lines is maintained as a
relative standard. For comparison with the FPSSS the PMS Model CSAS-100 was
selected. The particle sources used included natural ambient aerosol and atomized
glycerin. The instruments were configured in such a way that the aerosol flowed
through the sample sections of both instruments in series. This dismisses any
questions about differing flow rates. A uniform velocity of about 6 msec
was used. The results are shown in Figures 47, 48, and 49. The first two
figures are in ambient aerosol with fairly low number densities revealing a
rather steep distribution. Such a distribution is a good test of systematic
sizing error which is typically observed as a lateral shift. Figure 47 covers
entirely the smallest submicron particles of interest (FPSSS range 1), while
Figure 48 includes sizes up to 2 microns (FPSSS range 2). Figure 49 was obtained
using atomized glycerin, providing coverage of sizes from 2-11 ym, which cor-
responds to FPSSS range 4. Range 3 of the FPSSS is entirely overlapped by ranges
2 and 4 and its consistency was checked relative to measurements in them.
These results are probably as good as can be expected from a pair of dif-
ferent instruments. In fact, the FPSSS and the CSAS-100 differ considerably
in some respects—the most apparent being the 2-11° versus 4-22° collecting
angles. The magnification in the FPSSS is double that of the CSAS-100, resulting
in a much smaller sample volume, yet the comparisons stand on their own merit.
Finally, the FPSSS was wind tunnel tested to calibrate the velocimeter.
These tests were repeated at several different periods in the contract to try
106
-------
FPSSS INTERRANGE COMPARISON
1000^
Range 3 Range 4
100-
E
3.
rt
i
o
10-
Range 1 Range 2
• 0.5-11.0 ym
• 0.4-5.65 ym
0.1 1.0 10 10
DIAMETER (^m)
Figure 46
137
-------
FPSSS/CSAS COMPARISON
(SUBMICRON PARTICLES)
100
10-
A FPSSS
- CSAS
o
Z
•*,
1 -
0.1.
0.1
• A
AA
A
A
i—r
1
—i—
10
-i—r
100
DIAMETER (Mm)
Figure 47
108
-------
FPSSS/CSAS COMPARISON
(MEDIUM SIZED PARTICLES)
100
10-
A FPSSS
- CSAS
i
u
1 -
-i 1—i—i 1 1 \—r
10 100
DIAMETER
Figure 48
109
-------
1000-
FPSSS/CSAS COMPARISON
(LARGE PARTICLES)
100-.
E
a.
n
i
o
A
A
A FPSSS
- CSAS
A
A
A
10-
1.0
1.0
i i i r i i i i
10 100
DIAMETER
i r
1000
Figure 49
110
-------
to refine the measurement. In early tests the calibration was found to be size-
dependent and we found it necessary to select a fairly narrow range of sizes
for pulse width measurements to avoid amplitude (and thus size spectral) dependency
During early field studies the velocimeter always read about 50% high, and inde-
pendent velocity measurements were utilized. The velocimeter, of course, is not
as accurate as standard pitot measurement techniques but because the measurement
is made at exactly the point of interest it can be extremely useful even if of
reduced intrinsic accuracy. For this reason a considerable effort was made to
refine this measurement as the FPSSS developed. In its final delivered form,
the velocimeter calibration appeared to be within ± 10-15% over a range of 3-30
msec . Below 3 msec it is not reliable due to triggering of multiple par-
ticle transit pulses for each particle vent. Possible improvements to the
velocimeter are suggested in Section 9.*
7.2 IERL Test Summary
During the two week period from February 6 to February 16, eight days of
intercomparisons were performed with the FPSSS and various EPA samplers at IERL.
Facilities exist at IERL to test various particle sizing and sampling instru-
ments under simulated in-stack conditions. Of primary importance to this pro-
gram was the IERL wind tunnel, which, while operating at room temperature, was
useful in characterizing the FPSSS using a variety of fly ash collections. In
addition to measurements of the size distribution and the ability to correlate
with other measurement systems, the tunnel was useful to study effects of
orientation and rates of contamination and to observe its internal velocimeter
under a wide range of velocities and mass loadings. A diagram of this facility
is given in Figure 50.
The second IERL facility in which tests were made was a heated wind tunnel
that had an operating electrostatic precipitator. Although this particular test
facility was not capable of generating the range of velocities or the variety
of test conditions possible in the primary facility, it did offer the added
dimension of elevated temperature.
*Since completion of the prototype three additional FPSSS instruments have
been delivered with improved velocimeters (see Section 9).
Ill
-------
SCHEMATIC OF IERL WIND TUNNEL
SHOWING DUST HANDLING EQUIPMENT
DUST INJECTION APPARATUS
FLOW
TEST SECTION
Figure 50
-------
The third and final IERL test facility used was a fuel oil boiler. Meas-
urements were performed in the flue of the boiler while varying the fuel load,
excess air, and the amount of water injection.
Four types of particle environments were sampled using the three above
test facilities:
1. In the first series of tests, the IERL wind tunnel operated at
velocities from 2 to 30 msec and fly ash loadings from 0.06 to
_3
3 gm . The tunnel was operated at room temperature and with
room air. The ash had been collected from a coal burning utility
and redispersed into the tunnel air flow using a feedscrew, blower,
and vibrating lead shot fluidizing bed (see Figure 51). Water was
added to the feed system to minimize loss or agglomeration of ash
due to static attraction. A few Climet Particle Counter observa-
tions, 13 impactor runs, 12 Coulter Counter analyses, and on the
order of 100 FPSSS observations (resulting in 200 computed number
density and mass spectra and 100 numerical summaries) were made.
These tests included characterization of the velocimeter and tests
of the FPSSS sensitivity to flow alignment.
2. The same wind tunnel was also tested with dust collected from a
steel mill whose primary constituent was iron oxide. A limited
_3
range of mass loadings (0.06 to 0.12 gm ) was used due to diffi-
culties in feeding the material at higher rates and uncertainties
in the tunnel airspeed at low airspeeds. Three impactor runs, 8
Coulter analyses, and about 15 FPSSS observations were made.
3. The third kind of test involved the heated wind tunnel with rec-
tangular cross section and incorporated an electrostatic precipitator
(ESP). Three hours of FPSSS observations at the inlet to the pre-
cipitator were made. The tunnel was loaded with 1 gm of ash at
a velocity of 1 msec airspeed. Feed system characteristics were
largely unknown to us, but appeared to be considerably more rudi-
mentary than the tunnel used above, and indications of uneven mass
loading in the tunnel's cross section were noted.
113
-------
IERL SCREW FEEDER
IERL FLUIDIZED BED AEROSOL GENERATOR
DRIVE MOTOR
ROTATING SCREW
DUST /
A— VIBRATING HOPPER
FEEDBACK
SIGNAL FOR
CONTROLUNG
SCREW SPEED
HIGH PRESSURE AIR
AEROSOL DENSITY MONITOR
DIUTER
FLUIDIZED DUST
POROUS PLATE
HIGH PRESSURE AIR
FLOW REGULATOR
AEROSOL OUTPUT
ASPIRATOR
Figure 51. The two aerosol feed systems depicted above can be used
independently or jointly to achieve the desired aerosol dispersal
characteristics.
-------
4. In the final test series, an attempt was made to characterize a 38 cm
diameter #6 fuel oil boiler flue at 230°C. Particulate loading versus
water injection and percent excess air relationships was examined.
About 5 hours of FPSSS observations were made; no other reference in-
formation or other instruments were available for simultaneous inter-
comparison. Impactor runs at presumably similar boiler conditions
made prior to this experiment were available for comparison.
A series of 15 specific test runs were performed in the IERL wind tunnel.
Twelve of these tests were performed using fly ash from a coal-fired power plant.
The last three tests used an ash that was very high in iron oxide. The normal
fly ash was collected from the Duke Power Station electrostatic precipitator
plants. The collection procedure for the iron oxide rich ash was not known.
Figure 52 shows SEM photomicrographs of material from these two sources. An
obvious problem with the iron oxide is the tendency for the material to clump
into aggregates, which the results show remain after fluidizing procedures.
Difficulties with the friability of both materials require humidity control to
reduce static attraction. Coulter Counter analysis of both of these materials
is given in Figures 53 and 54 along with relative comparisons of measurements
using the FPSSS. In Figure 53 the Coulter analysis of the fly ash appears to
be in excellent agreement with the FPSSS, especially when the 100 ym orifice was
used; however, the results with the iron oxide (shown in Figure 54) showed dis-
agreements we have attributed to clumping.
For the Duke Power Plant fly ash, the initial run with the Coulter Counter
was made with a 200 micron orifice to make certain that all particles up to
approximately 128 microns were measured. Since the FPSSS has a limited range
extending only up to 11 microns, a sub-sample was taken and sieved to remove
all particles larger than 44 microns and that sample analyzed with the Coulter
Counter using a 100 micron orifice. These results are all presented in Figure
53. The set of distributions is obviously quite different for the Coulter
analysis using the different orifices. The median volume diameter computer
using the 200 micron orifice and the Duke fly ash was 14.5 microns, while
that for the 100 micron orifice is approximately 8 microns. In the relative
comparison shown, we truncated the Counter analyses at 11 microns to coincide
with the FPSSS. In this region of overlap a median volume diameter of 3.8
115
-------
.0 I'm
Fipiire 52. SRM photomicrographs display
the primarily spherical Duke Power Plant
fly ash (upper) and the clumped aggregates
of iron oxide (lower).
116
-------
FPSSS/COULTER DUKE POWER
FLY ASH SIZE DISTRIBUTION COMPARISONS
io3-
102H
E
a.
«
'E
o
Range 3 Range 4
10-
— 100 Mm ORIFICE
— 200 Mm ORIFICE
• FPSSS
I i
0.1 1.0 10 100
DIAMETER (/j.m)
Figure 53
117
-------
FPSSS/COULTER COUNTER IRON OXIDE
SIZE DISTRIBUTION COMPARISONS
10
10 -
n
i
o
0.1
„•
10 H COULTER COUNTER IRON
OXIDE 200 M"! APERATURE
• FPSSS
— COULTER COUNTER
200 /xm APERATURE
(cyclone sample)
10
100
DIAMETER (/urn)
Figure 54
118
-------
microns is computed.
The iron oxide measured with the Coulter Counter using a 200 \m orifice
revealed a higher median volume diameter of 25 microns. No measurements with
the 100 ym orifice were made due to sieving difficulties. Inspection of the
material as revealed in Figure 52 shows that the most numerous particles are
invariably submicron. Only through clumping are particles of large size gener-
ated. This problem will be discussed in more detail later.
Beginning now for the 15 test runs we should first state that the primary
means of comparison was through gravimetric analysis of a 9 stage Anderson Im-
pactor. Because of certain unknown factors in the impactor analysis, primary
cut points, and the lower resolution as compared to the FPSSS, the fairest
comparison is in terms of total mass loading. A summary of the mass loading
comparison is given in Table VI. During these test runs the tunnel speed ranged
from 2 to 30 msec . The sampling intervals for the impactor runs varied from
10 minutes to 2 hours depending on the mass loading and velocity produced. In
general, the impactor was run sub-isokinetic with the exception of the final
pair of runs. Sub-isokinetic conditions tend to over-sample large particles
while super-isokinetic conditions under-sample the larger particles. However,
with the mass distributions measured the average error would probably be less
than 15%.
_3
The first run was performed at a set mass loading of 0.06 gm and a
10 msec tunnel velocity. The average mass loading as measured by the FPSSS
— 3 ~3
was 0.055 gm , in good agreement with the impactor value of 0.049 gm . Run
_3
numbers 2 through 6 used a tunnel mass loading setting of 0.58 gm and tunnel
velocity of 2 and 10 msec . The FPSSS values range from 0.46 to 0.61; the
average value is 0.57. The impactor mass loadings for run numbers 2, 3, 5, and
6 range from 0.42 to 0.5. Run number 4 generated an unexplainable low impactor
value of 0.07 (initial weighing is suspect). Run numbers 7 and 8 used set mass
loadings of 0.06 gm and a tunnel velocity of 10 msec (essentially a repeat
of run number 1). Again, a very good agreement was found among the FPSSS,
impactor, and set mass loading values. The final four runs using the Duke fly
ash used a set mass loading of 0.2 mg and a tunnel velocity of 5 and 30 msec
The first of these runs found the FPSSS in good agreement with the set mass
119
-------
TABLE VI
IERL WIND TUNNEL MASS LOADING COMPARISONS
Run No.
