<>EPA
United States
Environmental Protection
Agency
Environmental Sciences Research
Laboratory
Research Triangle Park NC 27711
EPA 600 2-79-032
February 1979
Research and Development
Optical
Instrument for
In-Stack
Monitoring of
Particle Size
-------
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.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-79-032
February 1979
OPTICAL INSTRUMENT FOR IN-STACK
MONITORING OF PARTICLE SIZE
by
A.L. Wertheimer
M.N. Trainer
Leeds & Northrup Company
North Wales, Pa. 19454
Contract No. 68-02-2447
Project Officer
William D. Conner
Emission Measurement and Characterization Division
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory, U.S. Environmental Protection Agency, and approved for publi-
cation. Approval does not signify that the contents necessarily reflect
the views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
-------
ABSTRACT
A new light scattering instrument for in-situ measurements of particulates
in the 0.2 to 10.0 micron diameter size range is described. Two modes of
scattering are used, each with two wavelengths of light, to generate five size
fractions by volume from a distribution of particulates. One mode measures
polarized light scattered in two orthogonal orientations at an angle of 90° to
the optical probe beam. The second mode measures light scattered in near
forward angles (4 to 11°). Both modes allow the extraction of size data when
particles of different sizes are present simultaneously in the sensing region.
These principles have been incorporated into a prototype portable stack monitor,
consisting of a 1.5 meter long, 9 cm diameter insertable probe capable of
withstanding temperatures up to 260°C. The optical signals are carried through
fiber optic cables contained in the probe. An arc source and silicon photo-
detectors are outside the stack at the end of the probe, while a digital
microprocessor analyzes the set of measurements and calculates the size
fractions. The microprocessor, an air purge system, the lamp power supply, and
a digital printer are housed separately from the probe for ease of installation
and service.
The completed instrument was tested on coal-fired fly ash particulates in the
laboratory at a stationary source simulator facility and in the field at a coal-
fired power plant. Results and comparisons with other sizing techniques are
presented.
This report was submitted in fulfillment of Contract 68-02-2447 by Leeds and
Northrup Company under the sponsorship of the U.S. Environmental Protection
Agency. This report covers a period from October 1976 to October 1978.
iii
-------
CONTENTS
Figures vi
Tables viii
Acknowledgements ix
1. Introduction 1
2. Conclusions and Recommendations. ..... 2
3. Optical Research Activities 4
Theory 4
Additional Theory & Experiments on 90° . . 10
Scattering
Refinements of the 90° Scattering Response . 15
• Signal Level Computations 18
Studies of Alternate Configurations, ... 19
4. Theory of Instrument Design 25
Operating Environment 25
Optical Design 25
Spectral Balance 26
Electronic Filter 27
Purge System 28
Air Curtains 30
Summary Description 32
5. Testing of the Prototype 35
Theoretical Calibrations 35
In-house Calibration of the Completed Unit . 43
Field Trial at EPA Facility 53
Field Trial at a Coal-Fired Facility ... 58
Modifications During System Testing. ... 61
6. Conclusions 63
References 66
-------
FIGURES
Number Page
1. Log of scattered intensity per particle as a function of
observation angle, with forward direction at 180°. (Two
orthogonal polarizations of incident light are shown.) ... 5
2. Intensity difference per unit volume, (Ii - I2)/a3, for a
scattering angle of 90° 6
3. Configurations for 90° scattering for polarizations of
light 8
4. Experiment & theory for polystyrene spheres in water,
n = 1.20 9
5. Spectrophotofluorometer system used for experimental data in
Figure 4 10
6. Volume response for different collection angles around 90°. . 11
7. Volume response at 90° for index = 1.96 - Oi 12
8. Volume response at 90° for index = 1.96 - 0.66i (absorptive
flyash) 12
9. Schematic drawing of laboratory system used to verify the
volumetric response functions for 90° scattering 13
10. Experiments using system in Figure 9, with polystyrene spheres
in water 14
11. Theoretical and experimental curves for 90° scattering using
absorbing and transparent material, showing the flux difference
i2 - i2» per unit volume of material. Two sizes of polystyrene
spheres were used, 0.234 and 0.801 microns in diameter, and the
wavelength was varied to change the size parameter, a . .15
12. Comparison of different weightings of the difference, Ii-KI2 . 16
13. Effect of broadening the spectral bandwidth 17
-------
Number Page
14. Integral of primary peak as a function of real and complex
portions of refractive index 18
15. Stack and remote configurations for particulate size monitor,
as originally proposed 19
16. 90° scatter system with forward scattering capabilities. . . 21
17. Background/main beam ratio vs. scatter angle 24
18. Schematic drawing of optical system 27
19. Schematic drawing of electronic filter circuit 29
20. Schematic drawing of blower and pneumatic components. ... 29
21. Basic air curtain design and gas flow 31
22. Air chamber and an additional orifice plate used to create
serial mixing chambers 31
23. Air chamber with baffle to confine and direct air flow ... 32
24. Photograph of stack probe including all components .... 34
25. Comparison of forward scattering common volumes with and
without water in cell 44
26. Flux/Unit volume vs. particle diameter for OOP, Aerosols,
A = 0.45 microns 46
27. Flux/Unit volume vs. particle size for OOP, Aerosols,
X = 0.09. 46
28. Laboratory sample system for delivery of aerosol smokes. . . 47
29. Volumetric histograms of several smokes 51
30. Volumetric histograms for various dioctylphthalate concen-
trations and for cigar smoke 52
31. Size of fly ash as measured by impactor and optical monitor
at EPA, RTP source simulator 56
32. Volumetric response curves, as a function of particle diameter,
for the five channels of the prototype. Curves 1 and 2 are
obtained with 90° scattering, and curves 3, 4, and 5 from
forward scattering. These curves are for silica, with re-
fractive index of 1.5 63
33. Probe, electronics package, and digital printer used for
measuring the size particulate material in utility stacks . . 64
vii
-------
TABLES
Number Page
1. Volume Response as a Function of Wavelength 7
2. Background Normalized to Input Beam 22
3. Sample Region Purged with Dry Nitrogen 23
4. Histogram Intervals for Prototype Instrument. ..... 36
5. Forward Channels 37
6. Symbol Definitions 37
7. Volume and Transmission Factors 38
8. Watts of Radiant Power for 1 Part/Billion per Channel . . 39
9. Terms Used in 90° Calculations 40
10. 90° Signals for 1 Part/Billion 41
11. Detector Gains. 42
12. Preliminary Channel Equations 42
13. Normalized Response for Each Zone and the Inverse Matrix
Used for Decoupling 43
14. Theoretically Predicted Ratios for Three Particle Sizes . . 49
15. Experimental Data with Various Aerosols 49
16. Final Constants for the Prototype, as Built 50
17. Fly Ash Sample Tests at EPA-RTP Source Simulator-May, 1978 . 54
18. Impactor and Optical Volume Histograms for May, 1978 EPA/RTP
Tests 57
19. Test Conditions 59
20. Field Test Data from Coal Fired Facility 60
21. Characteristics of Prototype Light Scattering Instrument. . 65
vn
-------
ACKNOWLEDGMENTS
We wish to recognize the contributions of the following members of the
Leeds and Northrup staff to the design and development program described in
this report: Mr. Joe Haran, Mr. Howard Hart, Dr. Ernil C. Muly, Mr. Ernest
VanValkenburg and Dr. Hal Weber (Penn State University).
In addition, the EPA Technical Project Officer, Mr. William Conner,
provided considerable help and guidance throughout this program and his
assistance is gratefully acknowledged.
ix
-------
SECTION 1
INTRODUCTION
The purpose of this project was to develop a method for the real-
time measurement of the size of particles in source emissions and to
build a prototype instrument to verify performance of the methodology. The
particle size range of interest is 0.2 to 10.0 microns, and the method of
measurement of these particles is based on principles of light scattering.
Two modes of scattering are used: near forward angle for the larger part-
icles and 90° scattering for the smaller sizes.
Prior to the start of the contract work, research on the methodology
had been performed on determining volumetric fractions from the difference
in 90° scattered intensities of two orthogonally polarized beams. This
approach formed the basis for the work described here.
The work plan for this project included two phases. The first was an
experimental and laboratory demonstration phase, and the second involved
the design, construction and evaluation of an on-stack instrument for measur-
ing the size of particles of in-stack source emissions.
-------
SECTION 2
SUMNARY - CONCLUSIONS AND RECOMMENDATIONS
A prototype optical light scattering instrument for measuring the size
of particulate material in stacks has been designed, constructed, and tested.
The prototype combines two optical scattering modes into a single instrument
package for real-time measurement of volumetric fractions, according to size,
in the respirable range of 0.1 to 10 microns in diameter. The application to
stack analysis is unique and relatively new. Accordingly there is need for
continuing development and analysis. This section provides a summary of
the contract efforts: and the second section describes some recommendations
for future work in this area.
Optical techniques involving low angle forward and polarization depen-
dent 90° scattering have been combined in an instrument package for real-
time particulate sizing in stationary sources. A significant extension to
the technology of light scattering was the development and verification of
the technique to measure volume fractions of sub-micron material. The con-
cept involved measuring intensity differences of scattered, polarized light
at 90° to the direction of a probe beam. This technique was engineered
into a probe type configuration and combined with forward light scattering
to provide five size fractions in the 0.1 to 10 micron diameter range.
An insertable probe was fabricated, designed and engineered for use in
high temperature acid environments. Associated equipment; an air purge
system and sophisticated electronics detection and microprocessor package,
was designed and constructed.
Upon completion of the prototype, considerable effort was spent on
insuring the calibration by the use of various aerosol materials. Field
tests were conducted at EPA and at an East Coast coal-fired facility, and
the results were compared with impactors. The unit was subsequently de-
livered to EPA for their use and evaluation.
The instrument offers a method for in-stack analysis of particle size
without dilution or extraction of materials and directly produces a distribu-
tion of the particulates by volume fraction. This device thus provides a
unique and valuable measurement capability for analysis and monitoring of
particulates in source emissions.
-------
RECOMMENDATIONS FOR FUTURE WORK
The principles of light scattering for particle size measurement have
been well established. The forward scattering techniques are currently
being applied in a range of particle sizing instruments, and the 90° scatter-
ing principles were verified carefully before and during the contract work.
The combination of the two techniques in a configuration suitable for
stack particulate analysis produced a number of trade-off and compromise
situations. Many of these trade-offs could not be fully evaluated until
after the prototype was built and tested. While the unit works satisfactor-
ily, our experience suggests that a number of items might be appropriate for
future work. Some of these are listed here.
'l. The arc lamp light source is ideal for some of its optical proper-
ties, such as its point source nature and broad spectral band width. How-
ever, some of the associated electrical difficulties, the starting pulse,
in particular, required extensive circuit protection. With that in mind, a
review of light sources for subsequent units would be in order.
2. The forward scattering configuration proved to be very difficult to
calibrate, since each scattering angle covered a different region of the
probe. Forward scattering subsequently became much more important for
analysis of flyash samples than originally anticipated, as a significant
fraction of the material was one micron or larger. A review of the forward
scattering implementation would be appropriate for subsequent instrumentation.
3. The upper size limit could be extended in connection with the re-
view in item 2, since forward scattering, in principle, is useful to sizes
much greater than 10 microns.
4. Additional calibration experiments might be appropriate to verify
the relationship between forward and 90° channels. Such experimentation
would require a sophisticated particle or aerosol generation system to pro-
vide true Rayleigh scatterers, 0.1 microns in diameter or less, with
sufficient number to provide a good optical signal. Our experience with
conventional aerosols is that there is great difficulty in generating a high
percentage of the mass of material below 0.5 microns.
-------
SECTION 3
OPTICAL RESEARCH ACTIVITIES
Work under the contract began by measuring the optical flux scattered
by polystyrene spheres from the submicron range to a few microns in dia-
meter. These particles were illuminated by visible and near infrared sources
and the scattered flux was analyzed as functions of wavelength, polarization,
angular distribution, refractive index and absorption. Particular attention
was given to the 90° observation angle since we had previously verified that
size dependent information exists in the scatter flux at that angle.