1
2
3
1)
5
6
7
8
9
10
11
12
13
li)
15
Date
2-8
2-8
2-8
2-8
2-9
2-9
2-9
2-9
2-12
2-12
2-12
2-12
2-11)
2-\k
2-\k
Tunnel
Velocity
_ i
m sec
10
10
2
2
10
10
10
10
5
30
5
30
10
30
30
Time of Day
10:00-11 :00 am
2:1)0-2:50 pm
3:33-3:<>3 pm
3:5l-lt:OI pm
10:02-10:17 am
I0:l)l)-I0:59 am
11 :2I am - 12:21 pm
1 :05-2:05 pm
9:51-10:06 am
11:29-11 ;1)9 am
3:09-3:39 pm
<):06-1):26 pm
10:514 am - 12: 5*1 pm
1:52-2:52 pm
3:l3-'i:13 pm
Time
Min.
57
10
10
10
15
15
60
60
15
20
30
30
120
60
60
Set Mass
Loading
0.06
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
.58
58
.58
58
58
06
06
2
2
,2
.2
.06
0.035
0.
035
Vol ume
ACF
25.82
it. 149
1).62
k.62
5.66
7-35
27.87
26.7
6.2
6.5
12.0
7-5
61.06
38.95
28. it
SCF
26.5
l).61
k.y*
*t.7i)
5.75
7. ^7
28.33
27.15
6.3
6.7
12.3
7.7
62.66
39.7
29.0
% Isokinetic
12.1
72.1
86
86
63
82
77.5
Tt
85
72
83
83
85
I'll)
105
Impactor
Mass
Loading
g m
0.0*19
0.
-------
loading while the impactor was more than a factor of 5 lower. In run number 10
the reverse is true, that is, the impactor gave a high value while the FPSSS
was more than a factor of 3 lower than the anticipated value of mass loading.
The impactor yielded a low value in run number 11, while the FPSSS was quite close
with a value of 0.17 gm . Run number 12 showed the FPSSS and impactor in good
agreement, with values of 0.13 and 0.14, respectively, representing about two-
thirds of the anticipated mass loading. The average for the four runs was 0.14
gm for the FPSSS and 0.13 for the impactor. These results suggest possible
variability in the tunnel (FPSSS and impactor sample locations were as close as
possible but still separated by several inches) or errors of unknown origin.
It is a bit curious that only in the final run did the FPSSS and impactor show
consistency. In each of the prior three runs one or the other measuring device
was substantially low, suggesting possible tunnel variations and the need to
profile the tunnel, to be discussed later. Inspection of the FPSSS data reveals
4 3
that the measured number density decreased from 1.2 x 10 to 5.8 x 10 — thus
the discrepancies in mass loading are largely the result of a change in number
density and not size.
The final three tests used the iron oxide ash with set mass loadings of
0.06 and 0.035 and a tunnel velocity of 10 and 30 msec . In the first test
the FPSSS produced a slightly higher value than the impactor which was extremely
close to the set mass loading of 0.061 gm . In the final two runs the FPSSS
measured values that were about 40% low, while the impactor values were about
30% lower than the set values of mass loading. After these runs were completed,
we found that a lot of the iron oxide ash settled in the contraction section just
beyond the injectors. In fact, the material had accumulated to an inch in several
places. This would, of course, partially account for the lower readings by
both the FPSSS and the impactor; the particles simply never made it through the
tunnel. From the noticeable clumping of the iron oxide material one would ex-
pect median mass values outside the FPSSS size range. A Coulter analysis of the
cyclone sample (on the impactor) gave a median volume diameter of 6.3 microns
as compared to the FPSSS value of 5.1 urn (revealed in the accumulative mass dis-
tribution ash shown in Figure 55). However, almost all of the material ended
up in the cyclone (25 out of 31 mg in run number 14 and 17.2 out of 17.7 mg
in run number 15).
121
-------
IUW -
CO oft
CO 80-
S
S 60-
H
<
-1
S 40-
O
O
< on
85 SO-
rt-
A
— .
—
:
2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER
Figure 55. FPSSS accumulative mass distribution for
the iron oxide for run #14. A median volume diameter
of 5.1 microns is indicated.
122
-------
In general, what appears to be excellent agreement in most of the above
runs is thought to be at least partially fortuitous. As far as the impactor
is concerned, errors in weighing, non-isokinetic conditions, and siting errors
would restrict overall accuracy to perhaps ± 20% at best. For instance, the
samples were weighed twice. The difference between the first and second weigh-
ings showed decreases of 50% and 20% for run numbers 1 and 2 but an increase of
40% in run number 3. This, of course, suggests that one must use a statis-
tical approach to impactor sampling and average over many samples. In Table VI,
all impactor mass loadings are the final weighing results. With regard to the
FPSSS the two sources of greatest error are contamination and errors in assumed
density. Laboratory results have already confirmed that the FPSSS should size
accurately enough regardless of particle composition. From Figure 52 we can
recognize that at least the Duke fly ash is almost entirely composed of particles
with spherical morphology; thus shape is of no major concern. Even in the case
of the iron oxide, the primary particles are so small (submicron) that shape
would not be a strong issue. The large clumps remain an open questions, how-
ever.
Contamination of the optics in the FPSSS occurred several times during these
tests and, of course, any loss of optical signal results in an undersizing con-
dition generating low mass. In several cases where it was obvious (such as when
the probe was misoriented to expose and trap particles rather than shield the
optics from particles), results were discarded. The highest levels of contamin-
ation observed were during run numbers 5 and 6 which involved the highest mass
loadings and 10 msec velocities. We also had some difficulty with the oxide
ash during cleaning; its more abrasive character permanently damaged the coating
on the objective which later required replacement. In any case, contamination
was observed to produce up to 50% lower values of mass loading during worst
contamination conditions. This corresponds to (approximately) a 20% systematic
underestimate of particle diameter. The errors in density can be subtle. The
density of the source materials was measured fairly accurately through picnometry;
however, it is not known if the density varies as a function of particle size. We
have briefly mentioned problems associated with siting of both the impactor and
the FPSSS and noted material settling during the final three runs. We must also
acknowledge that uncertainties in the set mass loadings are known to exist and,
123
-------
though the magnitude is probably small, must still be regarded as an unknown.
In summary, generally good agreement was obtained between the FPSSS and
the mass loading rate fed into the tunnel. Some disagreement between FPSSS and
impactor samples was noted, but overall uncertainty averaged less than 20%.
Agglomeration and differing constituents are problems with the iron oxide, as
evidenced by microscopic examination. Generally poorer and somewhat erratic
intercomparisons were noted. The oxide tends to contaminate the FPSSS optics
more readily than ash. The FPSSS tended to compare better with the set mass
loading input than with the impactor samples.
Turning now to the size and mass distributions, the standard operating pro-
cedure was to obtain both for each population measured. For each impactor run,
typically 4 or 5 different size and mass spectra were obtained from the FPSSS.
There were similarities in all of the spectra for a given material so we will
only provide a few representative samples for discussion. The size distribution
for run number 1 is shown in Figure 56 and the mass distributions in Figures
57a and 57b. The size distribution shows two statistically significant modes,
one at about 0.8 microns and a second at 2.25 microns. The first mode at 0.8
microns is also apparent in the mass distribution while the primary mass mode
ranges from 2 to 6 microns, generating a mean mass size of 5.5 microns. The
slight tendency for an increase in mass at sizes between 8 and 11 microns is not
statistically significant, and, from a number of other distributions, it is
estimated that less than 20% of the given mass covered by the FPSSS would lie
above its upper 11 micron limit. Figures 58 and 59a show the size and mass
distributions characteristic for run number 5. They are strikingly similar to
those just discussed for run number 1. The FPSSS's ability to generate the cor-
rect (similar) size distribution with an order of magnitude higher mass loading
is illustrated. Because of the higher counting statistics the mass distribution
is quite smooth out to 11 microns. Extrapolating beyond the 11 microns one would
estimate that the mass contributions at 11 microns would be down an order of mag-
nitude from the peak at 6 microns and again would correspond to less than a 10%
residual error in unmeasurable mass at sizes larger than 11 microns. Run number
7 had the same mass loading as run number 1 but the greater sample time provided
less noisy statistics in the accumulative mass distribution, as revealed in
Figure 60. Comparing Figures 60 and 57b differences are noteworthy only in
124
-------
10
E
U
10 -
f
i i i r
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER
Figure 56. Size distribution for IERL run number 1,
ending at 11:20, February 6.
125
-------
to"1
io'2-
V-
E
=*• -3 Ll
co 10 - I In
1
E
o>
-4
10 "
-5
10
i
—
—
HIII
.5
3.5 5
DIAMETER (yum)
6.5 8 9.5 11
Figure 57a. Mass distribution for IERL run
number 1, ending at 11:20, February 6.
-------
en
en
Q
UJ
D
5
O
O
100
80-
60^
40 J
20 J
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER (//m)
Figure 57b. Accumulative mass distribution for
IERL run number 1, ending at 11:20, February 6.
127
-------
10'
E
=*
7 10
o
z
10 -
i ! i i r
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER
Figure 58. FPSSS size distribution for IERL
run number 5, ending at 10:07, February 7.
128
-------
10'2-
1 10-3 -
n
E
W -4
10 -
-5
j —
—
•^^n
- •
—
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER (/xm)
Figure 59a. FPSSS mass distribution for IERL run number 5,
ending at 10:07, February 7.
129
-------
100
CO 80 -
Q
w 60 H
O
O
40
20-
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER
Figure 59b. FPSSS accumulative mass distribution
for IERL run number 5, ending at 10:07, February 7
130
-------
CO
Q
LU
I-
O
o
100
80-
60-
40-
20-
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER (,um)
Figure 60. FPSSS accumulative mass distribution
for IERL run number 7, ending at 11:45, February 7,
L31
-------
channel-channel comparisons.
The iron oxide in run number 14 was of a lower mass loading and somewhat
different size and mass distribution. There is no size or mass mode similar to
that previously observed in Figure 59a and 59b at 2 microns and the mass dis-
tribution peaks somewhat more strongly. Again, because of the low mass loadings,
statistics in the mass distribution at sizes beyond 8 microns were marginal.
Comparison of the actual size distributions derived from the impactors and
as measured by the FPSSS were made for all runs. Comparisons show surprisingly
good agreement in all cases but particularly good agreement with the first im-
pactor run. This is largely due to the omission of the pre-cutter and cyclone
used on all subsequent impactor runs. The pre-cutter and cyclone captured over
90% of the sample, making detailed size resolution a much more difficult task.
For completeness we have provided comparisons of the FPSSS and impactor size
distribution analysis for the Duke and iron oxide fly ash in Figures 61 and 62.
Figure 61 shows the results of impactor run number 1 and the average of runs 5,
6, and 7 compared to appropriate FPSSS samples. Figure 62 provides a similar
comparison for run number 14.
The results illustrate the effect of the cyclone on the impactor size
analysis. The cyclone used on the impactor during runs 5, 6, and 7 removed a
considerable amount of material at smaller sizes when compared to the run 1
results in Figure 61. In spite of the effects of using the cyclone both com-
parisons in Figures 61 and 62 must be regarded as quite satisfactory.
Alignment Studies and Spatial Profiling between Run Number 1 and Run
Number 2. While gravimetric analysis of the impactor samples was being con-
ducted, we were able to run a series of tests of the FPSSS sensitivity to align-
ment orientations. At this particular period in time the velocity was a constant
_3
10 meters per second with a projected mass loading of 0.06 gm . Figure 63
shows a 23 minute run in which the mass loading and number density measured by
the FPSSS showed striking stability. Figure 64 shows a succeeding run over a
30 minute period in which the probe orientation was stepped from 0° to 45°. As
the figure illustrates, there is surprisingly little sensitivity to changes in
alignment until approximately 25° rotation. At rotations beyond 30°, the
effects of misalignment are rather dramatic in that 45° shows a drop of about an
132
-------
ERL IMPACTOR/FPSSS
MASS DISTRIBUTION COMPARISONS
100-
• •
A
A » A A
4
4 •••••
. .' •: *.
•••V..; *
*
•
• **
1.04 * * .
* ** RUN i£5 -6 AND 7 (Avg.)
A IMPACTOR WITH CYCLONE
» FPSSS
• FPSSS
+ IMPACTOR WrTH
NO PICK UP DEVICE
1 1
0.1 1.0 10 100
DIAMETER (/xm)
Figure 61
133
-------
IMPACTOR/FPSSS MASS DISTRIBUTION
COMPARISON FOR IRON OXIDE
100-
10-
O5
1.0 H
0.1-
• IMPACTOR
» FPSSS
0.1
1.0
DIAMETER
Figure 62
10
100
134
-------
TIME SERIES FOR ROTATION OF PROBE HEAD IN WINDSTREAM
1C1-
<0
U O
5 io3
V)
10
cc
UJ
CO
D 10
O)
id3
< -4
2 10 -
-5
10 .
MASS
NUMBER DENSITY
13:00
23:00
TIME
33:00
Figure 63. FPSSS time series and number density plot
in IERL tunnel. This 23 minute run was performed just
prior to the alignment sensitivity study and shows re-
markable uniformity in both mass loading and number
density.
135
-------
u>
co
'Sio»
Z
z 1°2
tu
a
oc
Ul
! 10
z
1
CO
I
E
O)
O
-------
order of magnitude in mass loading; the greatest change occurred in the range of
25° to 35°. An examination of Figure 64 shows one rather surprising feature:
while the mass distribution changes by an order of magnitude, the number density
falls off by about only a factor of 2. This, of course, simply means that the
large particles are being rejected more easily than the smaller ones under
cases of severe misorientation. Since these results were immediately obvious
from the FPSSS printout we elected to look in detail at the size distributions.
Figure 65 shows mass distributions measured by alignment errors of 0°, 15°, 30°,
and 45°. The 0° and 15° mass distributions are quite similar, the primary effect
being slightly less mass at all sizes, with perhaps some further reduction in mass
beyond 2 microns. The plot of 30° shows, in general an order of magnitude loss
at sizes beyond 2 microns, with a sharp cutoff above 7 microns. At 45° there
is a strong effect at sizes larger than 2 microns, showing increasingly greater
mass loss with increasing size until cutoff completely at 6 microns.
From this brief treatment it is difficult to say whether the structures
produced under the 30° or 45° orientations are entirely reproducible; however,
the stronger^ general tendency to reject the larger particles is very apparent.
The cutoff at some larger size is predictable from imaging considerations and the
timing circuitry which establishes the accepted sample volume. Referring to
Figure 64 it would seem that misorientations beyond ±20° would produce unreliable
data. In Section 8, we discuss field results in which measured asymmetries
in the flow were encountered yet acceptable results produced.
Upon completing the alignment sensitivity study, we were able to scan the
tunnel diameter to map the variations in number density and mass loading. As
_3
shown in Figure 66 the mass loading was maintained at the same 0.6 gm values
used for the previous studies of probe orientation. The same velocity was also
employed. The results show surprisingly uniform distribution in the aerosol a-
cross the tunnel. In the actual impactor/FPSSS comparisons, the two sampling
points were at most a few inches apart. The results of Figure 66 would only ad-
mit to errors of 10% for siting differences between the measurement techniques.
Tests in Heated Tunnel. Tests in the heated wind tunnel at IERL were made
on a limited basis but provided one of two opportunities to observe particulates
at temperatures nearer to actual stack temperatures. This tunnel is rather small
and is designed to study ESP efficiency. The inlet is rectangular (approximately
137
-------
3.5 5 6.5 8
DIAMETER (//m)
9.5
Figure 65. Effects of rotation in windstream on
mass-loading. The mass distributions indicate that
misalignment results in progressively higher losses
the larger the size and alignment error.
-------
SPATIAL VARIATIONS OF NUMBER DENSITY
AND MASS LOADING IN IERL TUNNEL
o
x
2-
(fl
Ul
Q
MASS LOADING,
NUMBER DENSITY
10
15
20
.8
CO
- .5
,.4 9
O
o
L .3
DISTANCE FROM WIND TUNNEL INNER WALL (INCHES)
Figure 66. As shown in the figure the tunnel is
quite uniform in terms of aerosol dispersal. The
higher values near the walls mostly reflect the
increase in air speed (measured with the FPSSS)
rather than any increased concentration in par-
ticulates.
-------
4' x 2') and the flow velocity a relatively slow 1 m sec . The same Duke fly
ash used in the other tunnel was also dispersed here with a set mass loading
of 1 gm
The size and mass distributions characteristic of essentially all of the
runs are given in Figures 67 and 68. The integrated mass loading averaged
_ o
0.52 gm . This lower measured value can be explained in two ways. First,
the size distribution revealed that with increasing size there is a faster roll-
off in the mass distribution as compared to results given previously (in Figures
57 and 59) for the same fly ash in the other tunnel. This suggests possible loss
of the larger particles in the hot tunnel which we attribute to sedimentation.
Our measurements were made in the upper part of the inlet cross section. Later,
measurements in a lower part of this cross section revealed larger particles and
higher mass loadings, a finding that correlates with the second explanation for
lower FPSSS mass loadings—uneven distribution in the vertical.
Because of limited access we were not able to profile the tunnel, but we
did perform several time series studies. Figure 69 shows a time series of
number density and mass loading covering 6 minutes of sampling. Each data
point represents either 4 or 12 second averages. The changes in averaging
times were made to determine the time scale of the observed fluctuations. As
revealed, there is significant structure at both time scales. When we contrast
this figure with Figure 63a, we find there is greater variation in the material
feed rates, manifesting itself in considerable temporal inhomogeneity. One can-
not help but observe what appears to be dropouts in the material flow. For
_3
instance, the maximum value of mass loading is invariably close to 1 gm ; how-
ever, values as low as 0.05 gm are noted in the 2-second averages. There is a
suggestion that there might be times when material is not adequately dispersed
but settles to lower levels as clumps. We were able to confirm that a lot of
the input fly ash did collect on the floor of the inlet and on the top parts
of the FPSSS probe itself with only a short period of exposure. One might sug-
gest further that the truncation of FPSSS measurements at 11 urn could account
for the underestimate of mass; but the mass distribution of Figure 68 would not
support more than a 15% unaccounted loss unless it were bimodal.
In summary, the FPSSS observed 40-50% lower mass loadings. No other instru-
140
-------
10
7
§ 102
10
III I
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER
Figure 67. FPSSS size distribution for IERL
heated tunnel using Duke fly ash, for run ending
at 11:05 on February 8.
141
-------
10
10
-1
7 10
E
-2
E 10
o>
-3
4
10
-5
10 ~i i i i i i r
.5 2.0 3.5 5.0 6.5 8.0 9.5 11.0
DIAMETER
Figure 68. FPSSS mass distribution for IERL
heated tunnel using Duke fly ash, for run
ending at 11:05 on February 8.
142
-------
iuu 10 ~
10- 105-
i
7 E
E =^4
* 1-V°-
E °
o> Z
0.1- 103-
OO1 10
_^_
NUMBER DENSITY j
~
I - -
--
MASS LOADING
* A. +
— — ~
- . -
\
\ ~ - _ _
* 1° »
U.U1 IU , j j ||l
So o o o o
S 9 999
Is" 00 O) O r- C4
« « CO 5? 5fr Tt
Figure 69. FPSSS time series plot of number density and mass loading for 6 minute
run ending at 43:00 on February 8. Note the large fluctuations in number density
and mass loading as compared to Figure 63.
-------
ments were intercompared, so conclusions on this phase of the experiment are
less definite. The consistent underestimate of mass loading by the FPSSS (com-
pared to input feed rate) may be entirely due to settling losses.
Tests in Fuel Oil Boiler Flue. Tests in the flue of the oil-fired boiler
were primarily implemented to provide an entirely different test environment for
the FPSSS. The tests were run with //6 fuel oil having '\> 1% sulfur. The flue
velocity varied from 3-4 msec . The mass loading was previously determined
to vary with the amount of injected water and excess air used in the combustion
process; tests were made at 10% and 18% water and with excess air up to 48%.
From prior impactor data, we anticipated the mass loading at 18% water
:s i
-3
-3 -3
to be 0.06 gm and at 10% water, 0.04 gm . However, FPSSS measurements gave
values at least an order of magnitude smaller—typically only a few mg m
Typical size and mass distributions given in Figures 70 and 71 correspond to
-3
mass loadings of 0.0065 gm . These distributions only cover sizes up to
5.65 um and perhaps a factor of two increase can be justified extrapolating to
larger sizes, but the underestimate will still be quite substantial.
With regard to excess air, the increase in mass loading with decreasing
excess air was quite dramatic. Table VII shows that mass loading increases by
nearly an order of magnitude for a small reduction in excess air—from 30 to
20%.
TABLE VII
% Excess Air FPSSS Mass Loading (gm~3)
48 0.006
37 0.006
31 0.009
21 0.045
One suspects that the impactor runs were carried out with reduced excess air,
but this was not supposed to be the case. The best information available sug-
gested that the predicted mass loadings were more than an order of magnitude
higher than that observed by the FPSSS. This might be attributed to a predom-
inance of particulates too small (optically) to be sized or to general undersizins
144
-------
§
U
10
103H
10 -
10 -
I r
.4 1.15 1.9 2.65 3.4 4.15 4.9 5.65
DIAMETER
Figure 70. FPSSS size distribution for IERL
test in fuel oil boiler flue, ending at 13:00
on February 14.
145
-------
10
-2
10
10'3H
£
o>
10
-4
-5
10 i~ 1 1 1 1 r
.4 1.15 1.9 2.65 3.4 4.15 4.9 5.65
DIAMETER
Figure 71. FPSSS mass distribution for IERL
test in fuel oil boiler flue, ending at 13:00
on February 14.
146
-------
While the FPSSS showed poor agreement with the expected result, its relative
accuracy seemed reasonable and showed it to be a useful tool for boiler align-
ment. An error in mass loading of an order of magnitude would require a factor
of 2 error in sizing. There is no theoretical basis for such a large under-
sizing and it cannot be supported by empirical work. We should also point out
that the fact that the activity was very low indicated a lower concentration
than expected and not simply smaller particles.
147
-------
8. FIELD TESTING
Field tests of the FPSSS during the course of the contract were conducted
at the Colorado Public Service Co. Valmont power plant in Boulder, Colorado, and
the Duke Power Riverbend plant at Charlotte, North Carolina. The Valmont power
plant served as the primary test site with a variety of short tests during early
development as well as final testing with comprehensive impactor intercompari-
sons. The first full-scale testing at Valmont during April and May 1978 was
not with the final electronics and data acquisition system but utilized the same
probe head as finally delivered. Size ranges and data processing were different,
as will be apparent in the data to be presented. The tests at Charlotte in
April 1978 were only partially successful due to a pre-amp failure in the
FPSSS which necessitated the more comprehensive final study at Valmont. The
final tests at Valmont during September 1979 provide the best overall data set
in terms of quality and quantity.
8.1 Initial Testing at Valmont Power Plant
The first field tests of the FPSSS in-stack were conducted at the Valmont
power plant on the afternoon of April 24, 1979. The Valmont plant is a 300 MW
coal-fired power plant where general operation and control devices are similar
to those depicted in Figure 72. Measurements were performed at the 80 foot
level on the stack, affording good access. At this level a Lear Siegler Model
RM41 Opacity Monitor was also operating. These initial tests were limited to
an hour of measurement time and the entire set-up, measurement, and pack-up
were completed in less than two hours.
The FPSSS was operated with a set of electronics borrowed from a PMS
Model CSAS-100 instrument.* The pulse height analyzer was configured to provide
four size ranges of 15 channels each as follows:
Size Range Size Interval
Range 1 2.5 - 10 ym 0.5 ym
Range 2 1 - 5.5 vim 0.3 urn
Range 3 0.5 - 2.0 ym 0.1 urn
Range 4 0.4 - 0.85 ym 0.03 urn
*Note that these measurements preceded those at IERL discussed in Section
7.
148
-------
PULVERIZERS
FORCED DRAFT
FAN
COOLING TOWER
ASH DISPOSAL
Figure 72. Schematic drawing of a steam electric generating station.
-------
All size intervals were linear, with maximum resolution in range 3 being 0.03
Vim. The ability to switch ranges was found to be particularly important in
field measurements. Calibrations with particles of known size had been per-
formed previously using latex spheres and glass beads.
The FPSSS performed satisfactorily in these first tests. Samples were
taken at two-foot intervals across the diameter of the stack to within 5 feet
of the far side. We elected not to extend the FPSSS to its full boom length
because of large oscillations induced by gusts inside the stack. We later per-
formed structural tests to determine the maximum loads the boom could withstand
fully extended. These tests suggested a heavier wall thickness was needed.
After these tests the wall thickness was increased on the final two boom sec-
tions.
The stack temperature was 260°F. No difficulty was found in operations at
this temperature. For these tests the external 7.5 cm radius mirror was purged
but the window was not in order that we could determine its potential exposure
problems. The window's accumulated particulates resulted in a 25% loss after
an hour's operation. The window had an aerodynamic shield which wind tunnel
tests indicated would be quite effective; however, the stack turbulence obviously
reduced its effectiveness. A purge-shield fitting similar to that on the 7.5
cm mirror was designed and installed on the FPSSS. We decided to over-pressure
the region around the two beam folding plane mirrors as a result of this initial
experience. Although no contamination was observed internally, the high humid-
ity in the stack penetrated into this cavity and condensed on the inner side of
the window when later extracted and allowed to cool. The above indicated changes
were also after completion of these tests.
Measurements were taken at two-foot intervals from 3 to 17 feet in-stack
penetration. The stack had an I.D. of 21 feet and we did not have sufficient
boom to reach completely across. However, the measurements revealed a symmetric
profile of particle flux, and the lack of measurement over the final four feet
was not regarded as significant.