A means was then developed to demonstrate in the laboratory the real-
time measurement of particulate size, and equipment was assembled to demon-
strate the feasibility of this method. During this process, forward angle
light scattering was proposed and investigated as a practical alternative to
measuring the portion of the particle size range which required long infrared
wavelengths when using 90° scattering.
The Phase I effort was concluded by developing preliminary optical de-
signs for the on-stack instrument which became the basis for the design
specifications of the prototype instrument to be developed in Phase II.
Inputs from the Project Officer concerning design and performance of the pro-
totype were incorporated into the design.
THEORY
Considerable work was done on collecting measurements of particles in
the 0.2 to 10 micron size range prior to September 27, 1976, when the con-
tract began. This work was based on a new approach involving polarized light
to determine volume fractions of material in well defined size ranges. The
theoretical background to that approach is presented here.
For visible light, optical scattering characteristics for the majority
of the particle size range of 0.2 to 10 microns must be determined through
rigorous theory, referred to as Mie Scattering Theory.1 Here, the exact size
of the particle plays an important role. For spherical particles, the wave-
length is usually expressed through the dimensionless quantity, a = ird/X,
with d as the diameter of the particle and X as the wavelength. Within this
realm, other parameters figure significantly in the characterization of
scattering. The refractive index (both real and complex portions) influences
the sngular distribution of flux. The size and shape of the particle, as
well as the polarization of the incident light, determines the pattern.
4-
-------
Mie theory has been studied by many researchers2'3, and numerous tables
of the angular distribution of flux have been prepared by others for use with
problems of this nature. In order to illustrate some characteristics of the
angular distribution of flux, Figure 1 presents graphs of the tabulated data"
from a = 0.5 to 40 for spherical particles witn an index of 1.33. Two orth-
ogonal polarizations are shown, referred to as Ii and I2.
Intensity Ii for values /4°
of a from 0.5 to 40
Intensity I2 for values j
'40
0.5
of a from 0.5 to 40
e.o
4.0
2.0
1.0
0.5
90
120
ISO
180
Angle of Observation, degrees
30 60 90 iaO 150 180
Angle of Observation, degrees
Figure 1. Log of scattered intensity per particle as a function of obser-
vation angle, with forward direction at 180°.
(Two orthogonal polarizations of incident light are shown.)
The log of the flux per particle is graphed, covering a hemisphere from
0° (light returning directly toward the source) through 180° (forward scat-
tering direction). For a = 0.5 and 1.0, the Rayleigh scattering character-
istics are evident, while for a = 40, only the envelope of the pattern is
shown; the flux is mostly in the forward direction. These curves change
somewhat as the refractive index is varied. However, one of the most stable
characteristics of the patterns for different indices of refraction is the
null at 90° for small values of "a" with I2, that is, for light polarized
so that the electric field is in the plane defined by the direction of propa-
-------
gation of the incident beam and the point of observation.
The difference in intensity can be calculated by subtracting the two
curves at 90°. The percentage change is greatest for a = 0.5 and 1.0, since
\2 goes to zero at 90°. However, the absolute value of the difference per
particle generally increases as the diameter increases.
It should be mentioned that these curves describe the light scattered
from individual particles. Suspensions of N identical particles widely
separated from each other produce an angular pattern N times as intense as
that of the single particle, if no significant shadowing or secondary scatter-
ing occurs. This assumption is referred to as single scattering.
Now let us examine the flux per unit volume of material in a given size
range. Typical particulate distributions contain a vastly greater number of
small particles than large ones and when the flux difference per particle,
Ii - lz, is divided by the volume of the particle and plotted against the
diameter, a well-defined peak is observed. The flux difference per unit
volume at about 90° for various size spheres with an index of 1.33 is shown
in Figure 2.
Q
X
Size Parameter,.. [a=
TTd
A
Figure 2. Intensity difference per unit volume, (Ii - I2)/a3, for a scatter-
ing angle of 90°.
-------
This curve has a peak at about a = 1.5, and the half width covers approx-
imately a range from a = 1 to 2. In order to identify a corresponding part-
icle size, a wavelength must be specified. For a helium-neon laser, with
A = 0.6328 microns, the half width covers diameters from .2 to .4 microns, and
the peak is located at 0.3 microns. Based on this response curve, other wave-
lengths can be used to measure volume components at other size ranges, as
shown in Table 1.
The curve in Figure 2 is for a specific refractive index, 1.33. Other
indices produce slightly different curves, although the basic characteristics
are preserved. The curve for glass spheres, for example (n = 1.55) has
its peak at approximately the same value (a = 1.5) and is quite similar in
shape to the curve in Figure 2 for the region around a = 1.5. The differences
occur principally in the fine structure to the right of the first peak.
TABLE 1. VOLUME RESPONSE AS A FUNCTION OF WAVELENGTH
(*Fundamental Laser Wavelengths)
Wavelength
0.4 microns
0.6328 microns*
1 .06 microns*
1.25 microns
2.2 microns
10.6 microns*
11.6 microns
Particle Diameter
at Peak of Response
0.2 microns
0.3 microns
0.5 microns
0.6 microns
1 .05 microns
5.06 microns
5.5 microns
50% Range
.13 to .25 microns
.20 to .40 microns
.34 to .67 microns
.40 to .80 microns
.70 to 1 .4 microns
3.4 to 6.7 microns
3.7 to 7.4 microns
Figure 3 shows the basic scattering configurations for measurement of
intensities, Ii and I2, for two orthogonal polarizations of light. Ii and I2
are related to the tabulated Mie scattering functions, ii and 1*2» through
the following equations:
with
IN
)±w r3
= I0NX2i2/4Tr2 = -^ I0N (i2/a3)f IT r3
= intensity per unit solid angle scattered when
N particles are illuminated by one unit of
intensity of polarized light of wavelength, X
(1)
(2)
-------
X =
r =
Io =
N =
a =
wavelength of light in scattering medium.
radius of the (spherical) scattering particles
intensity of polarized incident beam.
number of illuminated particles.
2rr r
Thus, if the size and number of particles remains the same, the response per
unit volume at one wavelength, (ii - i^/a3, can be determined experimentally
by varying the wavelength and multiplying X times the measured quantity
(Ii - I2).
Light Source
I~- Polarizer
11 E" Field
'£>\ ^Sample
Light Source
Polarizer
"E" Field
Sample
Detector
Ii
Figure 3. Configurations for 90° scattering for polarizations of light
The values for ii and i2 are the scattering functions for the polarizations
perpendicular and parallel to the plane of observation, respectively. These
functions depend on a, 6, and n, where 9 is the scattering angle and n is the
ratio of the refractive index of the particle to that of the dispersant.
Figure 4 shows a plot of |ii - i2|/a3 vs. a for n = 1.2 and 9 = 90°. By
changing X, the diameter corresponding to the maximum around a = 1.6 changes.
This procedure is used to scan the particle size distribution, thereby
creating a histogram of the distribution.
-------
Experiments were performed to verify these theoretical results. A
spectrophotofluorometer shown in Figure 5 was used to measure 90° scatter-
ing of polystyrene spheres dispersed in deionized water at 10 ppm. The ratio
of 1$ to Ip (scatter and reference signals, respectively) was measured for
two orthogonal polarizations and for sphere diameters of .234y and .801y.
For each particle size, the monochromaters were scanned from .35y to .65y.
The value X |Ii -I2| was plotted, normalized to peak at a = 1.6 for comparison
to the theoretical plot of (ii-ial/a3, (the dashed line curves in Figure 4).
Qualitative agreement between theory and experiment is evident. The
dominant feature, the large response lobe at approximately a = 1.6, is
quite evident. Similarly, the smaller alternating positive and negative lobes
are observed for "a" between 4 and 7. The lack of quantitative agreement
between these two plots is due primarily to the difference in reflectivity
of the beam-splitter for the two polarizations. Also, exact angular adjust-
ment of the polarizer was difficult in this instrument.
O.SOlu
Spheres
Figure 4. Experiment & theory for polystyrene spheres in water, n = 1.20.
-------
Beam
Splitter
Monochromator
\ I
Polarizer
Sample
O\ .
•*
Xenon
Arc
> ^
^
I,
\
H _
•.'• ::
.•;-;.".•
-|r
A
LI J
^
- 2mm Slit
Monochromator
' CO
Reference
Is
Signal
Figure 5. Spectrophotofluorometer System Used for Experimental Data in
Figure 4.
ADDITIONAL THEORY & EXPERIMENTS ON 90° SCATTERING
After the start of the contract, a computer program was written, based
on a subroutine obtained from IBM5, to generate data on the volumetric re-
sponse from rigorous computations of Mie scattering coefficients. One set
of curves generated from this analysis (see Figure 6) shows how the volu-
metric response changes as a function of angular collection for acceptance
angles around 90°, with an index of n = 1.5, and acceptance angles of +_ 1°,
+_ 10°, and +_ 20°. The dominant feature is the peak in the volumetric response
in the vicinity of a = 1.5, as expected, but the items of particular interest
are the secondary oscillations (indicated + and - on the curves). It is
desirable to reduce these secondary lobes to zero, if possible. Larger
acceptance angles give smaller contributions to the secondary lobes, owing
principally to the smoothing effect of integrating flux over a range of
scattering angles.
The effects of varying index of refraction were also studied. The effect
of absorption on the scattering pattern of fly ash was investigated. This
material has a nominal index of 1.96 - 0.66i. Figure 7 presents a curve for
a fictitious material that is non-absorbing (transparent), with an index of
1.96 - O.Oi. The secondary lobe dominates this pattern. Figure 8 shows the
response for an actual material, indicating the smoothing effects of absorp-
tion. This effect is related to well known phenomenon of how optical
extinction coefficients
smoothed as a function of absorption.
10
-------
An experimental system was then assembled, incorporating a lock-
in amplifier and an electronic ratio circuit to measure 90° Mie scattering
more accurately. This system enabled recording of continuous scattering
curves for both polarizations, Ii and I2; hence, the response curve for the
function (h - I2)X could be generated easily.
Volume Response at 90°
vs. a for
collection angles
of + 1*
+ 10*
Size Parameter, a
Figure 6: Volume response for different collection anoles arounr! 90°.
11
-------
ZO >»0 1-CC l.»C 1.80 ?.20 2.60 3.00 3-«0 3.80 ••SO «.»0 S.CC
80 1»?0 |<6Q Z.OO __?..«0 ?.BO _. 3.20 3«»0 _»«OO j>..«Q »r«
Size Parameter, a = ird
Figure 7. Volume Response at 90° for index = 1.96 j-_0i
o t.to 2-6C ..J?CO_ J««0_.».1C
C .1C l.tC I.«C Z.QO ?.«C
Size Parameter, a = ird
' \
Figure 8. Volume Response at 90° for index = 1.96 - 0.66i (absorptive fly
ash)
-------
The experimental system, shown in Figure 9, consisted of a xenon arc,
EG&6 monochromator, and two U.D.T. PIN diodes. The source flux was chopped
so that a P.A.R. 186-A lock-in amplifier could be used to filter ambient
radiation noise from the relatively small 90° scattering signal. The flux
passing directly through the sample was focused onto a PIN 6DP diode. This
reference signal was electronically filtered to present a D.C. signal to an
analog divider, which ratioed the scattered and reference signals to correct
for spectral variations in source intensity. Therefore, the output of the
divider was directly proportional to Ii or I2. While a rotating prism
polarizer alternately selected Ii or I2, a synchronous motor scanned the
monochromotor from 400 nm to 800 nm at a speed chosen so that a pair of
measurements, Ii and la. were recorded once every 5nm on a chart recorder.
BOTMINQ
SlANJ-THOWlPSOJ
POL M21 ZEE
XENON ABC
SOUPCE
Figure 9. Schematic drawing of laboratory system used to verify the
volumetric response functions for 90° scattering.
13
-------
This system was used to record data for .312 micron and .801 micron
polystyrene spheres, dispersed in a solution of 1% Triton X-100 in deionized
water at 20 ppm. These two particle sizes generated scattering data for "a"
values from 1.64 to 3.27 and 4.19 to 8.39, respectively. The angular
scattering field was 20° centered at 90° with respect to the incident beam.
The experimental results are plotted along with the theoretical response
curve for comparison (see Figure 10). Notice that the experimental data
agrees well with theory, both quantitatively and with respect to positions of
extrema. The only major disagreement occurs in the region around "a" = 2.8,
where the scattering signal became too low for accurate operation of the
analog divider.
Further experiments were performed to confirm the selective reduction of
secondary response peaks for absorbing materials. 90° scattering data was
recorded for transparent and dyed (absorbing) polystyrene spheres of .234y
and .801y diameter. This data is plotted along with a theoretical curve for
non-absorbing particles in Figure 11. Notice the excellent agreement between
experiment and theory for non-absorbing particles, and the selective reduction
of the secondary response peaks for absorbing particles. The primary peak
data (d = .234y) for absorbing and non-absorbing particles are almost ident-
ical, but the secondary response peaks (d = .801y) are greatly reduced for
absorbing particles.
There are two factors not accounted for in the experimental data. Back-
ground molecular scattering of the solution in which the particles are sus-
pended prevents the signal from going to zero where predicted. Also, the
changes in refractive index with wavelength may cause a slight shifting of
the location of peaks and valleys. However, even without correcting for these
factors, the experimental data appears to fit the theoretical curve quite well.
O,
OJ
cc
25,
20
15
o> 10
Experiment
Theory
W V
1 23456
Size Parameter a • i.333*>\/\
10
Figure 10. Experiments losing system in
in water.
9. with polystyrene spheres
14
-------
NON-A8SO!?B!N per unit volume of material. Two sizes of polystyrene
spheres were used, 0.234 and 0.801 microns in diameter, and the
wavelength was varied to change the size parameter, a.
REFINEMENTS OF THE 90° SCATTERING RESPONSE
Response functions for typical stack materials were evaluated during
this phase and three methods were investigated for reduction of the second-
ary lobes in the scattering signals.
1. Widen the flux collection field angle.
2. Differentially weight the polarization components Ii and la-
3. Broaden the spectral bandwidth of the source to convolve the response
curve.
The effect of increasing flux collection field angle is illustrated in
Figure 6. As the collection angle is increased to a 40° full field, with
uniform weighting of contributions, the secondary lobes are reduced signi-
ficantly.
15
-------
The second method involves reducing the weighting factor of I2 in the
Ii - I2 response function. Theoretical studies show that most of the large
secondary lobes are negative for refractive indices greater than 1.5. Since
the primary response peak is positive, these secondary lobes can be select-
ively reduced without losing primary peak response. Studies indicated that
a factor of approximately 0.7 for I2 best balanced the positive and negative
secondary peaks. Over the index range of interest Figure 12 shows the
resultant effect of differential weighting for Ii and I2. The secondary
negative lobes are now reduced to a size comparable to the secondary positive
peaks.
The third technique involves using a broad spectral bandwidth. All
previous data represents scattering over a very narrow spectral range. A
broad band source convolves the response curve in wavelength space, smoothing
the secondary lobes. The following equations describe the relationship of
the convolution interval between alpha or diameter space, a, and wavelength
space:
a =
TTd_
A '
so
lAal
AX_
X
(3)
(4)
1.0
Size Parameter ?'°
Cnmnaricnn n-f
16
-------
For constant -, Aa increases with a. Hence a broad spectral bandwidth
causes each point fh the convolved response curve to represent a summation
of points over larger alpha intervals as alpha increases. Thus, wavelength
convolution has little effect on the primary peak, but a greater and greater
smooting effect on secondary lobes.
The results of spectral convolution are shown in Figure 13. Note that
the primary peak shows little change with increase in ^- but the secondary
peaks are reduced significantly due to the larger a convolution interval.
The combination of wide collection angle, differential weighting, and
wide spectral bandwidth is generally effective in reducing secondary peaks
to a negligible level for "a" greater than 5.
The 0.7 factor for \^ causes a train of alternating plus and minus peaks
of comparable size and the spectral convolution smears these peaks together;
so they tend to cancel each other. The small secondary response which re-
mains between alpha of 2 and 5 can be reduced through additional data
processing.
Another study showed the variation in response as a function of the
refractive index of the material. The peak of the response curve remains
fairly stable as the index is changed, but the total scattered light energy
for a range of particle sizes in the region of the response peak varies some-
what. Figure 14 presents the integral of the scattered energy as a function
of the real and absorbing parts of the refractive index. Over the range of
material studied, a factor of two in scattered light energy can result. This,
however, is not of concern unless the material being measured has significant
differences in composition as a function of size.
».o Size Parameter
Figure 13: Effect of broadening the spectral bandwidth.
17
-------
1.8,
1.6
1.4
1.2
o- 1.0
£
to
0.8
o.
M-
O
CO
CT>
O)
0.6
0.4
0.2
;io Absorption (i =
Refractive Index
1.5 1.6
1.7
1.8
1.9
Figure 14. Integral of primary peak as a function of real and complex
portions of refractive index.
SIGNAL LEVEL COMPUTATIONS
The application of 90° light scattering was proposed so as to be com-
patible with both in-stack and remote in-plume situations. The geometry of
the two cases are shown in Figure 15. For the ir,-stack case, 5 reasonable
section of the center portion of the stack can be covered with a fixed source
and a detector of appropriate angular divergence. A thermal source or
multiple lasers could be used. If necessary, both the source(s) and
detector could be gimbaled to cover a range of locations within the stack such
that the axes remain perpendicular. It is expected, however, that some
reasonable assumptions about the distribution could allow a fixed source and
detector, thus allowing a mechanically simple execution.
18
-------
• N-STACK
REMOTE
IN-PLUME
DETECTOR
FIELD OF VIEW
COLLECTION
OPTICS
AND DETECTOR
Finure 15. Stack and remote configurations for participate size monitor,
as originally proposed.
An in-plume measurement would require laser sources due to the limita-
tions in intensity and directionality of conventional sources. However, the
theory of operations remains unchanged. The detector and the laser need to
be separated from each other and aligned so both are directed at the same
point in the plume, while making an angle of intersection with respect to
each other of 90°.
To evaluate the methodology, emphasis was placed on the in-stack config-
uration. A study was undertaken of the signal levels in a stack implementa-
tion for 90° scattering, when used to measure the size range of 0.2 to 10
microns. The study showed that the signal to noise ratio was less than one
for thermal sources and conventional detectors at wavelengths greater than
2.3 microns. The use of lasers is indicated to provide sufficient power for
measurements at 10.6 micron wavelength (5 micron diameter), and 77°K cooled
detectors are necessary for measurements at 4.6 and 10.6 microns wavelength
(2.3 and 5 microns diameter).
Due to the high cost of the optical sources and detectors necessary to
extend the operation of the instrument far into the infrared, and the add-
itional complexity of using a multiple source approach, a decision was made
to consider modifications to remote sensing of 90 degree scattering.
STUDIES OF ALTERNATE CONFIGURATIONS
Through communications and meetings with the Project Officer, it was
agreed to investigate modifications to the original proposal. An insertable
probe configuration seemed to be the most attractive approach, combined with
the use of fiber optic cables to collect the scattered light at the point of
measurement and bring the optical signals outside the stack to be sensed with
19
-------
conventional detectors.
ANALYSIS OF THE USE OF A FIBER PROBE FOR 90° SCATTERING
A feasibility study of the fiber probe configuration was performed.
Several considerations were involved in switching from the remote (port-
hole based) system previously proposed to the portable probe (fiber optics
based) system.
1. Fibers tend to depolarize light, so it is necessary to illuminate
the region with a polarized beam. When the incident beam has a
known polarization, only the intensity of the scattered beam need
be measured.
2. The spectral transmission properties of typical fibers limit the
range of wavelengths to between 0.4 and 2.0 microns. This allows
a range of particle response function peaks from approximately 0.2
to 1.0 microns.
3. The operating temperature in the stack can be as high as 260°C,
but special fibers are available that can withstand this temperature.
4. Since the fiber is much closer to the scattering volume, flux is
collected over a much larger range of angles than for the remote
case. As shown earlier, this was found to help in smoothing the
response curves.
This initial examination showed fibers to have advantages over remote
sensing for submicron particles but raised a question about ability to measure
the size fractions above 1 micron in diameter.
FORWARD SCATTERING CONFIGURATIONS FOR LARGE PARTICLES
As an alternative to using long wavelengths and exotic optics for meas-
uring large particles, forward scattering was considered in combination with
90° scattering. Studies were performed on several configurations. One
specific approach, evaluated in the early stages of system design, involved
measuring scattering at both 90Q and low forward angles, as shown in Figure
16. The light source is common for both modes. The forward scattering
measurement is made by using a mirror or prism reflector at the end of the
insertable probe to return a portion of the incident beam.
2Q
-------
Polarizer
Mi rror
f
y^fe
//*&
kl/
1 -- !
\ V n1?
1 Splitter
Mirror
Filter
Wheel /
/^' Mask Plane & Detectors (2)
Chopper
Figure 16. 90° scatter system with forward scattering capabilities.
During March of 1977 we met with the Technical Project Officer to discuss
some of the configuration options, and it was agreed to combine the two modes
in an insertable probe configuration using fiber optics.
Following that meeting, extensive analysis was performed on forward and
90° scattering combinations. Several changes and improvements were made, as
listed below:
1.
2.
3.
4.
An arc source was selected instead of the tungsten lamp.
Only two channels were to be measured using the 90° scattering
approach, at 0.421 microns and 0.850 microns in wavelength. These
wavelengths were later changed to 0.45 and 0.9 microns.
Three channels were to be measured using a conventional forward
scattering configuration, with the same wavelengths as for the 90°
scattering.
The forward scattered light is measured at the end of the probe in
the stack by use of three additional fiber bundles and a one inch
diameter lens located at the end of the probe.
The advantages for each of the above features is listed below:
1. A tungsten source is sensitive to damage from vibrations and has a
relatively large filament. The arc source is more stable in a vi-
brating environment. It has a more concentrated area of light, which
is more readily focused or collimated. This provides a well defined
beam, especially suited for the anticipated forward scattering work.
2. The third channel at 90Q, which was to use light at about 1.7 mi-
crons, would have required an additional set of detectors, special
polarizers, and very expensive fiber optic glass. It was found that
a forward scattering channel could be substituted for the 90° channel
21
-------
in the particle size region of 0.8 to 1.6 microns.
By changing the configuration of a true forward mode and adding a
third mask opening, the size resolution was considerably improved,
producing a minimum of overlap between adjacent channels. The
wavelengths being used for forward scattering are the same as for
the 90° case. Different algebraic combinations of the flux
measured at three points in the Fraunhofer plane are used to com-
pute the channel widths.
The reflected light system, proposed earlier, was simplified by
measuring the light at the end of the probe. This eliminated
difficult problems of vignetting unwanted reflected light and locat-
ing a mask and detector assembly in close physical proximity to
the light source. Improved size resolution was not possible in the
original configuration due to the limits on the range of angles
which could be measured using a fixed reflector.
OPTICAL BACKGROUND MEASUREMENTS
A laboratory system was set up to make background and scatter signal
measurements under conditions resembling the configuration of the planned
prototype instrument.
A photomultiplier tube was used for the detector to measure the back-
ground since some signals were lower than the range of sensitivity of the
silicon detector. For a given angle, the highest background levels were
found to exist for the shortest wavelengths.
Optical background levels were measured for different angles and wave-
lengths in the forward configuration. The intensity of the main beam was
measured with the same optical system by using neutral density filters to
attenuate the beam. These filters were calibrated in a separate test. A
ratio of background light to incident light was then computed, which removed
any need to make absolute calibration of the individual components. The
results are shown in Table 2.
Table 2: Background Normalized to Input Beam*
Angle
3.8°
4.8°
7.5°
10.3°
90° (+1°)
\ = .4500
1.1 x 10"8
7.3 x !0~3
3.9 x 10"9
2.7 x 10"9
6.5 x 10~n
A = .6500
—
2.U x 10~3
1.0 x 10"9
6.8 x 10"10
—
* Expressed as a fraction of the main beam
22
-------
Note the significantly lower background measurements at the longer
wavelength and the very low value at 90°.