Velocity measurements were manually computed from average pulse width
measurements. The in-stack velocity profile revealed a greater amount of tur-
bulence than anticipated with gusts of 5 m sec common near the stack center.
The mean flow varied from 5 to 12 m sec ; the highest values were at the center.
150
-------
No regions of reverse flow could be identified.
The particle size distributions measured at 2 foot intervals were surpris-
ingly uniform as viewed on the CRT spectral display of the CSAS-100 electronics
console. In fact, once corrections for variations in flow rate were made, the
distributions revealed only minor spatial spectral differences which were of the
same order as temporal changes at a fixed location. For this reason, all meas-
urements were integrated together to generate a single average particle size
distribution representative of the entire stack (see Figure 73). The particle
size distribution is seen to be highest in number at the smallest size (0.4 ym),
although the slope is decreasing. The overall slope was quite steep, elimin-
ating the utility of range 1 due to low count statistics. A size distribution
with such a steep slope is characteristic of air passed through coarse filters
and is steeper than background ambient aerosol size distributions.
The computed average distribution of extinction cross section (2nd moment)
as a function of particle size is given in Figure 74. The extinction cross
section was computed for a wavelength of 5500 8 in an effort to make comparisons
with the opacity monitor operating with a white light source having its spectral
maximum at this wavelength. The extinction cross section is seen to peak at
4 2-3
sizes of 0.6 urn and has a volumetric integral value of 3.0 x 10 x ym cm
The integrated cross section over the 6 meter stack diameter provides a computed
transmission of 0.83 or an opacity of 27%. The Lear Siegler RM41 had comparable
opacity values of 24 to 36% during the period of measurement, as revealed in
Figure 75.
The close agreement between the Lear Siegler RM41 and the FPSSS opacities
must be regarded as fortuitous at this time. A certain amount of particle cross
section also exists below the 0.4 ym lower limit of sensitivity, although prob-
ably less than 10% of the total. The exact spectral characteristics of the
Lear Siegler RM41 were not available, and in this size range extinction is a
strong function of wavelength. Furthermore, an ideal transmissometer still
gains contributions from multiple scattering that were neglected in our treat-
ment. Finally, the soiling of the window resulted in a small but obvious under-
estimate of particle size and opacity.
The average particle volume (mass) distribution (3rd moment) is given in
Figure 76a. This is probably the most interesting of all the spectral plots. It
151
-------
DIAMETER (Ml
Figure 73. Average in-stack size distribution. The higher resolution
data for size range number 3 has been summed into 0,1 urn size classes.
152
-------
105.
104 _
103-
1Q2 -
101
DIAMETER (/J.)
Figure 74. Average in-stack extinction cross section distribution.
The higher resolution data for size range number 4 has beem summed
into 0.1 ym size classes* The mean cross section size is 0=79 ym.
153
-------
Ul
.p-
Figure 75. Lear Siegler RM41 opacity record for April 24, 1978
between 14:00 and 15:00, during which the FPSSS performed measure-
ments in the stack. The sharp spikes in the data record are related
to the rapping cycle, which is further discussed in Section 8.3.
-------
10S-
104 —
<7
103 —
102 —
101.
TOTAL PARTICLE VOLUME: 7.3 mm3nT3
DIAMETER (/j.)
Figure 76a. Average in-stack volume (mass) distribution, The higher
resolution data for size range number 3 has been summed into 0.1 ym
size classes. The mean volume (mass) size is 1.31 urn.
155
-------
reveals several dominant mass modes, each being statistically valid. Perhaps
most surprising is the dominant mass mode at 0.65 ym. It is more striking when
presented with the full resolution of range 4 as shown in Figure 76b. Here
there is little doubt that the FPSSS has captured nearly all of the mass. The
integrated mass loading was lower than anticipated, being on the order of 0.01
~3
to 0.2 gm m , depending on the assumed density. The known undersizing due to
the soiled window should have only resulted in underestimates of mass less
than 30%.
We expected mass loadings nearer a tenth of a gram; however, the only way
in which higher mass loading values would be compatible with opacity values of
less than 20% (regulatory compliance value) would be with an increase in size
and decrease in number density over that observed. A simple increase in size
of a factor of two provides an eight-fold mass increase but also a four-fold
increase in cross section and opacity. Since the mass-to-cross-section ratio
is proportional to radius, or size, it is necessary to increase size by a factor
of 10 and decrease concentration by a factor of 100 to maintain the same opacity
by a factor of 10 higher mass loading. Either of the above is totally unreason-
able from an instrumental standpoint. We were thus left to conclude that this
particular stack did indeed have a lower than expected mass loading, character-
ized by particles of about 1.3 vim diameter.
These initial measurements were, of course, of great interest. The meas-
urements revealed a fairly uniform concentration across the stack. Within
-3
the operating range of 0.4-5.5 microns, the number density averaged 50,000 cm
The size distribution indicated the maximum number density at the smallest size
whereas the peaks in the extinction cross section and mass distribution were at
0.79 microns and 1.3 microns, respectively. The computed mass loading was only
-3
0.01 to 0.2 gm m — much smaller than expected. However, the integrated ex-
tinction cross section produced an integrated opacity of 17%, in close agreement
with the opacity monitor reading of 14-16%. These results are fortuitous
in that corrections for the reduction in sensitivity due to window contamination
were not made and the exact transmission properties of the Lear Siegler opacity
monitor were not known; overall, the results were regarded as preliminary, but
extremely encouraging.
156
-------
14,000
12,000-
10,000-
8,000-
6,000-
4,000-
2,000-
SIZE RANGE #3
0.03/M PER SIZE CLASS
0.4
0.5
0.6-
0.8
I
0.9
1.0
0.7
DIAMETER (fj.)
Figure 76bo Average in-stack volume (mass) distribution from 0.4
to 0.85 UHI.
157
-------
On May 11, 1978 a second set of field measurements was performed at the
Valmont power plant to obtain data upstream of the control devices. These
tests were conducted at two locations on the input to the scrubber and pre-
cipitator. When compared to the previous in-stack measurements the results
indicated a significant difference in particle size distribution when plotted
as a volume (mass distribution). Most noticeable was the abundance of par-
ticles larger than 5 microns which were essentially completely absent in the
stack. A mean mass size of 4 microns was found as compared to 1.3 microns in
the stack. Perhaps most importantly comparing the two mass distributions re-
vealed an efficiency between 95 and 99% in the mass removal above 3 microns but
little apparent efficiency at the submicron sizes, shown in Figure 77 as a
removal efficiency plot. The data are of course only approximate in that over
two weeks elapsed between the pair of measurements. However, the opacity
values were nearly identical and no change in coal or operating parameters had
taken place; thus, the measurements should be characteristic of conditions
at both times.
8.2 Field Tests in Charlotte, North Carolina
An attempt was made on April 17, 1979 to perform comparative measurements
with the FPSSS and impactors on a coal fired power plant at Charlotte, North
Carolina. This is the Duke Power Riverbend coal fired generating station which
has previously been used as a test-site for EPA studies. Measurements with the
FPSSS were only performed on one day due to the failure of a programmable
amplifier in the probe head and the lack of suitable spares. This short period
of measurements is discussed here primarily in the interest of being complete.
Impactor data analysis was performed by personnel from Northrup Services.
On the 17th, three impactor runs were made, each of 45 minute sample
duration, initiated at 10:15, 13:00, and 14:24 LT. The mass loadings measured
by the impactor and FPSSS are summarized in Figure 78. A single FPSSS point
was obtained around noon and seven samples taken after 14:00. The FPSSS in-
dicated that the mass loading was approximately 2^ times less than the average
of the first two impactor runs. In the afternoon period the first FPSSS run
at 14:20 was nearly an order of magnitude higher in mass loading than subse-
quent runs; its value was 13 mg m . In this period the FPSSS was set up to
158
-------
VALMONT CONTROL DEVICE REMOVAL EFFICIENCY
99.9
Q
111
O
01
DC
III
I
0.
u.
o
LU
O
CC
SSI
CO
CO
99-
90 -
I
234
DIAMETER (/xm)
Figure 77
159
-------
CHARLOTTE TOTAL MASS COMPARISONS
100
E 10-
o>
E
• FPSSS
A IMPACTOR
\ r
10 A.M.
I I
12 NOON 2 P.M.
TIME OF DAY
4P.M.
Figure 78
160
-------
sample for four minute periods, and considerable variability in measured particle
activity was noted. Our analysis of these data has shown that the FPSSS sample
intervals were much too short to provide statistical coverage of the stack tem-
poral variability. The size and mass distributions measured by the FPSSS are
shown in Figures 79 and 80. Both distributions are extremely steep and, because
of the relatively small particle size observed, only the 0.4 to 5.65 micron
range was utilized. The only change noted in the size and mass distributions
at other times was the reduction in the number of particles. However, the
total number density was approximately four times higher for the run at 14:20 LT,
which thus largely accounts for its higher number density and mass loading
values.
The steepness of the size distribution increases the potential error in
mass computations, and in this case the mass distribution rolls off nearly two
decades between 1 and 2 microns; thus an undersizing error of only 10% could
account for a factor of two in mass. It is also important to examine the im-
pactor data accumulated: 16, 12, and 10 milligrams for each of these three
runs. These values are regarded as much lower than desired for gravimetric
analysis.
Turning back to the size and mass distributions in Figures 79 and 80, it
is interesting to compare these with those measurements at the Valmont power
station presented in Figures 73 and 75. The Riverbend data are slightly steeper,
with over four decades of roll off in number density between 0.4 and 2 microns.
The average number density is less than that found at Valmont, as was the
loading, mean mass size, and computed opacity. The range of opacity values
computed for the Riverbend plant was from 1% to 3%.
Both Riverbend and Valmont give a significantly steeper distribution than
obtained during the IERL wind tunnel test series. Of course, the feed fly ash
used at IERL is the material removed between precipitators instead of that which
passes through. Comparison of the mass distributions between the FPSSS and
the impactors is shown in Figures 81 and 82 for the sum of impactor runs 2 and
3 and the average of the first four afternoon FPSSS samples. The accumulative
mass distribution in Figure 82 shows the expected difference due to the 0.4
micron cut-off of the FPSSS. Comparison of sizes larger than 0.4 ym would show
reasonably good agreement up to about 1 to 2 microns. Above 1 to 2 microns the
impactor data indicate the presence of considerable mass, which is not noted
161
-------
10'
i
-------
10
-1
10
-2
1C
'3
E
0)
10
-4
1 0
-5
.4 1.15 1.9 2.65 3.4 4.15 4.9 5.65
DIAMETER
Figure 80. FPSSS mass loading for sample
run ending at 14:55 on April 17.
163
-------
CHARLOTTE OUTLET MASS DISTRIBUTION
10-
1.0-
E
a.
E
en
0.1 -
• IMPACTOR RUN
» FPSSS RANGES
0.01
0.1
1.0
10
10C
DIAMETER
Figure 81. Note the large decrease in
mass distributions at sizes larger than
1 micron for the FPSSS relative to the
impactor data.
164
-------
CHARLOTTE CUMULATIVE MASS DISTRIBUTION
10
1.0 -
£
o>
0.1 -
0.01
. IMPACTOR (SRI)
» FPSSS RANGES 1 & 2
0.1
—T~
1.0
10
100
DIAMETER
Figure 82. The above comparison illustrates
the effects of the 0.4 micron lower limit of
the FPSSS and the lack of mass contributions
at sizes larger than 1 micron more easily
observed in Figure 81.
165
-------
in the FPSSS data. This is more readily observed in Figure 81 which also shows
a much larger contribution of particles of about 0.4 to 1.0 micron diameter
than observed by the impactor.
In summary, the data must be considered of a spotty nature due to the
limited time of FPSSS operation. The integrated mass and opacity values from
the FPSSS are in reasonably good agreement with those derived from the impactor.
The size spectral data show, however; considerable mismatch at the larger sizes
which is difficult to explain from this data set alone. It is, however, a
systematic difference that we see exhibited in other field comparisons between
impactors and the FPSSS.
8.3 Final Tests at Valmont
Final field testing of the FPSSS was completed at the Valmont power plant
on September 13-15, 1979. Testing was done at both the precipitator inlet and
the stack. Primary comparison was with impactor measurements made by Southern
Research Institute (SRI). Initial FPSSS comparisons were made in mass loading
(with impactors) and opacity (with opacity meter). After analysis of the im-
pactor data by SRI, size distribution comparisons were also made. Transient
phenomena were particularly well characterized with the FPSSS during these
tests.