Several experiments were attempted to define the source(s) of the back-
ground signals in the forward direction. The most dramatic improvement was
achieved by purging the sample region with dry nitrogen. In the forward
scattering mode, the background signal dropped significantly at all measured
forward angles when the ambient room air in a transparent tube was replaced
by flowing dry nitrogen. The resultant data appears in Table 3, and implies
that approximately 90% of the background light level observed in the forward
direction is due to light scattered from components (dust, moisture, etc.) in
ambient room air. The remaining background contributions are probably due
to molecular scattering and residual diffraction effects from the beam form-
ing optics.
Table 3: Sample Region Purged with Dry Nitrogen
Angle
4.75°
7.50°
10.25°
Background with
Dry Nitrogen
(Referenced to Main Beam)
3.06 x 10"9
4.08 x 10"10
4.49 x 10"10
Reduction (%)
over Ambient Air
76° and 70% Rel . Humidity
-58%
-90%
-83%
During the course of these experiments, some variations in the back-
ground levels were observed in ambient air, perhaps related to changes in
humidity in the laboratory due to prevailing weather conditions. Figure 17
provides a graph of the background level as a function of angle, as measured
for three different situations, two in ambient air, and one in dry nitrogen.
23
-------
ID'7,
10'8-
OJ
CO
-o
(D
CQ
I0-ia
o Ambient Air
* Ambient Air
* N2 Purge
4812
Scatter Angle (Degrees)
Figure 17. Background/Main Beam Ratio vs. Scatter Angle
Other experiments were performed to determine the background effects of
scattering from dirty windows and refraction due to thermal gradients. Back-
ground readings were taken at 0 = 4.75°, 7.5°, and 10.5° for three cases:
1. With dirty glass window on the probe side of the beam forming lens.
2. With dirty glass window directly in front of collector lens.
3. Hot air flowing through the sample region of the probe.
In all three c*<;p<;; no significant change in background was observed.
24
-------
SECTION 4
THEORY OF INSTRUMENT DESIGN
This section describes the design of the instrument: its operating
environment, optical design, component selection, electronic filter for
noise suppression, purge system, and air curtain design for protecting
optical surfaces in the stack from contamination, and then, a brief overall
description of the instrument.
OPERATING ENVIRONMENT
Application to oil and coal combustion processes at various field lo-
cations requires an instrument to meet diverse operating, measurement, and
deployment conditions. In addition, an on-stack instrument must withstand
the rigors of vibration, corrosion and gas flow variabilities, characteristic
of in-situ measurements. Within a duct, the probe may be exposed to temp-
eratures between 90° and 260°C, gas velocities up to 18 meters/second, and
acid and water vapors and condensation as well as the fouling effects of
particulate. Particles of interest range between 0.1 and 10 microns in
diameter, have a specific gravity near 2.5 and indices of refraction in the
range of 1.5. The size distribution may be broad, evenly spanning the in-
strument five size channels; or narrow, with as much as 80% in a single
measurement channel. Total loadings are anticipated to fall between 0.01 and
0.1 grams per cubic meter (4-40 ppb) while the instrument's measurement span
extends to 1.0 grams per cubic meter. At controlled intervals, precipitator
rapping creates more dense, transitory clouds of particulate.
The instrument is to be used at a variety of sites where access to the
sampling point may be restricted to a four inch port and service limited to
110 volt power. Near the sampling point, instrumentation will be exposed to
combustion fumes, particulate, and the weather and protected only be a
tarpaulin.
OPTICAL DESIGN
The interior stack temperatures rule out the use of in-situ detectors,
so a set of high temperature fiber optic cables are utilized to sense the
scattered light at five points(three forward angles and two orthogonal posit-
ions at 90° to the probe beam). These cables are installed inside a 1.5
meter long stainless steel tube, whose outside dimension is approximately 9
25
-------
cm. The active sensing region is towards the tip of the probe and is a slot
36 cm long by 3 cm wide. A lens located in the tube tip collects scattered
flux for the forward angles, while the two 90° fibers look directly through
protective windows and air curtains into the sample region. Each fiber cable
runs to an individual detector, located in the transceiver outside the stack,
at the end of the probe. A sixth detector is used to monitor beam strength,
as reflected from a tilted glass plate in the transceiver section.
The chopped optical probe beam is generated by a xenon arc source and
chopper wheel, located in the transceiver. Thin-film interference filters
with center wavelengths of 0.45 and 0.90 microns are combined with linear
polarizers in an indexing filter wheel. The filtered light falls onto a pin-
hole, which serves to adjust beam position, and then passes down the center
of the tube in a nearly collimated beam of alternating blue or infrared
polarized light, with a beam diameter of approximately 1.25 cm. To avoid
spurious signals the light beam is not reflected internally, but exits the
probe through a hole at the tip. A schematic drawing of the optical system
is shown in Figure 18.
SPECTRAL BALANCE
The components in the optical system were selected so that equal load-
ings of particles in either the blue or the infrared beams generate nearly
equal detector outputs and full scale readings at the computer input. In
general, less flux is scattered from the infrared beam than is scattered
from the blue beam, but better transmission through the fiber optic bundles
and response to infrared by the silicon detectors actually produce higher
detector outputs for the infrared beam. To equalize outputs, a neutral
density filter (with 70% transmission) was inserted into the infrared beam.
Other techniques for improving the sensitivity and spectral balance of
the optical system were investigated and discarded. Substitution of a
mercury bulb for a xenon doubled the power of the blue beam but decreased the
infrared power by a factor of five, resulting in an overall loss in instru-
ment sensitivity. Specifications of photomultiplier tubes indicated a poten-
tial 10 fold improvement in detection of scattered blue light but only compar-
able performance in the infrared region. The increased technical requirements
and expense of these detectors outweighed this performance advantage. Silicon
avalanche detectors were rejected because commercial units required more
electronic support and could only match the performance of the common, large
area silicon detectors used in the final design.
The fiber optic bundles have stainless steel jackets and incorporate
commercial fibers with specially molded ends secured with Emerson & Cummings
2762 FT high temperature epoxy (good to 315 C°). Of possible vendors, only
Dolan-Jenner, Inc. had stock molds for producing tight bends and custom
dimensions needed in the probe.
26
-------
rDetectors
\ rFiber Optic
\ \ Bundles
Collection Lens for
Forward Angle—, Scatter
Beam Monitor
Xenon Lam
lens
Rotating Chopper
Wheel (15 cps)
Indexing Filter Wheel
Neutral Density Filter-
(Only with 900nm Filtep*X
Interference Filter /
Polarizer-^
Fiqure 18. Schematic Drawing of Optical System
Probe
Exit Hoii
Sample Slot
tack Gas
ELECTRONIC FILTER
The power of the light scattered by particles, collected by
:ics and relayed to the silicon detectors is as little as 10~15
the fiber
optics and relayed to the silicon detectors is as little as 10~12 watts.
Model HUV-4000B silicon detectors, made by EG&G, were selected because
these detector-amplifier assemblies offered signal gains in the order of 108
and low level noise by virtue of component selection and hybrid construction.
To improve noise rejection each amplifier is followed by a switched filter
circuit which responds to the frequency of the chopped probe beam (15 cps)
but rejects higher and lower frequency signals and has an effective band-
width in the order of 10 cps.
The circuit is shown in Figure 19 and, operationally, consists of a
switched up/down integrator to achieve a low-noise response. Light chopped
at 15 Hz impinges on the photodiode detector and the diode amplifier charges
27
-------
and discharges the coupling capacitor (lOOyf) without saturating. (The cap-
acitor negates the amplifier drift and off-set.) The integrator's input IT
is sequenced by switches Si and S2 between outputs OA and -0. to achieve
up/down integration of the scatter signal. These switches are synchronized
with the chopper so that when the light beam is projected, the signal
(scattered light) is added in the integrator and, conversely, when the light
beam is blanked, the signal (electrical background) is subtracted. Since the
chopper timing is symmetrical, the integrator retains a value proportional to
the scatter signal and cancels out longer term background variations, detec-
tor off-set, and low frequency noise.
PURGE SYSTEM
A single blower is used both to provide cooling air for the electronic
and optical components in the transceiver and to deliver filtered air to the
air curtains which protect optical surfaces in the probe as shown in Figure
20. In operation, the blower draws approximately 20 CFM through the
transceiver, through a flexible hose, and into a high pressure, high volume
regenerative blower (Rotron SL4P2). The output of the blower is split;
about 2-8 CFM is driven through air delivery hoses to the probe and the
excess volume is exhausted directly to the atmosphere. The inlet and out-
let openings in the transceiver are sealed with filters so that only filtered
air flows over components in the transceiver and twice filtered air passes on
to the blower and the probe. The mounting of the filters to the transceiver's
housing provides continuous protection for components in the transceiver
even when the hose to the blower is detached. The Balston (Grade D) filter
tubes, in this configuration, provide 99% retention of 0.3 - 0.5 micron
particles and greater retention of both larger and smaller particles.
The velocity of the gas in the duct determines the flow rate of the air
to the air curtains in the probe which is controlled by an orifice and the
exhaust valve. When duct velocities are low, a small restricting orifice
is employed to limit the flow to the probe and yet permit high volume flow
through the transceiver and blower for cooling purposes. (High flow through
the blower prevents thermal cut out of the Rotron motor.) When higher duct
velocities prevail, less restrictive orifices are used to provide greater
flow to the probe. Finer adjustments are made with a manual exhaust valve.
After the orifice at the blower, the flow to the probe is divided and
delivered through five 1^ cm diameter hoses and then, inside the probe,
through 1% cm stainless steel tubes to the air curtains. Balancing of the
flow to each air curtain is accomplished by throttling orifices located in
the delivery connections to the probe. The large delivery hose and tube
diameters produce moderate pressure drops, in the vicinity of 10-40 inches,
which serve to moderate the effects of pressure fluctuations in the stack.
28
-------
Figure 19: Schematic Drawing of Electronic Filter Circuit
Transceiver
Housing
Exhaust Valve
Hoses
Probe
Air Curtail
— Flexible
•Filter Hose
Throttling
Orifices
Figure 20: Schematic Drawing of Blower and Pneumatic Components
29
U.S tP
Mail code i-rl-V:
1200 Pennsylvania Avenue UW
Wasnlngton, DC 2046C
2C2-566-G556
-------
AIR CURTAINS
The optical surfaces which look into the sampling slot in the probe are
sealed behind glass plates which are protected from fouling participate by
air curtains. The air curtains extend the time between cleanings to approx-
imately 20-40 hours depending on stack conditions. The curtains are unusual
in that they pass widely diverging light rays, are located in close proximity
to the sampling area without blowing away sample, and are small enough to
fit into the available space in the probe. The curtains are formed by only
a small amount of purge air which is made to circulate non-violently in a
mixing chamber and then exhaust into the gas stream. In operation, most
stack gas and particulate bypass the opening to the mixing chamber while a
little seeps into the chamber where it is entrained by circulating purge air
and carried back into the stack. Inevitably, a little particulate remains
in circulation and deposits on the window generally protected by the mixing
chamber.
A basic air curtain (mixing chamber) is shown in Figure 21. It consists
of an orifice opening into the stack (or sample slot), a relatively large
mixing chamber, a window to be protected, and an inlet for purge air which
directs clean air over the face of the window. The direction of incoming
purge air cooperates with the flow of the stack gas (beyond the orifice) to
encourage circular flow in the mixing chamber. The volume of purge air is
adjusted to generate a mixing line between the stack gas and purge air of 10
degrees as indicated by 0. Higher purge flow rates (and greater mixing
angles 0) blows sample further away from the opening but does not appreciably
improve performance. Reduction in flow rates to half this volume (and a
mixing angle of 5°) results in only modest loss in performance. Still lower
rates, producing mixing angles of less than 5°, allow substantial migration
of stack gas into the mixing chamber and significant loss in effectiveness.
The air chambers (air curtains) used to protect the collimating lens
and the collector lens at the tip of the probe employ the above design with
the addition of one or more orifice plates in the mixing chamber. These
plates create serial mixing chambers and multiple improvement in effective-
ness as shown in Figure 22. In wind tunnel tests, the addition of an inter-
mediate plate in the tip mixing chamber reduced the rate of contamination
by a factor of ten.
The air chambers used to protect the windows over the fiber optic
bundles at 90 degrees to the slot operate on the same principle but were
drastically modified to meet severe space limitations. In these instances,
the 5ha1 lev; depth dimension permitted only a shallow mixing volume that would
not maintain consistent circular flow. The required flow is mechanically
established by inserting a baffle which confines the incoming purge flow to
a narrow dimension and imposes a tight, 180 degree bend on this stream as
shown in Figure 23. A short overhanging lip completes the small chamber.
In trials, the air curtain proved more effective with the lip than without
one.
30
-------
Mixing Line
Stack Gas
Mixing
Angle
Purge Air
Purge
Outlet
Window
Circular
Flow
Figure 21. Basic Air Curtain Design and Gas Flov.1
Mixing
Chamber
Stack Gas
•Orifice
Orifice Plate
Wi ndow
Purae Air
Serial
Mixing
Chamber
Figure 22. Air Chamber and an Additional Orifice Plate Used to Create
Serial Mixing Chambers
31
-------
Purge Air
Overhanging Lip
Window
Baffle
Figure 23. Air Chamber with Baffle to Confine and Direct Air Flow
SUMMARY DESCRIPTION
The entire instrument is shown in Figure 24. To sample inside a duct,
it employs a 1.5 meter long, 9 cm diameter stainless steel probe with a
36 x 3 cm sample passage near its tip (Key 1). A beam of light is pro-
jected down the axis of the probe, through the sampling slot, and out
through the tip of the probe so as to avoid reflections within the slot.
Located about the sampling area are five high temperature fiber optic
cables which collect light scattered at three forward angles and at two
orthogonal positions 905 to the beam. Glass windows protect a lens
located in the tip of the probe (which collects scatter at forward angles)
and the faces of two 90° fibers which look directly into the slot. These
windows are protected by air curtains which are formed by aerodynamic
cavities and metered flows of filtered, purge air. Within the duct, the
probe is oriented so that gas passes straight through the slot and its
direction satisfies the one-directional requirements of the air curtains.
32
-------
In the transceiver (Key 2), a xenon arc source and a rotating disc
generate a beam chopped at 15 Hz. The beam passes through interference
filters (with center wavelengths of .45 and .90 microns) and linear
polarizers mounted on a filter wheel, through a pinhole, a collimating
lens, and a protecting window and air curtain. A geneva mechanism indexes
the filter wheel at one second intervals causing the polarized beam to
alternate from blue to infrared light while the pinhole (and micrometer
adjusting screws) centers the beam in the exit tube at the tip of the probe.
Light scattered in the sampling slot is returned by five fiber optic
cables to silicon detectors located in the transceiver. A sixth detector
collects reflected light from the beam. Filter circuitry associated with
the detectors is synchronized with the chopping rate of the beam so as to
negate extraneous noise and ambient light. The transceiver is enclosed by
a light tight cover with four filter protected ports which pass cooling
air.
The control console (Key 3) includes a microprocessor which reads
twelve detector signals from the transceiver (five angular scattering
signals and one beam strength signal for each of two colors). The signals
are combined into five volumetric responses which are then decoupled using
a matrix routine to produce a five channel histrogram of loadings on the
printer (Key 4). Total loading and mean particle diameters are also
reported on an LED display. Since computations normalize signals to beam
strength, results are insensitive to minor contamination of the beam
generating optics in the transceiver. Operator controls include a function
switch for measuring and storing background signal values or making measure-
ments in the stack. Five span ranges between 0-40 ppm and 0-400 ppm and
eight sample periods between 2 and 256 seconds are available. To accomodate
precipitator rapping within a sample period, individual one second samples
which exceed the measurement range are excluded from computations but are
reported as bad samples at the printer. Power to all subassemblies except
the lamp is controlled at the console panel.
The air blower (Key 5) is used both to draw cooling air through the
transceiver and to deliver purge air to the air curtains in the Probe. The
xenon lamp (in the transceiver) is driven by a lamp power supply (Key 6)
which is a standard unit (Model CA-75-8267, Illumination Industries, Inc.)
packaged in a commercial drip proof enclosure.
33
-------
I
Figure 24: Photograph of Stack Probe
including all components.
-------
SECTION 5
TESTING OF THE PROTOTYPE
This portion of the report describes several phases of calibration and
testing of the completed prototype. The first section discusses the theo-
retical calibration of the instrument as built. The next section describes
controlled in-house testing of the unit to verify its performance. Sections
three and four describe tests performed at Research Triangle Park at the EPA
controlled source laboratory, and tests carried out at an East Coast coal-
fired facility. During the testing phases several changes and modifications
took place, and the most significant of these are described in the last section.
THEORETICAL CALIBRATION
This section summarizes the calculation procedure used in the final
calibration of the EPA prototype stack monitor. In the process of calib-
ration several assumptions were made, as listed below:
1. The particles are spherical6, and rigorous scattering theory
(Mie theory) is used to predict scattered light distributions.
2. The refractive index is that of silica, n = 1.5 - O.Oi, a common
stack material.
3. The particle distribution is to be represented on a log diameter
scale, within the range of 0.1 to 10 microns.
4. Five equal log increment intervals are to be measured, and within
each interval, the distribution is approximately flat.
5. The density of the material is 2.5 grams/cm3, which makes the
loading of 0.01 to 0.1 grams/m3 equivalent to 4 to 40 parts per
billion by volume.
6. The electronic gains will be set to anticipate 40 parts/billion
in any one of the five channels.
Two computer programs incorporating Mie scattering subroutines
(supplied by IBM5 and modified at L&N) were used for this analysis.
MIESCAT: This is a forward angle program which computes the flux in
watts which passes through an aperture in the Fraunhofer plane of a forward
35
-------
scattering optical system when a particle is illuminated by one watt per
cm2 of unpolarized light. Cosine fourth losses are included. Refractive
index is an input parameter.
MIE: This program generates the Mie Coefficients ii and i2, which are
functions, the refractive index, n, the angle, 0, and the size parameter, a,
which is
a = TrdA (5)
Summation over angles is possible with individual loss coefficients, pro-
vided the coefficients and angles are symmetric about 90°. When one part-
icle is illuminated by 1 watt per cm of polarized light, the scattered
intensity per particle, per unit solid angle is
Air*
(6)
with A in cm.
Calculations were performed to predict signal levels to each of the
five detectors in terms of radiant power for a given loading. The particle
size distribution used for this analysis is flat on a logarithmic scale,
with 1 part per billion by volume in each of the five channels. The sample
regions are listed in Table 4 by particle diameter. The actual diameter
values are taken from computer generated diameter intervals, with adjacent
diameters having a ratio of 1.1297.
Table 4: Histogram Intervals for Prototype Instrument
Designation
A: "0.2"
B: "0.4"
C: "1.0"
D: "3.0"
E: "7.0"
Size Range
(Diameter in Microns)
0.114 to 0.268
0.302 to 0.710
0.802 to 1.884
2.128 to 4.998
5.646 to 13.259
Log Interval
.423
.423
.423
.423
.423
Measurement Wavelength
(Microns)
0.45
0.90
0.90
0.45
0.90
The three forward scattering channels range from 3 to 11 degrees.
Descriptions of the collection apertures, as built, are listed in Table 5.
-------
Table 5: Forward Channels
Angle
3.9°
7.8°
11°
Instrument
Designation
F5/45/95
F4/44/94
F3/43/93
Radial Dimensions
(Inches)
.0477 to .0667
.0943 to .134
.134 to .196
Width
(Inches)
.044
.103
.103
Area (cm2)
.00539 cm2
.02638 cm2
.04120 cm2
Each of the zones was modeled using the MIESCAT program. Since the
aperture definition for the program is based on a radial coordinate system
and these apertures are rectangles, a slight adjustment of inner and outer
radius was made, preserving the total area of the aperture.
Computer runs for an index of n = 1.5 - O.Oi were generated, and
within each size range the average was calculated by summing the individual
flux per unit volume of material values and dividing the number of log
intervals.
Table 6 provides a list of parameters used in computing the power
incident on each detector.
Table 6: Symbol Definitions
Symbol
W
L
V
d
ID
T
F/d3
V
Description
Radiant Power Striking the Detector Face
The loading in parts/billion
per log interval channel
The volume of material both illuminated and seen
by the detector (including vignetting losses)
Particle Diameter
The intensity of the source, as unpolarized light
The transmission of each window/lens/fiber train
The average power coefficient/d3 per unit log
interval, as computed from the MIESCAT program
The volume of one spherical particle
Units
Watts
Vol ume/Vol ume
cm3
Microns
Watts
1 /micron3
Micron3
-------
The number of particles observed is expressed as
„ - L V
no - -
101* microns
cm
(7)
The calculation of W is done for each angle, 0, size interval, S, a:id wave-
length, X. The resulting formula is
W(0,S,X) = L V(0) rF(e.S,X)/d3"|6|'lO't microns]3 I0(X) T(X,6) (8)
L -I IT I- cm •*
The most difficult part of this procedure is the determination of common
volume, V, including the effects of vignetting. Computation of the common
volune is quite complicated with a non-point detector due to the off axis
geometry of the lens system. An experimental approach was attempted, based
on comparing the signal scattered for the actual aperture to the signal
measured at a region of no vignetting for a 220 micron pinhole aperture.
Table 7 provides a summary of the experimental and theoretical results
and also includes the experimentally determined transmission functions of
lens/window/fiber combination. These transmission losses did not contribute
to the experimental uncertainty in common volumes, since the calculations
used ratios of signals at each wavelength.
Table 7: Volume and Transmission Factors
Channel
3 (11°)
4 (7.8°)
5 (3.9°)
Common Volume
(Experiment)
5.5 + 1.5
8.2 +_ 2.0
17 + 4.0
Common Volume
(Calculations)
7.1 cm3
10.1 cm3
12.3 cm3
Transmission
Blue (0.45) Red (0.90)
.201 .334
.232 .372
.205 .352
Finally, the power of the probe beam was measured as .485 milliwatts in blue
and .412 milliwatts in red. This was polarized light, so the equivalent
intensity was 1/2 the measured intensity. When distributed over the 1/2
inch diameter beam, the intensity, Io> was
T 1.914 x 10'" watts/cm2 (0.45 microns) (9)
10 1.626 x 10"" watts/cm2 (0.90 microns)
38
-------
Using these coefficients, the power levels, W, were computed as listed in
Table 8.
Table 8: Watts of Radiant Power for 1 Part/Billion per Channel
Wavelength
0.45y
(.485 x 10"3 watts)
0.90v
(.412 x 1(T3 watts)
Size
A
B
C
D
E
A
B
C
D
E
3.9°
5.7968 E-12
5.8733 E-ll
5.576 E-ll
5.934 E-ll
5.596 E-12
4.292 E-13
1.0471 E-ll
4.948 E-ll
4.602 E-ll
3.029 E-ll
7.8°
2.496 E-ll
2.407 E-10
1.631 E-10
4.307 E-ll
8.771 E-12
1.678 E-12
4.185 E-ll
1.820 E-10
9.700 E-ll
1.017 E-ll
11°
2.181 E-ll
1.975 E-10
8.863 E-ll
1.374 E-ll
2.952 E-12
1.5245 E-12
3.804 E-ll
1.471 E-10
4.189 E-ll
6.658 E-12
For the polarization measurements, two fiber bundles are oriented at
90° to the incident beam. The optical axes of these two fibers are perpen-
dicular to each other, so the fibers see the two orthogonally polarized planes
of scattered light.
The fiber faces are 0.5 inches (along the direction of the probe beam)
by 0.024 inches, for a surface area of 7.742 x 10~2cm2. No lens or optical
system is used, so the fibers collect over a wide angle, from 50° to 130°,
with most of the sensitivity between 60° and 120°.
Using the expressions in (6) for scattered flux per unit solid angle,
W, the power incident on each detector can be computed from,
W(p,S,A) =
(10)
Table 9 provides a list of parameters used in this equation.