The impactors sampled the inlet and the outlet streams of the ESP control
device system in parallel with sampling by the FPSSS. University of Washington
Mark III Cascade impactors with grease substrates were used exclusively. The
flow rate was 0.3 ACFM with a 3/16" nozzle, giving a sampling velocity of 7.5
m sec
During the testing the boiler was stable at 165 MW, burning Rosebud coal
with 0.7-0.9% sulfur content. At the inlet ports, the gas stream exited the air
heaters at 138°C through two 8.5 x 1.8 meter ducts approximately 3 meters in
length. Impactor sampling was performed from above and near the center of one
of these ducts with the FPSSS in the adjacent port 1 meter away. The gas
velocity at this point was found to vary between 7.6 and 10.1 m sec
The outlet sample ports were in the stack at the same location initially
sampled. The temperature was 121°C. The impactors and the FPSSS were inserted
to a depth of 3 m at diametrically opposite sides of the stack, leaving about
0.6 m between the two instruments.
166
-------
These tests were designed to provide coverage for impactor intercom-
parisons that were lacking or incomplete in prior tests at Valmont and River-
bend. Three impactor runs were conducted on the stack outlet and four at the
precipitator inlet at essentially the same sampling points as those conducted
in the first test series in April, 1978. Of particular significance to the
results obtained during this test series was the fact that the scrubber was off-
line during all of these tests and one of the five banks of precipitators was
also non-functional. Stack emissions were considerably higher than in prior
visits-, evidenced by careful scrutiny of the appearance of the plume as well as
in the actual measurements. The set-up time on the impactor sample runs as
well as data quick checks allowed for other tests using the FPSSS to be con-
ducted between runs. Of particular interest here were time series outputs of
mass loading, number density, mass median diameter, and opacity that were corre-
lated with the rapping sequence used on the precipitators.
In addition to the FPSSS and impactor data we were able to compare the
Lear Siegler KM41 opacity meter with the FPSSS values on a near real-time basis.
Also on September 15 we were able to use a CSAS-100-HTS (CSAS-100 modified for
flue gas sampling with a hot test section) to monitor the outlet while we were
making inlet runs. The CSAS-100-HTS has greater time resolution than the FPSSS
and provided further detail on the precipitator rapping sequence. Finally, the
last impactor run from both the precipitator inlet and the stack outlet was
analyzed by optical and scanning electron microscope as an aid in interpreting
comparative results.
Prior to measurements at either the stack outlet or the precipitator in-
let, velocity profiles were taken with a pitot boom arrangement furnished by SRI
and with the FPSSS velocimeter. Of particular interest were the velocity
measurements in the stack which revealed a rather asymmetric flow and about a
factor of two higher velocities on the south port (where the FPSSS was eventually
located for all measurements) as compared to the north port (impactor sampling
location). Figure 83 shows the measurements taken by the FPSSS and pitot systems.
Of course, the two measurement systems could not be co-located during either the
velocity measurements or the sampling runs. Because of variations in what ap-
peared to be average stack draft velocities, it was necessary to monitor the
FPSSS velocimeter and enter an estimated average velocity for each sample run.
167
-------
VALMONT OUTLET VELOCITY PROFILES
20
16 -
12-
u
4)
-------
TABLE VIII
SUMMARY OF VALMONT FPSSS, IMPACTOR, & OPACITY DATA
Run
Outlet
1
2
3
Inlet
1
2
3
4
FPSSS
(mg/m3)
140. 0-L
46.5-S
71.2-L
45.0-S
79. 0-L
62.3-S
236. 0-L
290. 0-L
495. 0-L
325. 0-S
332. 0-L
Opacity %
44 -L
33 -S
34 -L
36 -S
27 -L
35 -S
MMD
(ym)
2.99 -L
1.37 -S
2.10 -L
1.27 -S
2.89 -L
1.78 -S
1.88 -L
2.20 -L
2.0 -L
2.1 -S
2.1 -L
(L.S.RM41)
Opacity
Outlet
34
35
34
32
32
35
35
Impactor
(mg/m )
53
58
62
1800
2700
1300
980
MMD
(ym)
4.6
4.2
5.0
3.4
3.6
3.3
3.3
-S Indicates FPSSS Ranges 1 & 2
-L Indicates FPSSS Ranges 3 & 4
MMD Mass Median Diameter
169
-------
It is noteworthy that the FPSSS in general gave higher velocities than the pitot
system; however, the pitot system is a directional sensor, whereas the FPSSS
velocimeter measures the resultant velocity through the laser beam irrespective
of direction. Using the pitot we were able to determine that flow components
other than axial were generally measurable. Thus, the higher velocity values
for the FPSSS are probably more correct for use in its data interpretation
than simply using the prior pitot measurements.
Table VIII summarizes the FPSSS and impactor measurements and provides a
direct comparison of integrating mass loading and FPSSS and KM41 opacity values.
During the first day, two impactor runs were made at the outlet late in the
afternoon (#1 and #2). On both of these runs the FPSSS produced four separate
reports overlapping each impactor run. For outlet run #1 the 0.5 to 11 micron
size range was used on the first report, followed by three reports using the
0.4 to 5.65 micron size range. On outlet run #2, we used the 0.5 to 11.0 micron
size range for the first three reports and the 0.4 to 5.65 micron size range for
the last report. The size distribution as measured by the FPSSS is given in
Figure 84 and the mass distribution in Figure 85. This distribution reveals a
much broader size and mass spectrum than previously observed; however, we were
not surprised in view of the shut down scrubber and precipitator malfunctioning.
We will later see that in fact the mass peak from 8 to 11 microns is an artifact
of heavy EMI induced by particle charge accumulation on the FPSSS and corona
discharge through the electronics console. It is noteworthy that the distribu-
tions observed on the 0.4 to 5.65 micron range at 14:11:01 shown in Figures
86 and 87 reveal a much steeper distribution than the same portion shown in
Figures 84 and 85, where the 0.5 to 11 micron range was used. Approximately
three times as much mass loading was observed on the 0.5 to 11 micron range
as compared to the 0.4 to 5.65 micron range. It is also worth noting that
the mass median diameter is approximately twice as large on the 0.5 to 11 micron
range as compared to the 0.4 to 5.65 micron range. The impactor total mass
value of 53 milligrams is in good agreement with the three FPSSS reports using
the 0.4 to 5.65 micron range. We will defer discussion of impactor based
size and mass distribution with the FPSSS until later.
The second impactor run followed completion of the first run by two hours.
The size and mass distributions observed by the FPSSS were only slightly dif-
170
-------
rt
10
«ft
104^
103H
10^
10 -
—T—
12
—r~
16
20
DIAMETER (/xm)
Figure 84. Average FPSSS size distribution
observed using ranges 3 and 4 during first
impactor run (Valmont outlet).
171
-------
CO
E
0)
10
1 -
10
r2_
10
-3
8
12
16
20
DIAMETER
Figure 85. Average FPSSS mass distribution observed
using ranges 3 and 4 during first impactor run (Valmont
outlet).
172
-------
E
o
10'
105H
103H
10 H
6
T
8
10
DIAMETER ()Ltm)
Figure 86. FPSSS size distribution for run
ending at 14:11 on September 13 (Valmont outlet)
173
-------
10
-1
10
ID
'3 -1
E
o>
10
-4-
10
-5
4
6 8 10
DIAMETER
Figure 87- FPSSS mass distribution for run ending
at 14:11 on September 13 (Valmont outlet).
174
-------
ferent from those indicated for the first run when comparing the two sets of
ranges. However, the strong second mode observed in Figure 85 was much less
pronounced. The observed median mass diameter was 2 microns rather than 3
microns (0.5 to 11.0 micron range) on run 2 versus run 1. The mass loading ob-
served when the 0.4 to 5.65 micron range was used agrees most favorably with the
impactor data value of 58 milligrams, although the data from the 0.5 to 11
micron range had fewer of the artifacts at the largest sizes it also showed
relatively good agreement.
The third impactor run was on September 14 at 10:00 a.m. Six FPSSS reports
were taken averaging over 8 minutes each, four from the 0.4 to 5.65 ym range
and two from the 0.5 to 11.0 urn range. The samples from the 0.5 to 11 micron size
range suffer from some of the large particle artifacts, as is evident in the
higher masses and mass median diameters. The size and mass distributions at
10:14 and 10:23 exhibit some differences from those observed the previous day
(See Figures 88-91). Perhaps of greatest relevance is the appearance of a mass
mode around 3 microns which was not observed the previous day. It occurs on
both the 0.4 to 5.65 micron and 0.5 to 11 micron size ranges. The start of a
mass mode (third) between 4 and 6 microns must be regarded with some suspicion
due to low statistics, although it may well be real. It is also noteworthy
that during this period of operations we observed a number of interesting trans-
ient phenomena with opacity values exceeding 60% at several points in time.
After completion of the third impactor run the impactor sampling apparatus
was moved to the inlet and the FPSSS left to explore the outlet transient
phenomena. We will discuss these in greater detail later; however, it was also
during this time period that we noticed fluctuations in the laser reference
voltage of sufficient magnitude to make the data suspect. These fluctuations
only occurred with the probe in the stack and did not appear to be related to
operating temperature. We attributed them to static discharging from collected
particle charges. The temporary fix in the form of a large filter capacitor was
added to the reference output, eliminating the problem (but not the source). The
size distribution also lost its large particle tail in the 8 to 11 micron range,
making the 0.4 to 5.65 micron range sufficiently wide to cover the largest par-
ticles observed. Note the difference in the size and mass distributions shown
175
-------
o
10
105-
! 3
> 10 H
2
10 H
10 -
4
8
10
DIAMETER
Figure S3. FPSSS size distribution covering third
impactor run, ending at 10:14 on September 14 (Valmont
outlet).
176
-------
10
-1
10
1C
r3 -
10
-4 .
10
-5
8 10
DIAMETER (,um)
Figure 89. FPSSS mass- distribution covering third
impactor run and ending at 10:14 on September 14
(Valmont outlet).
177
-------
10
10
6
10 -
O
10 -
!
6
8
10
DIAMETER (/xm)
Figure 90. FPSSS size distribution for time
period ending at 10:23 on September 14 (Valrnont
outlet).
178
-------
-1
co
10
10'2H
-3
10 H
10 -
10
-5
i
6
i
8
10
DIAMETER
Figure 91. FPSSS mass distribution for time period
ending at 10:23 on September 14 (Valmont outlet).
179
-------
in Figures 92 and 93 for this period of time as compared to Figures 84 and
85 during impactor run number 1. The mass loadings observed after the EMI
fix are somewhat lower than previously observed; however, the stack was also
producing lower opacity values; thus, it is not possible to say that the 0.4
to 5.65 micron range did or did not have some contamination from EMI artifacts
that were noticeable on the 0.5 to 11 micron range. However, in our judgment
the effect is smaller than the other known sources of error.
On the afternoon of the 14th, two impactor runs were taken at the precipi-
tator inlet. These sample runs were each 5 minutes in length and, because of
the higher mass loading, collected about 2 grams of fly ash during each run.
The FPSSS coverage included two reports during the first run and a single re-
port during the second run, with a fourth report in between. The total mass
loading average from the two FPSSS runs is seen to be approximately one-third
of that reported by the impactor. The size and mass distributions for a 4
minute FPSSS sample completed at 16:49:14 are given in Figures 94 and 95. These
data indicate much larger particles present than found at the outlet. For the
FPSSS mass loadings the activity correction was nearly a factor of two (see
Section 6.2 for collection relationship). Because of the high number density,
activities of 70-80% were observed on the most sensitive size range.
The velocity used for all FPSSS computations was 9 ms in this section of
inlet. Because the flow is unrestricted here we did not expect or find the
large fluctuations in the flow found in the stack. We thus felt the pitot
measurements might be more accurate than those reported by the FPSSS. The
FPSSS measured systematically higher values, more typically 12 ms , which we
knew to be somewhat high; however, the FPSSS mass loadings computed would be
underestimated by up to 30% if the FPSSS velocities measured were to be accepted
Because considerable mass lies well beyond the range of the FPSSS, this dis-
crepancy is in reality of minor concern.
During the final day of tests (September 15) two additional impactor runs
were conducted. The third precipitator inlet impactor run was from 10:08 to
10:13, with the fourth from 10:26 to 10:31. Bracketing these runs were 8
FPSSS reports, each of 4 minutes duration. The mass loading values from the
180
-------
10
10'
6
n
10*-
10 -
10 -
i
6
i
8
10
DIAMETER
Figure 92. FPSSS size distribution at 14; 00 on
September 14. This distribution lacks the spurious
counts at the larger sizes (Valmont outlet) .