39
-------
Table 9: Terms Used in 90° Calculations
Symbol
W
L
V
V
a
T
TP
1P
A/R2
Description
Radiant power striking detector face, as function
of size, S, polarization, p, and wavelength, A
Loading, in parts per billion
per log interval channel
Volume of region seen by fiber
Volume of one spherical particle of diameter, d
(Alpha) - dimensionless size parameter, irdA
Transmission of fiber bundle
Intensity of polarized light
ii or i-2» as defined in equation 2
Solid angle subtended by fiber face, with A,
the area of the face and R the distance
to a scattering object
Units
Watts
—
cm3
cm3
—
—
Watts/cm2
1/Steradians
Steradians
Due to the wide field of coverage, V, R, T, and ip are all functions of
the collection angle. Using the computer program, MIE, a numerical integra-
tion was performed over the range of 50° to 130°, for both polarizations, and
the response per unit volume was summed over the appropriate log interval
ranges.
The individual R- and V. values were computed from a scale drawing of
the collection system1. Experimental data on the fiber transmission was
measured as a function of wavelength and angle. The MIE program then was
used to calculate
.Mn(a) = 10 £ E 1p(9) V(9) T(6)
P a 0 asR(6)£
(11)
From this, equation (lo) could be evaluated readily, as,
- T 3 i A * 10** microns 1 1 M , \
- TP 2 L A X cm 8 TO Mp(a)
(12)
40
-------
based on the relationship
1 1 6TT2
7 "5P" T7"
(13)
The 1/8 factor in equation (12) is due to the summation in M over 8 inter-
vals. The results are given in Table 10, expressed as watts of energy for
1 part per billion in each size range.
Table 10: 90° Signals for 1 Part/Billion
Size
A (.2)
B (.4)
C (1.0)
D (3.0)
E (7.0)
A (.2)
B (.4)
C (1.0)
D (3.0)
E (7.0)
1i 90°
1.4912
9.4152
3.4194
8.8710
2.0620
1.8457
7.9515
'3.0741
7.8898
2.5030
E-ll
E-12
E-12
E-13
E-13
E-12
E-12
E-12
E-13
E-13
1? 90°
4.0539
1.1577
3.8246
8.0443
2.5974
1.8271
3.5484
4.2514
1.3776
2.6722
E-12
E-ll
E-12
E-13
E-13
E-13
E-12
E-12
E-12
E-13
(Watts of Power at Detector)
The computed signal levels were then used to anticipate what signal
ranges would be encountered in the instrument, so that the individual detec-
tor gains could be set. Analysis of published data suggested that we might
encounter relatively narrow distributions, so the gains were chosen to
accomodate the maximurrloading (40 parts per billion) in any one channel.
A preference was expressed for gains that were simply related to one another,
to facilitate computations and calibration.
The relative gains are listed in Table 11, column 2, These gains were
set by adjusting feedback resistors, and the experimentally measured values
are given in columns 3 and 4.
41
-------
Table 11: Detector Gains
Angle
3.9°
7.8°
11°
90°Pl
90°P2
Selected
Gain
4
2
2
40
40
Experimentally Determined
Red
4.47
2.34
2.34
40.00
42.88
Blue
4.29
2.184
2.046
40.00 .
40.87
The next step involved adjusting the scattering equations for the various
gain, transmission, and common volume differences. Then it was necessary to
correct each volume response equation to produce the same signal for unit
loading at its peak region. The results are summarized in Table 12,
Table 12: Preliminary Channel Equations
Channel (Size)
Multiplier x Formula
A (.2)
B (.4)
C (1.)
D (3.)
E (7.)
2.143 x (1.0(41)- 0.8@)/(46)
4.89 x (1.0©- 0.8(92))/@
6.013 x (1.0(93)- 0.3686(94) + 0.0317(95)
5.56 x (0.3519@- 0.7832@+ 1.0©;
9.168 x (0.3739(93)- 0.8387(94)+ 1.0 (95) ) / (9~6)
The circled values represent signals from the respective colors and
detectors, including gains. No correction is needed for relative spectral
sensitivity of the detectors, since the ratio removes spectrally induced
gain differences.
MATRIX DECOUPLING
There is some spillover of response between adjacent channels, and to
reduce this problem, a matrix decoupling is applied. The spillover is
computed by normalizing the response functions into five equal log interval
regions. This produced a five by five matrix of values. The inverse of this
matrix provide^ the sequence of numbers to unfold a more accurate distribu-
tion from the measured one. The coupling matrix and its inverse arc
in Table 13. This decoupling is applied after the individual channel
42.
-------
signals, C^, are measured. The final histogram components are designated as
HI through He-
INVERSION 0r A MATRIX WITH 5 VARJARLES
l.COOC »C130 .0310 .0210 .0000
•338C ItCOUO -.065C -.0620 ,0070
tC125 .27*0 1.0000 .1090"" .0580
.0000 .OQ4Q .1760 1.0000 .0590
.0000 -.00^0 .0250 .4830' 1.0000
THIS IS JHE INVERSE 8F THr MATRIX
1»OC21 -.0053 -»028C -.0197 .0028
-.3343 .9868 »0628 .Q684 -.0146
• C803 "-.2745 :".999i - • i035 ' " • . 04g9
-.0130 .0451 ".'.-tl798 1.0476 -.0517
.0036 -.0130 .0620 -.5033 1.0262
NORMALIZED RESPONSE
FOR
0.2 (Microns)
0.4
1.0
3.0
7.0
X
"c/
C2
C3
d,
C5
r:
"H/
H2
H3
H4
H5
Table 13: Normalized Response for Each Zone and the Inverse Matrix Used for
Decoupling.
IN-HOUSE CALIBRATION OF THE COMPLETED UNIT
Calibration of the finished unit was complicated by several factors. The
most difficult portion was obtaining a sample material covering the size range
of interest that was sufficiently well characterized to be used as a reference
standard. Several iterations were made before coming to a satisfactory mat-
erial and set of instrumental constants which gave consistent results.
Initially, attempts were made to use suspensions of diamond dust in
water. A specially designed cuvette, 35 cm long and 2.5 cm wide was
built to hold the suspension, and a significant amount of effort was put
into sizing the diamond dust distributions using electron microscopy.
However, there are some optical difficulties involved with liqud sus-
pensions that had not been recognized until the actual testing began.
During the measurements, two artifacts of the cuvette system were discovered.
43
-------
The longer wavelength (0.9 microns) is severely attenuated by the
path through water. Approximately 90% of the light is lost in each
of the three forward channels to absorption by the water. Accord-:
ingly, all forward measurements at this wavelength appeared init-
ially to be only 10% of expected value.
The refraction that occurs at the air/glass/water interface is
accentuated by the off-axis position of the forward scattering
system, as illustrated in Figure 25. Refraction changes the
effective common volumes, slightly increasing the observed volume
in the 11 and 7.8 degree channels, and substantially reducing
the illuminated volume observed by the 3.9 degree channel. These
effects may be color dependent, since beam cross-section is not
identical in both colors.
Water Path
Air Path
Fimire 25: Comparison of Forward Scattorinn Connon Volumes vith and Without
Water in Cell
In addition to diamond dust, polystyrene spheres in two size ranges,
1.1 microns and 5.7 microns were evaluated in liquid suspension. The
results of the diamond emu polystyrene tests were consistent with th«> arti-
facts defined above, but it was not possible to correlate them with theory
44.
-------
without doing extensive additional testing. The use of water suspended
material was then abandoned in favor of aerosol suspensions, consistent with
the intended use in the stack.
The aerosol experiments were performed to determine the combined effect
of detector gain, fiber transmission, common intercepted sample volume, and
aperture size for each of the forward scattering channels, and to verify the
balance between red and blue signals.
Considerable thought went into the choice of a proper aerosol. For
particles larger than 1 micron, the Mie scattering properties show reson-
ances, and accurate determination of particle size to within a few tenths
of a micron is essential. In addition, the refractive index of the material
effects the resonances. The index should be known to within 0.01 for
accurate analysis of the data if monodisperse particles are used for calibra-
tion.
The absolute index and the particle size are much less critical for
scattering material in the Rayleigh region. Strictly, this is on the order
of 0.1 micron or smaller, but in practice, the transition region between
Rayleigh and Mie scattering allows some flexibility in the upper size limit
of a broad sub-micron aerosol used for calibration, as will be discussed.
Rayleigh scatterers are desirable for these experiments, since the
ratio of any two forward scattering signals is independent of diameter.
Therefore, the actual particle size distribution need not be known when all
the particles are in the Rayleigh region. Figures 26 and 27 show the forward
scattering signals vs. diameter for small particles at wavelengths .45y and
.90y, respectively. The region of constant slope is characteristic of
Rayleigh scattering. The upper limit is not well defined but occurs for
diameters of approximately 0.4y arid 0.8y, respectively. Below this size
the ratios of any two forward scattering signals are identical at both
wavelengths. This is a convenient method of determining if an aerosol meets
Rayleigh scattering criteria.
Four different sources of small particles were investigated; the Phoenix
aerosol generator, cigar smoke, ammonium chloride smoke, and cigarette smoke.
The distribution of the Phoenix generagor is specified by the supplier to be
92% below ly. This distribution can be shifted to smaller diameter by
diluting the aerosol source material, OOP, with isopropyl alcohol. The
alcohol in each primary droplet evaporates, leaving a OOP droplet of reduced
size. For example, a 1% solution of OOP in isopropyl is expected to reduce
the particle size by a factor of 0.215. One percent and 0.5 percent
solutions were used in the experiments.
45
-------
-p.
01
10
,-q
5 10
o
ID
c
'11
I
.13 .20 .32 .50 .30 1.27 2.20
DIAMETER IN MICRONS
Fl9ure 26: Flux/Unit Volume vs. Particle Diameter
for OOP, Aerosols, X • 0.45 microns
•io-'«
01
I
I
.13 .20 .32 .50 .80 1.27 2.02
DIAMETER IN MICRONS.
Figure 27: Flux/Unit Volume vs. Particle Site
for OOP, Aerosols, X • 0.09
-------
Cigarette smoke has been measured by many people using different
techniques. Most references place the size well below one micron, and one
recent article defines the mass mean diameter of total particulate matter
in the 0.5 to 0.75 micron range7 (filtered cigarette). Ultimately,
cigarette smoke was found to contain the smallest size particles, and was
used as the technique for calibrating the forward scattering channels. Cigar
smoke was also tried, but subsequently determined to be bimodal.
The system used in the laboratory to produce the smoke aerosols is
shown in Figure 28. A filtered cigarette was connected to the inlet of a 1
cfm diaphragm pump. The output of the pump filled a 2.4 cubic foot mixing
chamber before entering the probe sample region, to promote homogeneous
aerosol concentration throughout the probe. The purge system could not
be used at such low flow rates. Therefore, microscope slides were placed
over the 90° apertures and at each end of the slot to contain the aerosol to
the sampling region. The same procedure was used with the Phoenix aerosol
except that the flow was powered by a compressed nitrogen tank.
Plexiglass
Window
Sample Volume
Bleed £
Valve
Exhaust
EPA Probe
Diffusion Grating
Mixing Chamber
N2 25 psi
Phoenix Aerosol
Generator
Figure 28: Laboratory Sample System for Delivery of Aerosol Smokes
47,
-------
Data on all three samples was taken at various smoke concentrations
testing for multiple scattering by checking whether all angular scattering
signals increased by the same percentage as the loading increased. No mul-
tiple scattering was detected.
Uniform sample flow was very important, since each of the three forward
angles intercepts a different volume of material in the sample region.
During the experimentation, a plastic window was inserted on the exit side of
the sample chamber, after the probe, to permit visual inspection of the
material flowing through the sample slot. Occasional turbulence and non-
uniformities in flow were observed, and the sample flow system was changed
to eliminate those problems.
The 7° and 3° signals were divided by the 11° signal in each of two
colors to cancel the effect of particle loading. Comparisons to theory were
then made on the basis of these ratios. Then, since true Rayleigh scatter-
ing scales as inverse 4th power with wavelength, the ratio of blue to
red at any low forward angle indicates the extent of Rayleigh scattering.
The ratio at 11° was used throughout. These five parameters are compared
in Tables 14 and 15. Table 14 lists the results predicted by the preliminary
theoretical constants for particles of .1, .4, and .5 microns diameter.
Notice that 3° and 7° ratios change very little with particle size, but the
blue to red ratio changes rapidly. The experimental data is listed in
Table 15. This data shows an interesting trend. As all the normalized sig-
nals decrease from left to right, the blue to red ratio increases. This
behavior is consistant with decreasing particle size from left to right.
The only exception is cigar smoke data which appears to be bimodal.
None of the experimental data exhibited the full inverse 4th power
wavelength dependence characteristic of pure Rayleigh scattering. However,
cigarette smoke came closest and is in the range predicted by values at the
bottom of Table 14. The blue to red ratio of 11.4 suggests that most of
the cigarette smoke is at or below .5y in diameter. This size was sub-
sequently confirmed by the 90° scattering data which indicated 85% volume
in the 0.4y channel and 15% in the 0.2 channel. In this range the blue to
red ratio is sensitive to particle size but the forward scattering normalized
signals are not.
48
-------
Table 14: Theoretically Predicted Ratios for Three Particle Sizes
Size (Microns)
X = .45y
7°/n°
3°/H0
X = .90y
7°/H°
3°/H°
Blue/Red
11°
0.1
1.100
.491
1.063
.507
20.
0.4
1.133
.517
1.069
.511
15.
0.5
1.202
.544
1.105
.514
10.2
Table 15: Experimental Data with Various Aerosols
x = .45W
7°/n°
3°/n°
X = .90
7°/H°
3°/n°
Blue/Red
11°
Cigar
1.058
.558
.979
.398
1.99
1% OOP
1.033
.412
1.142
.541
4.27
0.5% OOP
.966
.405
1.102
.418
5.08
Cigarette
.914
.315
.940
.306
11.4
Cigarette Data
at 0.5 microns:
Theoretical/Experiment
1.315
1.727
1.175
1.680
49
-------
Table 16: Final Constants for the Prototype, as Built.
The matrix (Table 13) remains unchanged.
Cl = .605 (@- .7462@)/@
C2 = .520 (©- . 7830 (92) )/ (96)
C3 = .639 (1.0(93)- .4331(94)+ .0533 @ )/(96)
Ck = 1.570 (0.3619@- 1.030(44) + 1.727 (45) )/
C5 = .974 (0.3739(93)- 0.9855(94)+ 1.680 (95) ) /
Therefore, by comparing the 0,5y theory and experimental cigarette data,
as shown in Table 15, corrections to the original equations can be made,
corresponding to the performance of the instrument as built. The five for-
ward scattering equations are listed in Table 16.
Figures 29 and 30 present instrumental measurements on materials made
during the final calibration. Figure 29 shows cigarette smoke runs at
several different concentrations. This was the only material investigated
that consistently showed all sub-micron particles. Separate experiments
were performed with a combination of ammonia and hydrochloric acid fumes,
which combined to form a white cloud of ammonium chloride smoke. Although
reported to be submicron material under ideal conditions, agglomeration
is claimed to be a problem, and this seemed to be the case for our samples
as prepared in the test system. Two different arrangements of the sample
presentation were attempted.
Figure 30 shows the progression of the increase in size with changing
concentration in a combination of di-octyl phthalate (OOP) and alcohol.
There is occasionally a small amount of material shown in the largest size
channel, which may be artifact of the system, or may represent a truly
bimodal distribution.
The bimodal nature of at least one material, cigar smoke, was con-
firmed by separate analysis of the scattering data. Measured data is
shown at the bottom of Figure 30.
50
-------
QJ 1001
CT>
ro 80 -
s-
s: 60 4
o
S 40-j
20 J
10
«c
Cigarette Smoke
dV • .46
.2
.4
1.0 ' 3.5
7.0/1
Particle Diameter Microns
o
ro
100X,
80 •
60 .
40 -
20 .
: Cigarette Smoke
dV - 1.22
.2 ' .4
1.0 ' 3.5 ' 7.0
Particle Diameter Microns
CD
C
ro
o
ro
O)
c
to
tn
1001-
80 .
60 •
40 .
20 •
Cigarette Smoke
dV - .41
1 * K
in'^c 1 7/*"1
.2
0
c.
I
0
ro
Q>
V}
loot-
80 .
60 .
40 -
20 -
Cigarette Smoke
dV » 1.51
•3 I *
in ' •> c 'in *
Particle Diameter Microns
JQ ••• ••• i.u J.J /.U
E
&« Particle Diameter Microns
O)
ro
s-
0)
lOK-i
ao •
60 .
40 .
20 .
NH..CL Agglomerated
•
CJ>
c
ro
o
ro
-------
en 1001 1
c
2 80 .
"5 60 '
03
V 40 •
C
10°*1
c
« 80.
i_
-j= 60.
(_>
IT3
<" an.
HU •
c
'•" 20-
io
•a
23.4% OOP
f
' .2 ' .4 * 1.0
1 — 1
3.5 7.0
** Particle Diameter, Microns
OJ
CD
c
to lias*
s-
.c CO -
(J
fO
S 60 -
= 40.
y, 20 .
to
100S OOP
1
.2
.4
1.0 3.5
7.0 '
Particle Diameter, Microns
CD
c
tc
^
-------
FIELD TRIAL AT EPA FACILITY
During the period of May 15 through 19, 1978, ten experiments were
performed at an EPA source simulator facility in North Carolina. Three
different size distributions of fly ash were measured in the simulator. The
simulator is a closed loop wind tunnel with accompanying subsystems to simu-
late source emissions. Particulate emissions are simulated by dispersing
powders into the tunnel with a fluidized bed particulate generator. A com-
plete description of the simulator and particulate generator is given in
reference 8. Size distribution data was recorded with the stack emissions
monitor and a seven stage cascade impactor. The impactor was located 5.5
feet up stream from the monitor. Tunnel flow was maintained at 35 ft/sec
for all runs.
Each sample of fly ash was measured three times, a total of nine runs.
The tenth run, which lasted one hour, tested the integrity of the purge
system for long term measurements. Table 17 lists the fly ash sample
number, span, sample time and duration of each experiment. The span in-
dicates full scale in parts per billion; it is inversely porportional to
gain. The sample time is the total measurement time for each of two wave-
lengths. The total time between data printouts is twice the measurement
time.
Raw data from the scattering signals was collected, but technical
difficulties with the microprocessor constants prevented on-the-spot
calculations of the size distribution. Many sets of optical scattering data
were taken for each impactor run, and transmissometer readings were made
continuously during the tests.
The amplitudes of the raw data from the optical scattering measurements
tracked very well with the transmissometer data, rising and falling to-
gether, as the loading varied during each test. Only one impactor sample
was collected per run, representing the average over the entire period.
After the final calibration was complete, the raw data sets from optical
scattering were processed by computer as they would have been with the
associated digital microprocessor. To compare the optically generated
signals to the distribution measured by the impactor, some manipulation of
the impactor data was required. It was corrected for a density of 2.3
gm/cm3, typical of fly ash, and graphed on a cumulative distribution plot.
53
-------
Table 17: Fly Ash Sample Tests at EPA-RTP Source Simulator - May, 1978
Test
Number
1
2
3
4
5
6
7
8
9
10
Fly Ash
Sample
Number
1
1
1
2
2
2
3
3
3
3
Span
(ppb)
100
100
40
100
40
100
TOO
100
100
100
Sample
Time
(sec)
64
64
64
64
64
64
64
64
64
128
Date and
Duration
of Test
5/17/78*
9:48 - 10:18
5/17/78
1:15 - 1:55
5/17/78
2:35 - 2:55
5/18/78
9:37 - 9:57
5/18/78
11:02- 11:27
5/18/78
1:10 - 1:25
5/18/78
2:05 - 2:20
5/18/78
3:00 - 3:10
5/18/78
3:40 - 3:50
5/18/78
3:50 - 5:00
Number of
Light Scattering
Data Sets Taken
11
17
**
9
**
7
7
4
4
No Data Taken
* Run made without glass beads in the fluidized bed particulate generator.
** Overrange conditions at one wavelength prevented acquisition of complete
54
-------
Next, channel edges, corresponding to those listed in Table 4, were ident-
ified and the mass was read in each of the five size regions measured by
the scattering instrument. Finally, the histogram of the impactor measured
mass distribution was computed by normalizing by the total mass in the size
range of 0.11 to 13.3 microns.
Graphs of five runs represent different size distributions used in the
testing are presented in Figure 31. The scattering data is shown as the
solid line histogram, and the impactor data is indicated by the dashed
lines and circles. Only one set of scattering data is graphed for each run,
but the size variations in most cases were minimal. The optical data
indicates that most of the material is in the range of 1 micron or greater,
while the impactor usually defines a distribution with a more well defined
peak but a tail extending over a broader range. Table 18 contains the
numerical results as graphed in Figure 31.
Difference between impactor and optical data as shown may be due to
the following things:
1. The two measurement techniques define boundaries for size fractions
in different manners.
2. The reduction of impactor data to optical scattering channels
introduces some errors.
3. The raw data printed out in 3 digit format and used in computer
processed optical data is not as accurate as that available to
the actual microprocessor. A 10% error in raw data for channels
1 and 4 (0.2 microns and 3.5 microns) is possible.
4. The sample collection procedure for the impactors showed different
results from the free stream state of the material. This is evident
in at least one case, runs 1 and 2, where the sample in the free
stream was consistently measured differently for the two runs,
while the impactor data showed them both very similar. Large
glass beads were added to the sample conditioner hopper for the
second and subsequent runs, and this appears to have changed the
characteristics of the air entrained material between otherwise
identical runs one and two.
55
-------
a,100
«.
o
£ *o-
.= 20-
Run 1
(Small Size)
e-
-
.2 ' .4 1.0 ^3.5 ' 7.0
Particle Diameter (Microns)
in
03
BO-
40-
20
Run 2
(Small Size)
.2 ' .4 ' 1.0 ' 3.5 ' 7.0
Particle Diameter (Microns)
til
CD
-C
0
to
O)
cl
03
O
tr:
c
(O
$-
o
(O
0)
(/)
in
£
100-
80.
60-
40.
20 .
Pa
100-
80.
60.
40.
20 •
Run 4
(Intermediate Size)
i
- o-i o .i- e- -
1.
.2 .4 ' 1.0 ' 3.5
rticle Diameter (Mi
Run 8
(Large Size)
p -e —
i
—o--. -o~- -o-
...
7.0
cron.<
-•-
(_)
C 100 •!
to
*~ 80.
o
S 60"
c 40-
10 20.
ra
0 ° P
Run 6
(Intermediate Size)
re~i
•2 .4 1.0 3.5 7.0
article Diameter (Microns)
II ~~" ll
-•- Impdctor Data II
J' _l|
s-e
3.5
7.0
Figure 31: Size of Fly Ash as Measured by Impactor and Optical Monitor at
EPA, RTP Source Simulator.
56
-------
Table 18: Impactor and Optical Volume Histograms
for May, 1978 EPA/RTP Tests
Channel Center
Particle Diameter, ym
Center
0.2
0.4
1.0
3.5
7.0
0.2-
0.4
1.0
3.5
7.0
Width
0.14 to 0.30
0.30 to 0.75
0.75 to 2.0
2.0 to 5.0
5.0 to!3.3
0.14 to 0.30
0.30 to 0.75
0.75 to 2.0
2.0 to 5.0
5.0 to!3.3
Impactor Run Number
1
3.6%
2.9
31.8
54.5
7.2
2
3.2%
3.7
31.0
54.7
7.4
4
4.9%
3.0
9.0
63.7
19.4
6
3.8%
4.9
10.0
62.0
18.0
8
7.0%
10.2
12.5
35.9
34.4
Optical Data Run
1-5
0%
5.5
26.3
26.0
42.2
2-5
0%
9.3
43.1
19.6
28.0
4»5
0%
5.2
16.3
29.4
49.1
6-5
0%
6.4
23.1
31.3
39.2
8-2
0%
1.3
0
16.4
82.4
57
-------
FIELD TRIAL AT A COAL-FIRED FACILITY
On June 27 and 28, 1978, the prototype on-stack instrument was tested
at an East Coast coal fired power plant. Optical data was to be taken from
a 15,600 mVrtlin horizontal flow duct, which led directly to the base of a
stack. In addition, Ray Steward of EPA and Bruce McElhoe of Northrop Services
(subcontractors to the EPA), took data with an eight-stage cascade impactor.