181
-------
E
O)
-1
10
10"2H
10 -
10" H
10
-5
0
8
10
DIAMETER
Figure 93. FPSSS mass distribution at 14:00 on
September 14. This distribution lacks the
spurious counts at the larger sizes (Valnont outlet)
182
-------
o
10
4
10
5 to3-
2
10
10-
I
i
4
8
I
12
i
16
20
DIAMETER (/urn)
Figure 94. FPSSS size distribution for run
ending at 16:49 on September 14 (Valmont inlet)
183
-------
10
io" -i
-3
10
-4
10 ^
10
-5
i
8
i
12
16
20
DIAMETER
Figure 95. FPSSS mass distribution for run
ending at 16:49 on September 14 (Valmont
inlet).
184
-------
FPSSS are somewhat higher than those on the previous day while the impactor
shows a reduction in mass values. A typical set of size and mass distributions
is shown in Figures 96 and 97 for the FPSSS report ending at 10:13:24. The
primary difference between the data for the final two runs and the data on the
14th is found to be that the number density is systematically higher. The size
and mass distributions show remarkable similarities. The mass median diameter
at the inlet averages 2.2 microns for all the FPSSS runs, but, of course, its
bias to the low side is due to the cut-off at 11 microns. A closer examination
of Figures 96 and 97 also indicates a potential flattening in the mass distri-
bution at sizes above 8 microns. We were somewhat suspicious that, in fact, a
second mass mode beyond the instrument range might well be indicated. Of course,
with FPSSS mass values running one-third to one-fourth those of the impactor,
one would suspect either a second mass mode or at least a fairly flat mass dis-
tribution out to some tens of microns.
The difference between the FPSSS/impactor values on the 15th and the 14th
is difficult to explain. However, it may be due to actual velocity differences
and the fact that the FPSSS computations used a constant 9 m sec velocity.
Support for this argument comes from the observation of measured (uncorrected
5 -3
for activity) number densities above 10 cm on the 15th. Laboratory compu-
tations and testing indicate that the FPSSS cannot measure this high. If the
velocity was, in fact, higher (supported by FPSSS velocimetry), the actual
number density would decrease and the observed mass loadings on the 15th would
be similar to those on the 14th.
There were several opportunities during this test series to utilize the
FPSSS to observe transient phenomena. Perhaps most illustrative of the observed
changes in stack emissions is the opacity. Figure 98 is a plot of opacity
and mass median diameter for an 8 minute period on the afternoon of the 14th,
while we were waiting for SRI to move the impactor sampling equipment to the
inlet. The results clearly show a 2 minute precipitator rapping cycle. Notice
that the maximum values of opacity are 60-80% while the quiescent values are
15-20%. This sharp maximum is about a half minute in duration. It corresponds
to the period when the final row of precipitation plates are rapped. The pre-
cipitator is configured with three rows of plates, each having 12 individual
subsets all rapped at 2% second intervals. Thus, it takes 30 seconds to rap
185
-------
6
10
105H
104H
u
102H
10 -
n
0 4 8 12 16 20
DIAMETER (/xm)
Figure 56. FPSSS size distribution for run ending
at 10:13:24 on September 15 (Valmont inlet).
186
-------
8
I I
12 16
20
DIAMETER
Figure 97. FPSSS mass distribution for run ending
at 10:13:24 on September 15 (Valmont inlet).
187
-------
00
CO
VALMONT RAPPING SEQUENCE AS OBSERVED
IN OPACITY AND MASS MEDIAN DIAMETER
1
5
CC
1-
Ul
SE
<
0
z
o
UJ
2
(0
<
2
10- 100-
8- 80-
ye
6 - >. 60 -
t
O
4- | 40-
2 - 20-
Oo .
— U -"
-
-
OPACITY
~
-~ - - --
~™ "-— — _ ~—
• "-_ ----- ---"-""_ ~-_
— . _ _^-;— _ __ "— — —
MASS MEDIAN DIAMETER
i i i i i i i
0
0
0
0
O
0
O
0
O
0
O
0
O
9
O
9.
TIME (MIN:SEC)
Figure 98
-------
all plates in each row. The first row (closest to the inlet) is rapped twice
as often as rows 2 and 3, thus generating a 1-2-1-3 sequence. What is revealed
in the data peak is the rapping of row number 3 plates which clearly recollect
most of the material exhausted when rows 1 and 2 are rapped.
We observed the same phenomenon with a PMS model CSAS-HTS instrument on
September 15th (when the FPSSS was making measurements at the inlet). Figures
99a and 99b show two sets of data taken at 2-second intervals. Here we see a
factor of 5 increase in the 2 ym sizes, but essentially no change in the sub-
micron sizes (the observed count decrease on the graph is cancelled when the
activity correction is applied). These compare well with the changes in the
size distribution measured by the FPSSS during a rapping sequence. Both the
FPSSS and CSAS-HTS essentially show zero or little change at sizes of 0.5 ym
but a factor of 2 to 5 increase at sizes of 1-3 ym.
The opacity values seen in Figures 98 and 99 correspond rather well
to the RM41 data seen in Figure 100. Although the KM41 records are noisy, it
is not difficult to verify that the opacity seldom goes below 10%, in corres-
pondence with the FPSSS, and the peaks have similar amplitudes. In fact, during
one 1-hour time period, which included times from Figure 98, we were able to
match the 2-minute rapping peaks with nearly one-to-one correspondence, as indicated
in the scattergram of Figure 101. Figure 101 reveals excellent correlation
between the minima and maxima for the FPSSS/RM41 instruments. The maximum
opacity and number density variations are typically a factor of 3, while the
mass median diameter increases approximately 30%. This may be due to agglomer-
ation, which occurs in the electrostatic precipitator. The rapping cycle is not
so apparent in the number density, but is again easily observed in mass loading.
Figure 102 shows a time series of mass loading and number density values taken
just after the sample in Figure 98. A three-fold increase in mass loading is
evident, while the total number density changes slightly. This is consistent
with the particle sizes that are responsible for the observed changes in opacity
and mass loading being close to the mean mass size or perhaps the mean cross
section size, but certainly not much smaller. Otherwise, there would be a larger
change in number density than opacity or mass loading. This is, of course, con-
sistent with the rapping process that literally shakes collected material off
the ESP plates into hoppers. Some of this material is unavoidably re-entrained
189
-------
RAPPING SEQUENCE FOR SEPTEMBER 15th
AT APPROXIMATE MEAN MASS SIZE
240
220-
200-
180-
160-
UJ
< 140
X
o
cc
a.
(fl
I-
o
o
120-
80-
60-
40-
20-
CHANNEL *13
2.3 - 2.45
CHANNEL *15
2.60- 2.75
9
9.
9.
TIME OF DAY
un ir> \f> in
0066
i/) ip in ip LO in in
6666666
Figure 99a
190
-------
RAPPING SEQUENCE FOR SEPTEMBER 15th AT SUBMICRON SIZES
(CSAS-HTS)
UJ
z
z
<
I
o
cc
UJ
a.
«J
H
Z
3
O
O
600-
500-
400-
300-
200-
100-
CHANNEL #1
0.5 - 0.65
CHANNEL
0.8-0.95
CHANNEL
1.1- 1.25
in
6
If)
6
in
6
o
CM
in
6
CN
in
6
in
6
o
o
in
6
n
in
c
in
6
o
t
in
6
in
6
TIME OF DAY
Figure 99b
-------
. . . x
Figure 100. Lear Siegler RM41 opacity record for September 14. Note
the much greater amplitude in the opacity data spikes as compared to
Figure 75.
-------
VALMONT -
% OPACITY SCATTERGRAM
100
80-
60-
40-
20-
r
20
• TEMPORAL PEAK
o AVERAGE BASELINE
FOR SAMPLE PERIODS
60
40
L.S. RM41
i
80
100
Figure 101
193
-------
VALMONT RAPPING SEQUENCE AS OBSERVED IN
MASS LOADING AND NUMBER DENSITY
106-
10-
5-
10
I io«-
no3-
io2-
.01-
.001
NUMBER DENSITY
MASS LOADING
o
9.
o>
Ul
O
o
O
0
I
O
o
1
O
9
O
-------
into the outlet flow. One would expect it to consist of larger sizes, as
is observed.
After extensive analysis of the impactor data by SRI, the size distribu-
tions measured at the ESP inlet and outlet were compared with the FPSSS measure-
ments. Essentially all four of the impactor samples from the outlet were indis-
tinguishable from each other, as were all three inlet samples. It is thus
sufficient to select one of the impactor runs from both the inlet and the outlet
for comparison purposes. These results are presented in Figures 103-106. In
addition to the impactor and FPSSS data at the outlet we have included data
from the CSAS-HTS operating on September 15th.
The mass distribution comparison on an accumulative basis (Figure 104)
looks quite good at the outlet; however, noticeable differences arise when
examining the differential mass distributions of Figure 103. The FPSSS and
CSAS-HTS instruments are in close agreement, while the impactor only reveals
equal contributions around 2-3 urn, with reduced particles of smaller size yet
increasingly larger numbers of larger sized particles. This is essentially
the same result observed at Charlotte and leads us to conclude that the dif-
ferences are systematic and that it's necessary to account for them.
The mass distributions shown in Figures 105 and 106 at the inlet are both
in poorer agreement primarily due to one additional factor. This is that there
exists a considerable amount of mass beyond the 11 ym limit of FPSSS and no
attempt has been made to account for it. Activity on the FPSSS was also much
higher at the inlet than at the outlet and certainly the smaller sizes could be
easily off by 10-20% in corrected number density — and possibly by much more
if the potential velocity change previously discussed is real. However, a
quick comparison of Figures 104 and 106 reveals the same discrepancy at sizes
larger than a few microns. It is this discrepancy that we were most concerned
with because the lower impactor values at submicron sizes are not so large and
easily explained. We will address this problem in the remainder of this section.
First, we must recognize that the differences between the FPSSS and the
impactor data at sizes larger than 3 microns appear to be systematic. If we
had found the difference to be of a random nature we would probably not be
terribly concerned, since considering all of the possible errors involved, vari-
195
-------
VALMONT INLET MASS DISTRIBUTION
1000-
100 -
E
4
cn
10 -
1.0
0.1
• IMPACTOR
» FPSSS RANGES 1 AND 2
• FPSSS RANGES 3 AND 4
' *
I
1.0
I
10
DIAMETER (Mm)
Figure 103
100
196
-------
VALMONT INLET IMPACTOR RUN
CUMULATIVE MASS LOADING
10
10 -
o>
E
10 -
• FPSSS
A IMPACTOR
DIAMETER (/xm)
Figure 104
-------
VALMONT OUTLET MASS DISTRIBUTION
100
10 -
. A A
A A
v^.-""--
•V+A
E
a.
n
'* 1.0
o>
E
0.1 -
• 1MPACTOR RUN 3
A FPSSS RANGES 1 and 2
• FPSSS RANGES 3 and 4
+ CSAS RANGE 2
« CSAS RANGE 1
A
A
0.01
—I—
10
0.1
—I—
1.0
DIAMETER (/j.m)
Figure 105
100
198
-------
VALMONT OUTLET CUMULATIVE MASS
LOADING FPSSS/IMPACTOR COMPARISON
10 -\
01
o
5
o
3
10
10
-1
10" 10
PARTICLE DIAMETER
Figure 106
199
-------
ations of this magnitude are to be accepted. But the fact that the com-
parisons show systematic differences incites one to believe that with proper
resolution either or both techniques might eventually be substantially im-
proved.
We must bear in mind that there are uncertainties with both the FPSSS
and the impactor methodologies. The FPSSS is a high resolution instrument
measuring physical (optical) diameters of particles. By comparison, the impactor
is a lower resolution device that measures aerodynamic diameters. We, of course,
do not in reality know that the density does not vary with particle size; how-
ever we were able to obtain samples of the Valmont fly ash at both the outlet
and the inlet and have carefully examined and photographed them with optical
and scanning electron microscopes (SEM), examples of which are shown in Figure
107. The samples shown were obtained from the actual impactor stages for run 3
at the inlet and run 4 at the outlet. One can, of course, attempt to extract
size distributions from the SEM data; however, that task is extremely difficult
in view of the large depth-of-field produced by the SEM and the difficulty in
establishing the actual volume in any size analysis performed.
The resolution of the impactor is not limited simply to the fact that only
7 discrete size bins are afforded, but also by the mixing in sizes collected
between stages. In other words, the bin edges are "soft" and a given size has
a finite, and perhaps significant, probability of being sampled on more than one
stage. It is a known fact that limited resolution will produce a decrease in
the slope of a distribution parameter; thus, one would expect the FPSSS to
produce a slightly steeper distribution than the impactor from resolution con-
siderations alone. The SEM samples in Figure 107 show that sizes ranging over
factors of 3 to 4 are common within any one stage; however, a detailed examination
has revealed that quantitatively, the predominant effect is one of small particles
trapped on the upper stages as opposed to larger particles trapped on the final
stages. This, of course, has the interpretation that some of the mass of smaller
particles is being assigned to stages reserved for larger particles. It again
has the effect of flattening the distribution shown. But can it produce the
magnitude of the discrepancy indicated? We must finally keep in mind that at
sizes larger than a few microns the impactor analysis is admittedly less certain
and the second mode in the bimodal mass distribution of Figure 103 should not be
200
-------
10 ym
Figure 107. SEM photomicrographs at equal magnifications display Valmont Impactor samples. The
outlet samples (below) demonstrate the effect of the E.S.P, when compared with inlet samples (above)
Stage_l_photos are on the left, with Stage 2 on the right.
P T *V>>
^.-fe
-------
assigned much significance.
It is also worthy to note that the outlet FPSSS size distributions provided
cross sections and opacities in excellent agreement with the opacity monitors.
Impactor size distributions tend to underestimate opacity supporting the idea
that the small particle fraction is underestimated.
With regard to the FPSSS, it is intrinsically more accurate in sizing smaller
particles than larger ones. This arises from two sources: (1) the calibration
curve becomes steeper at sub-micron sizes, and (2) the rejection process is
less subject to error when the actual image size is infinitesimal. With the
above in mind we performed two kinds of analyses to shed further light on the
problem. First, all indications are that the impactor underestimates the number
of particles less than 3 microns relative to particles larger than 3 microns.
If we accept the FPSSS values as being correct below 3 ]m, we can artificially
adjust the impactor data upwards to match in this region and then reduce the
mass sizes larger than 3 microns by an equivalent amount (conservation of mass).
One might, of course, ask what justification we have in moving the impactor data
upwards at sizes less than 3 microns in view of the fact that the CSAS-HTS data
are still less. We know the opacity arises from sizes in this region and thus
believe the CSAS-HTS should read less in this region. Furthermore, the CSAS-
HTS has associated plumbing losses that cannot be accounted for. In any case,
requiring the impactor data to match the FPSSS at sizes less than 3 microns re-
sults in significant lowering of the curve at sizes larger than 3 microns. This
results in substantial improvement in the comparison but there remains some re-
sidual disagreement as one proceeds to larger sizes. The discrepancies increase
in magnitude owing to the difference in distribution slope generated by the two
classes of devices. Of course, we should be aware that any fair comparison must
end at about 11 microns, the upper size limit of the FPSSS.
From past experience with the FPSSS we know that it can reject large par-
ticles if it is misoriented. If the flow is turbulent, a certain percentage of
particles would invariably be misoriented. Recall the observed cutoff of larger
sizes observed in Figure 64; could we be losing large particles because of mis-
orientation produced by turbulence? Upon examining Figure 105, we find that the
FPSSS and the CSAS-HTS are in excellent agreement at sizes larger than 3 microns
and conclude that misorientation cannot be effecting the FPSSS; however, a
202
-------
closer look at the data indicates some support for higher counts by the CSAS-
HTS relative to the FPSSS and we have previously mentioned that the CSAS-HTS
has some unmeasurable line losses. One would thus conclude that indeed the
FPSSS may be rejecting higher numbers of larger particles than in fact it should.
Of course, the CSAS-HTS was not even operating on the same day as the FPSSS but
from the opacity records in Figure 100 one would suspect that the stack was
overall significantly cleaner on the 15th day than on the 14th and, if anything,
the CSAS-HTS should have seen fewer large particles by comparison. At the pre-
sent time, we probably are not warranted in adjusting the FPSSS data upward at
the larger sizes since we do not in fact have proof that it is an error. Suf-
fice to say that with suitable randomization of particle trajectories, some loss
of larger particles due to unfavorable rejection is likely. Further study of
this effect is warranted. It is, however, abundantly clear that it is not in
error by the amount indicated by the impactor data. Correction of the impactor
data for the small particle contributions previously indicated would bring about
excellent agreement. We do not believe the rejection of large particles by the
FPSSS to be unduly large. On the other hand, in laboratory comparisons at IERL,
where the flow was sufficiently laminar to preclude any rejection bias by the
FPSSS, no systematic discrepancies were observed. Differences between IERL and
SRI impactors and analyses must be recognized, however. We are thus left to
conclude that the "best guess" as to the size distribution from 3 to 11 microns
lies somewhere between the impactor and the FPSSS but most likely closer to the
FPSSS than to the impactor.
Finally, we should mention that it is not logical for the FPSSS to be cor-
rect below 3 microns and at the same time significantly undersample at larger
sizes. For such to be true, the opacities would be in error by as much as a
factor of 2 and we would be computing much higher masses with poorer agreement
than indicated. We must, therefore, conclude that the FPSSS, being in agreement
with indicated opacity and measured mass loading using two other independent
techniques, must be providing a pretty fair representation of the size distri-
bution, and furthermore, that the optical and mass properties of the particles
are probably adequately covered by its finite size range at the stack outlet.
At the inlet the FPSSS covers only a fraction of the size range contributing
to the observed impactor mass loading. It, however, again provides sufficient
size range coverage to adequately characterize particle optical properties.
203
-------
9. CONCLUSIONS
From the results presented in the theoretical analysis and laboratory and
field testing we are encouraged to believe that the FPSSS will be a satisfac-
tory instrument for in situ size distribution measurements in a hot stack
environment. The theoretical response computations discussed in Section 5 in-
dicate that systematic errors typically associated with light scattering de-
vices have been precluded in the selected 2-11° collecting solid angles. The
FPSSS has been tested with monodispersed latex and polydispersed glass spheres
and intercomparisons made with several other non-optical, as well as optical,
particle size spectrometers. In field comparisons with impactors and opacity
meters, the agreements in derived mass loading and opacity are probably within
the known expected errors in measurement. There is little doubt that the FPSSS
has the required accuracy and flexibility, at least potentially, to greatly aid
in the evaluation of control devices and stack stationary emission sources.
The above mentioned positive results do not fully indicate the real value
of the FPSSS. The FPSSS was not really intended to be a replacement for im-
pactors; it is simply much too expensive a device to compete with impactor
techniques, although it gives higher resolution in the measurement of the size
distribution. Clearly, its advantage lies in its much greater resolution in
size and time, the latter being a much more subtle parameter. Although the
FPSSS does not sample but a minute fraction of the air measured by impactors
(the ratio of their equivalent sample areas being about a factor of 200), it
is capable of measuring a mass increment as small as 10 g. Probably the
-4
minimum weighable mass difference on an impactor stage is 10 g. This dif-
9
ference in intrinsic mass resolution of 10 gives the FPSSS the capacity to meas-
ure relatively low outlet mass loadings (with minimal statistical sampling
error) in less than a second, compared to an hour of actual sampling by an
impactor followed by an even larger analysis time, particularly if size analysis
is desired. This much higher temporal resolution gives the FPSSS the ability
to reveal temporal phenomena that clearly cannot be addressed by impactors and
to make that information available immediately.
The size distribution in our analysis is not sharply resolved by impactors
due to the lack of sharp limits to the size bin associated with a given stage.
The problem is accentuated when the size distribution is particularly steep.
204
-------
This effect may be only of academic importance, since, for all practical
purposes, the sizing differences are small and the resulting spectra from either
an FPSSS or an impactor would probably not surprise anyone. Nor would anyone
be able to decide easily which spectrum represents a more desirable emissive
product.
We have found it of great benefit in our field evaluation to be able to
compare opacity values computed by the FPSSS with those generated by in-place
opacity meters. The good opacity correlation, simultaneous with reasonable
impactor mass comparisons, leaves little possibility of sizable FPSSS size
distribution errors. It may also not be apparent, but the FPSSS, like all near
forward scattering particle sizing devices, is capable of generating essentially
correct opacity values irrespective of the particle response oscillations asso-
ciated with Mie computations (as described in Section 5). This is because the
full extinction cross section (from which opacity is computed) follows the scat-
tering response (or partial scattering cross section) computed in Section 5 in
magnitude as well as in qualitative signature. In other words, a particle size
that corresponds to a resonant peak and generates a larger scattering signal
than some larger size also generates a correspondingly larger extinction loss.
You may not be able to measure its true size exactly, but its scattering and ex-
tinction cross sections are essentially error free at wavelengths close to those
of the He-Ne laser line. The passband of the opacity meters is sufficiently close
to the He-Ne line that the errors are negligible. Of course, over a large,
smooth size spectrum such effects would largely average out anyway. We would
thus encourage future users of the FPSSS to utilize opacity measurements as a
valuable method of cross checking.
It is helpful when analyzing the performance of a new instrument to bear in
mind the tradeoffs made in its design. With regard to the FPSSS two come to
mind immediately. One has resulted in the current lower size limit of 0.4 ym
and the other limits the maximum measurable number density to a lower level
than desirable. These two factors are synergistic.
The EPA initially asked for a size range of 0.5 to 5.0 Jim. We felt that a
0.3 ym sensitivity could be obtained since standard PMS instruments currently
size that small. Our current instrument has a 0.4 to 11.0 ym range. The dif-
ference between 0.3 and 0.4 ym is about a factor of 3 in signal. An expansion
205
-------
of the collecting angles from 11 to'20° could pick up this difference. How-
ever, this would require moving the sample region closer to the prime objective
since its size cannot be increased by the rough factor of 2 required and still
allow the FPSSS head to fit in standard 4 in-stack ports. Our wind tunnel re-
sults do not preclude moving the sample region closer but it obviously will not
be as aerodynamically clean as it currently is. In view of the turbulent con-
ditions observed in the Valmont stack, clean aerodynamics may not be a profitable
design feature and movement of the sample volume closer could indeed be a better
choice.
The current instrument design can handle number densities up to several tens
of thousand per cubic centimeter although coincidence losses must be corrected
for. An increase in magnification from 20 to 30X should provide measurements
approaching 10 cm . An increase in the inner collecting angle from 2 to 4°
also doubles the maximum number density where measurements can be made. Halving
the laser beam diameter has a four-fold effect.
We previously mentioned that achieving a higher maximum number density
limit and greater sensitivity were strongly related. First, it should be ob-
vious that increasing the sensitivity automatically increases the number of
particles viewed and thus coincidence losses. Conversely, doubling the inner
collecting angle reduces the light collected. The overall sensitivity is ob-
viously reduced as well, but there is a more subtle effect. The 2° inner angle
allows for very accurate measurements out to 10 ym by collecting the diffracted
light at small forward scattering angles. A deterioration in sizing performance
would result if the inner angle were increased above 3° for 10 ym particles or
about 5° for 5 ym particles. The latter choice would further require enlarging
the outer collecting angle to achieve adequate sensitivity. Halving the laser
beam diameter aids sensitivity as well as maximum number density limits. How-
ever, here one must be absolutely sure that alignment will be extremely stable.
To decrease the beam diameter by a factor of 2 requires either using a condensing
mirror of half the radius of the current one or doubling the laser beam diameter
input to the mirror.
Probably the easiest means would be to increase the magnification to 30K.
The only penalties paid are in increasing the detector bandwidth (admitting more
206
-------
noise) to see the relatively shorter pulses that would result and possibly in-
creasing the probability of unfavorable trajectory of large particles. An
increase in bandwidth is required because the photodiode is sized to view the
full transit of a particle and any increase in magnification will truncate the
time a particle is viewed.
There remain a number of unanswered questions that further use of the
FPSSS in the field should clarify. For instance, we indicated the possibility
that large particles might be rejected more easily if turbulence indicates
significant trajectory deviations from normal flow. Though we were not con-
vinced that any problem existed during our field measurements, it is something
that deserves further study. Users of the FPSSS in situations where large
particles are dominant would be wise to rotate the probe head and ascertain
whether any position other than normal to the flow maximizes count rate at
large sizes and leave it in that position if it does. Further use of the FPSSS
in the near future should clarify this open issue. There are ways to reduce
the sensitivity to orientation. One requires a reduction in the gain ratio
used to establish the rejection criteria and thus sample volume. The mask on
the beam splitter can also be modified so it is wider at the edges than at the
center (like an hourglass), permitting a wider spread in trajectory alignment.
We currently cannot say whether either is warranted.
We are also somewhat unsure of the stability of the internal velocimeter.
We indicated that its accuracy is probably 10-15%. It may well be the limiting
factor in the overall accuracy of the basic measurement system. Unlike impactors,
which only measure velocity to attempt to provide isokinetic matching but then
accurately meter the actual sample flow, the FPSSS is an in situ device requiring
in situ flow measurements as accurate as the final desired result. Again, pos-
sible improvements in the circuitry could alleviate instrumental sources of error,
but there is little one can do if the laser beam width is not stable. The
laser beam width is, of course, directly related to the transit time and thus
velocity computation. Multimode lasers can undergo temporal fluctuations in
output beam diameter amounting to at least 10%. Thus, it is probably impossible
to achieve better accuracy without additional optical hardware. For instance,
one could provide a separate detector (photodiode and beam splitter) that measures
transit time across a mask. As long as the mask is always smaller than the laser
207
-------
beam, fluctuations in the laser beam diameter would not be a factor. One other
attractive possibility is to use a laser operating in TEM mode. This is a
doughnut-shaped beam profile where size is invariant. It is also a smaller
beam that could allow for measurements at still higher number densities with-
out changes to the optical system.
We should finally recognize that the FPSSS is a new and highly sophis-
ticated device. There are more opportunities for failure than with conventional
techniques. It becomes contaminated with time, requiring cleaning. It is thus
viewed primarily as a research or investigative tool and not a routine monitoring
device. As is the case for all measurement processes, an experienced operator
is the best guarantee of successful use. Anyone who has tried to make measure-
ments around stacks knows of the difficulties that can occur. In designing
the FPSSS, we have tried to remain fully cognizant of the user's environment
and attempted to maximize the usefulness of the instrument. The real utility
of a device like the FPSSS is not in extending sensitivity to the smallest
particles or in having the ability to handle the highest number densities.
Rather, it is that it provides measurements of sufficient spectral quality to
gain insight into processes that produce changes in particle size and number
density and it retains reasonable verisimilitude in integrated properties such
as mass loading and opacity.
208
-------
REFERENCES
Knollenberg, R. G., 1977: Light Scattering and Imaging Array Particle Sizing
Methods for Stack Measurements, presented at Air Pollution Con-
trol Assn., Ontario Section Symposium, April 27, 1977
, and Hunten, D. M. 1979a: Clouds of Venus: Particle Size Distri-
bution Measurements, Science Volume 203, 23 February 1979, pp
792-795
, and Hunten, D. M. , 1979b: Clouds of Venus: A Preliminary Assess-
ment of Microstructure, Science Volume 205, 6 July, 1979, pp
70-74
, and Hunten, D. M., 1980: The Microphysics of the Clouds of Venus:
Results of the Pioneer-Venus Particle Size Spectrometer Experi-
ment, Journal of Geophysical Research Volume 85, No. A13, pp
8039-8058, December 1980
_, and Gilland, J. R., 1980: Sounder Probe Particle Size Spectrometer,
IEEE Volume GE-18, No. 1, January 1980, pp 100-104
, and Luehr, R., 1975: Open Cavity Laser 'Active* Scattering Par-
ticle Spectrometry from 0.05 to 5 Microns, Fine Particles, Aerosol,
Generation, Measurement, Sampling, and Analysis, Benjamin Y. H. Lui,
Editor, pp 669-696
, 1976a: Three New Instruments for Cloud Physics Measurements: The
2-D Spectrometer, The Forward Scattering Spectrometer Probe, and
The Active Scattering Aerosol Spectrometer, presented at the
International Cloud Physics Conference, July 25-August 6, 1976.
A. Meteor. Society Preprint Volume, pp 554-561
, 1976b: The Use of Low Power Lasers in Particle Size Spectrometry,
presented at the Optical, Electro-Optical, Laser and Photographic
Technology Symposium, August 23-27, 1976. SPIE, Volume 92
Borne & Wolfe, 1972: Principles of Optics, 3rd Edition, Pergamon Press Ltd,
England
Cooke, D. D. and Kerker, M., 1975: Response Calculations for Light-Scattering
Aerosol Particle Counters, Appl. Optics, Volume 14, No. 3, March
1975
209
-------
APPENDIX I. THE PMS CALIBRATION WIND TUNNEL FACILITY
The PMS wind tunnel is designed as a complete calibration facility for
particle environments encountered in flight and other high flow rate conditions.
The tunnel is an open circuit tunnel with an energy ratio of 6:5. It is aero-
dynamically clean and has a completely variable velocity control using an axi-
vane variable pitch fan, (see Figures Al and A2). The tunnel is capable of
velocities in excess of 200 knots in a 30" test section. A calibrated water
injection system duplicates flight conditions in cloud and precipitation en-
vironments. Complete monitoring of actual particulate size spectra with PMS
spectrometers enables thorough calibration of various instruments as well as
providing test conditions for research studies. A dedicated data acquisition/
recording system provides high resolution detailed data samples. During winter
months the tunnel can operate as an icing facility during below freezing weather.
Precipitation in the form of snow (or rain) is drawn into the tunnel for certain
test work. When combined with water injection mixed phase studies are possible.
The wind tunnel is fully instrumented with standard PMS spectrometers, full
state parameter sensors and data acquisition systems. Computer software exists
to assist the experimenter in handling all of these data. The following instru-
mentation is dedicated to tunnel operations.
Spectrometers:
1. Forward Scattering Spectrometer Probe (FSSP)
Size Range: 0.4 - 45 ym
2. 2-D Cloud Droplet Optical Array Spectrometer (OAP-2D-C)
Size Range: 25 - 800 urn
3. Experimental Airborne 2-D Grey-Scale Optical Array
Spectrometer Probe
State Parameter Measurements:
1. Total Air Temperature (2)
2. True Air Speed
3. Volumetric Airflow
4. Pitot Pressure
5. Static Pressure
6. Dew Point Hygrometer
210
-------
Data Systems:
1. PMS Data Acquisition System - DAS-64
2. PMS Dual 2-D Data Acquisition System
3. Particle Data Processing System - PDPS-11C
4. Pertec Tape Recorder and Formatter
5. Versatec Printer/Plotter
Of particular importance to many programs is the existence of aerosol
sourcing apparatus which can be used to study and.characterize in situ measure-
2
ment capabilities by the proposed FPSSS. A large 4 m vessel is normally used
to generate controlled populations of aerosol particles (see Figure A3). Popu-
lations of monodispersed latex, glass beads, salt nuclei or polydispersed
smokes are generated in the vessel and exhausted into the tunnel to provide
calibration points during tunnel calibration runs. As an example this set-up
has been used to calibrate the sample-inlet tube for NCAR's Electra/Sabreliner
Air-Sampling program. An inlet of this type is shown installed in the tunnel
in Figure A4. Note the paired FSSP's which provide in situ characterization of
the aerosol populations in free-stream. Besides the obvious sampling efficiencies
that can be determined for such inlets, problems with regard to sample lag and
acoustical resonance (standing waves) are being studied.
The PMS Calibration Wind Tunnel Facility was of considerable benefit to
this program. It was the primary test-tool in validating how well the FPSSS
design maintained its in situ characteristics.
211
-------
ho
h-1
NJ
Figure Al,
Photograph of PMS Calibration Wind Tunnel
Facility
-------
PMS CALIBRATION WIND TUNNEL FACILITY
ROOF OUTLET LOUVERS
24 AIR ATOMIZATION
NOZZLES FOR AEROSOL
NEBUUZATION-.
HOOF INLET LOUVERS
CLOSE-PACKED HEXAGONAL
SPRAY PATTERN —1» AIR
ATOMIZATION NOZZLES
FAN OUTLET
' EXPANSION
SECTION
FLOW
STRAIGHTENING
SECTION
Figure A2
-------
Rxhaust Blowerr,
4 M" Vessel
Aerosol Nebulizers
Figure A3. Wind Tunnel Monodispersed Sourcin^ Subsyste
m
214
-------
Sample Inlet
FSSP'S
Figure A4. Sample Inlet Tube Installed in Tunnel Test
Section with FSSP'S. Pumping System and"
other Samplers are Behind"Tunnel.
215
-------
ADDENDUM
Since the completion of the prototype FPSSS, four more Systems have been
fabricated. Two of these were purchased by Southern Research Institute (SRI),
Montgomery, Alabama. The third unit was purchased by Shell Development Labora-
tory of Houston, Texas. The fourth unit will be retained by PMS for contract
services work. The units purchased by SRI are being used in field studies while
Shell's unit is being applied in a research combustion facility. Feedback from
these initial users has brought about certain modifications which we believe
have increased the utility of the FPSSS.
From work with the prototype unit the one area of greatest concern was with
the velocimeter. These additional four units had velocimeters of an improved
design. Improvement primarily resulted from increasing the bandwidth of the
signal amplifier train and by providing separate pulse amplifiers for processing
pulse widths (velocity information) tailored to the particular size range selected.
These changes brought the velocimeter to within ±10% over the 2-30 msec velocity
range. This improved velocimeter was the only noteworthy change in these four
units as compared to the prototype.
With regard to early feedback through customer use of the three shipped units,
no problems were encountered in stack measurements; however, contamination prob-
lems were encountered when performing measurements prior to control devices. The
particulate matter was found to be contaminating the inside surface of the window
and the primary objective. Since this region is supposed to be in an overpressure
area, sealing of the stainless steel jacket had to be faulty.* However, since
some additional contamination was noticed on the outside surface of the window,
we elected to correct the problem by modifying the window purge-air shrouds and
increasing the purge-air supply as well as improving the sealing of the stainless
steel jacket. The single diaphram air pump housed within the heat exchanger has
been replaced with the double diaphram unit with twice the airflow capacity. The
air lines have also been increased in size. We are considering the installation
of an airflow meter to enable the operator to adjust the purge-air according to
particle flow velocity. This will preclude possible purge-air disturbance of
sample volume at low flow velocities.
*In one case corrosive vapors condensed upon the surface of the objective's
first element and etched into the glass sufficiently to require that the element
be replaced.
216
-------
completion of these modifications, contamination of optical elements
essentially ceased. There still are, however, when measuring before control
devices, instances where number densities exceeded the instrument's capacity at
the smaller sizes. There remains a number of additional possible modifications
that could benefit the user. We have recently found a superior insulation material
(to the fiberglass) in the form of a moldable insulation material. The fiberglass
wrap blankets are difficult to handle and deteriorate with each disassembly/
reassembly of the probe head. In certain cases it would be desirable to provide
a single shorter section of boom for use in small work areas. We have found the
lateral bearing to be awkward and unable to withstand the constant exposure to
corrosive gases. A new support clamp has been designed to replace the bearing
for stationary installations. An air seal on the lateral support bearing has
also been devised.
With regard to data treatment and presentation it would appear that the FPSSS
will cover the needs of most users. In fact, it appears desirable to reduce the
number of size classes, providing one single size range of 32 size classes, in the
interest of instrument simplicity. We plan to build such a unit in the near
future. We would also like to increase the maximum particle number density the
FPSSS can handle although here only small gains are possible. The laser beam di-
ameter at the sample plane can be further reduced by upwards of a factor of 2
which will allow measurements at 4X higher number densities. This will require
the addition of a negative lens assembly in the condensing path and necessitate
an x-y adjustment on the secondary objective for final alignment, the latter a
possible worthwhile improvement in any case. Thus it would appear that 100,000
3
particles per cm is a practical upper limit for the FPSSS in the future. It
is clear that number densities can be encountered well in excess of this limit.
Accommodation to higher numbers will require fundamental changes in head design
to allow scattering measurements at 90°. Using 90° scattering number densities
in excess of 10 cm are accessible.
217
-------
TECHNICAL REPORT DATA
{Please read Instructions on the reverse before completing)
1. REPORT NO
EPA-600/2-84-096
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
In Situ Field Portable Fine Particle Measuring
Device
5. REPORT DATE
May 1984
6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S)
Robert G. Knollenberg
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING OROANIZATION NAME AND ADDRESS
Particle Measuring Systems, Inc.
1855 South 57th Court
Boulder, Colorado 80301
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-2668
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 9/77 - 8/79
14. SPONSORING AGENCY CODE
EPA/600/13
15. SUPPLEMENTARY NOTES IERL~RTP project officer is D. Eruce Harris, Mail Drop 62,
919/541-7807.
16. ABSTRACT
report describes the design, development, and testing of an in situ
fine particle measuring device-- the Fine Particle Stack Spectrometer System
(FPSSS). It is a laser-fed optical system with detection by near-forward light scat-
tering. Sample volume is established by a high- resolution optical system that views
particle images in a dark field through a masked beam splitter. The FPSSS covers
an 0. 5 to 11.0 micrometer size range with 60-channel resolution. Absolute theoret-
ical accuracy is + or - 20% of size for completely unknown refractive index. The
FPSSS is designed to operate continuously at in- stack temperatures up to 250 C at
flow velocities up to 30 m/sec. It has been laboratory characterized and field tested
on coal-fired power plants at both the inlets and outlets of control devices. Its per-
formance indicates good agreement with impactors and excellent agreement with opa-
city meters in computed mass loading and optical opacity. Its size resolution is
greater than other currently known techniques. Its eventual use will be directed at
characterizing particulate emissions of stacks or other stationary sources and
qualitatively evaluating the performance and collection efficiencies of particulate
control devices now in operation.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
COSATi Meld/Croup
Pollution
Spectrometers
Particles
Dust
Lasers
Optics
Light Scattering
Pollution Control
Stationary Sources
Fine Particulate
13 B
14B
14G
11G
20E
20F
B. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
226
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
218
------- |