The tests were designed to determine the reliability, ease of use, and per-
formance of the probe under actual field conditions.
The week before the test we went to the site to clear out the two
test ports (one for the impactor and one for the prototype probe) and to
build a wooden deck to serve as a work platform during the tests. The test
site was outside, along the horizontal run of the duct, accessible by a
catwalk. The ports were difficult to clear and the positive pressure and
sulfuric acid fumes made this a very unpleasant task.
The instrument was set up on Tuesday morning, June 27. The xenon
lamp required several starts before it fired, but the high humidity that
date may have caused this. Vibration of the stack and the platform caused
the focus of the lamp to oscillate one half inch laterally in the pinhole
plane and final adjustment was difficult. The amount of beam misalignment
was not reproducible when the transceiver cover was attached. The system
was finally aligned through trial and error and varying tension on the
cover. Steps were later taken to correct this problem by adding adjust-
ment access ports to the cover, so that final adjustments are now made with
the cover on.
The stack flow was measured at 5.5 m/sec and the stack temperature
was 154°C. The impactor was inserted and we made a four minute measure-
ment. Then the probe was pulled out and allowed to cool for window inspect-
ion. Both 90° windows were clean, but the forward window had a thin layer
of fly ash. This layer did not cause appreciable optical attenuation.
The windows were cleaned and the lamp was restarted. The system was
aligned and probe inserted into the stack with the blower at a (dial
indicated) head pressure of 127 cm. This pressure was maintained while
waiting for the impactor, to avoid any fouling of the forward window. After
five minutes the blower shut off due to thermal overload; the high pressure
had caused the motor to overheat. Operating instructions now indicate the
proper operating pressures to avoid this problem. After the purge shut off,
the windows were exposed to flyash, and the probe was withdrawn from the
stack to clean the windows. The forward window showed the same amount of
dirt as when the purge system was operating. In later discussions with a
consultant in flow dynamics we decided that the fouling occurs during insert-
ion, when the positive pressure forces dirt laden air in through the beam
exit hole, near the forward windows. Purge operation during normal running
conditions seems satisfactory.
The windows were cleaned and new background readings taken. The probe
was inserted in the port and measurements werp recorded during two consecu-
58
-------
five impactor runs. The blower was run at a head pressure of 68.6 cm to
avoid fouling of the forward window. After these two tests, more data
was taken at a reduced head pressure of 43.2 cm to see the effect of the
purge on the common volumes. Table 19 summarizes all the test conditions.
Table 19: Test Conditions
Test
Number
1
2
3
4
Impactor
Number
1
2
3
—
Span
400
400
400
400
Function
1
3
3
3
Sample
Time
(Sec.)
64
16
16
16 and 32
Head
Pressure
48.3 cm
68.6 cm
68.6 cm
43.2 cm
Duration
12:20-12:25
4:47-4:52
5:00-5:03
5:09-5:19
A summary of the impactor data, adjusted as described earlier to fit
the optical instrument channels, is included in Table 20. The correspond-
ing optical data is shown below the impactor runs.
In these tests, there was a great disparity between optical and impac-
tor readings, with the prototype indicating approximately 87% of the mat-
erial in the largest size fraction while the impactor data shows a broader
distribution, peaked at 3.5 microns. No single factor explains the
difference but several observations were made at the site.
1. The loading was extremely high,
2. Some re-entrainment of fly ash was occuring from the collection
hoppers,
3. There was considerable turbulence and flow variation in the region
of measurement, and
4. The impactor and the optical probe were inserted in separate
ports, several feet apart.
The optical instrument was working properly before and after the test,
so we feel that the scattering data is accurate. However, the discrepancies
were larger than anticipated, based on the RTF tests in May, 1978. The
major interest in performing this test was to gain practical experience
with the instrument in that environment, and that goal was accomplished.
59
-------
Table 20: Field Test Data from Coal Fired Facility
Impactor Data
(Microns)
0.2
0.4
1.0
3.5
7.0
Optical Data
0.2
0.4
1.0
3.5
7.0
Run 1
3.4%
5.9
15.7
38.6
36.4
0.0%
2.2
3.5
7.3
87.3
Run 2
2.25%
4.5
18.3
45.0
30.0
0.0%
2.5
5.6
4.4
87.5
Run 3
1.14%
4.09
15.2
43.2
36.4
0.0%
2.6
5.9
4.2
87.3
60
-------
MODIFICATIONS DURING SYSTEM TESTING
During calibration and testing, a number of modifications and changes
were made to various parts of the system. The most significant ones are
described here.
1. A large random drift of the electrical background was discovered
and corrected. The cause was traced to high frequency electrical
noise at the ADC input, which was successfully eliminated with
the addition of a capacitor filter. Additional stability was
added by installing a lower resistance ground cable for the 50
foot connection between the lamp housing on the probe and the
digital microprocessor ccbinet. Backgrounds are now stable to
within one or two counts.
Other work involved elimination of unstable offset signals from
the ADC. These signals had a dependence on gain setting due to
a small tuning error. More careful tuning of the analog circuitry
and addition of a positive overrange trap eliminated spurious
signals when excessive scatter signals are present.
2. A high level of optical background in the 3.9° channel at 0.9
microns was traced to specular reflection in the interior of the
probe sheath. It was found that infrared illumination was being
scattered from dirt on the collimating lens and then reflected
from the sheath wall into view of that detector. This problem
was solved by proper sizing and location of intermediate baffles
inside the probe sheath. However, it is important that cleanli-
ness of the lens be maintained even though small deposits of
dust no longer drastically change the background.
In addition to final adjustment of the anti-reflection baffles, an
enclosure was installed around the chopper and filter assembly.
This was done to eliminate stray light from this part of the sys-
tem being reflected into the detector region.
3. During field trial testing, the main beam monitor showed excessive
sensitivity to alignment. Realignment after shipping produced
different relationships between beam strength and ADC counts, which
resulted in changes in calibration of the system. This problem
was solved by removing the fiber optic line and placing the silicon
detector at the focus of the reflected reference beam. This change
produced a much more stable beam strength signal, free from
variations due to realignment or residual flexing of the probe.
The entire focused spot, rather than a portion of the beam, now
falls on the detector.
4. There were several problems relating to the xenon lamp and power
supply. During testing, one of the two power transistors shorted.
Both were replaced with higher rated transistors. Additionally,
a plexiglass window was installed in the xenon power supply box so
the voltage and current meters could be monitored during the opera-
tion of the lamp.
61
-------
5. Arcing to the chassis was a problem when starting the lamp, so the
lamp housing was isolated from the chassis by adding a Delrin plate.
Additional electrical wiring changes were made to reduce the in-
duced current during the starting pulse. The lamp electrical
power now comes directly from the line cord to the power supply and
the starting controls were moved to the lamp power supply from the
control console. These and other associated changes were sufficient
to significantly reduce the arcing difficulties.
62.
-------
SECTION 6
CONCLUSIONS
Optical techniques involving low angle forward and polarization
dependent 90° scattering have been combined in an instrument package for
real time particulate sizing in stationary sources. An insertable probe
was fabricated, designed, and engineered for use in high temperature acid
environments. Associated equipment, such as an air purge system and
sophisticated electronics detection and microprocessor package, was
designed and constructed.
The response per unit volume of material for each of the five channels
is shown in Figure 32.
VOLUME
BESPON5E
0.10
s.o
IO.O
PARTICLE. DIAMETER
Figure 32: Volumetric response curves, as a function of particle diameter,
for the five channels of the prototype. Curves 1 and 2 are
obtained with 90° scattering, and curves 3, 4, and 5 from
forward scattering. These curves are for silica, with re-
fractive index of 1.5.
63
-------
A photograph of the completed prototype is included as Figure 33,
and the operational characteristics of the unit are defined in Table 21.
The instrument offers a method for in-stack analysis of particulate size
without dilution or extraction of material, and directly produces a
distribution of the particulates by volume fraction. This device thus
provides a unique and valuable measurement capability for analysis and
monitoring of particulates in source emissions.
Figure 33: Probe, electronics package, and digital printer used for
measuring the size of oarticulate material in utility stacks,
• ...
-------
Table 21: Characteristics of Prototype Light Scattering Instrument
Size Range (Particle Diameter)
Size Discrimination
Anticipated Loading Range
Operational Range
Mode of Operation
Power Requirements
Operational Temperature
0.1 to 10.0 microns
Five volume fractions with centers at
0.2, 0.4, 1.0, 3.5, and 7.0 microns
0.01 to 0.1 grams of material /meter3
or 4 to 40 parts/billion by volume
(with s.g. of 2.5)
4 to 400 parts/billion
Low angle forward scattering and 90°
polarization dependent scattering
115 volts, 60 Hz, 2 20 amp outlets
Probe portion in stack: up to 260°C
(500°F)
Transceiver portion and other electronics
outside stack: 0° to 43°C (32° to
Speed of Measurement Sequence
(Acquisition of Data)
Integration time selectable from 4 to 256
seconds
65
-------
REFERENCES
1. Mie, 6., Beitrage Zur Optik truber Medien. Annalen Der Physik, 3:
377-445, 1908.
2. Kerker, M., The Scattering of Light and Other Electromagnetic Radiation.
Academic Press, New York, 1969. 666 pp.
3. Van de Hulst, H.C., Light Scattering by Small Particles. John Wiley &
Sons, New York, 1957. 470 pp.
4. National Bureau of Standards. Tables of Scattering Functions for Spher-
ical Particles. Applied Mathematics Series - 4. U.S. Gov't Printing
Office, Washington, 1948. 119 pp.
5. Dave, J.V., Subroutines for Computing the Parameters of the Electro-
magnetic Radiation Scattered by a Spheres. Report No. 320-3237. IBM
Scientific Center, Palo Alto, CA, 1968. 65 pp.
6. Fisher, G.L., et. al., Physical and Morphological Studies of Size-
Classified Coal Fly Ash. Environmental Science and Technology. 12(4):
447-451, 1978.
7. Morie, G.P., and M.S. Baggett. Observations on the Distribution of
Certain Tobacco Smoke Components with Respect to Particle Size.
Beitrage Zur Tabakforschung. 9(2): 72-78, 1977.
8. Moran, M.J., Simulated Stationary Source Facility Users Handbook.
TN-262-1535, Northrop Services, Inc., Huntsville, AL, October 1975.
66
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
REPORT NO.
EPA-600/2-79-032
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
OPTICAL INSTRUMENT FOR IN-STACK MONITORING OF PARTICLE
SIZE
5. REPORT DATE
February 1979
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
A.L. Wertheimer and M.N. Trainer
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Leeds & Northrup Company
Dickerson Road
North Wales, Pennsylvania 19454
1O. PROGRAM ELEMENT NO.
1AD712 BC-11 (FY-78)
11. CONTRACT/GRANT NO.
68-02-2447
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory-RTF,NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final 10/76-10/78
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
16ABSTRACT
A new light scattering instrument for in-situ measurements of particulates in
the 0.2 to 10.0 micron diameter size range is described. Two modes of scattering are
used, each with two wavelengths of light, to generate five size fractions by volume
from a distribution of particulates. One mode measures polarized light scattered in
two orthogonal orientations at an angle of 90° to the optical probe beam. The second
mode measures light scattered in near forward angles (4 to 11°). Both modes allow the
extraction of size data when particles of different sizes are present simultaneously ir
the sensing region. These principles have been incorporated into a prototype portable
stack monitor, consisting of a 1.5 meter long, 9 cm diameter insertable probe capable
of withstanding temperatures up to 260°C. The optical signals are carried through fibei
optic cables contained in the probe. An arc source and silicon photodetectors are
outside the stack at the end of the probe, while a digital microprocessor analyzes the
set of measurements and calculates the size fractions. The microprocessor, an air
purge system, the lamp power supply, and a digital printer are housed separately from
the probe for ease of installation and service.
The completed instrument was tested.on coal-fired fly ash particulates in the
laboratory at a stationary source simulator facility and in the field at a coal-fired
power plant. Results and comparisons with other sizing techniques are presented.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
*Air pollution
*Particles
*Particle size distribution
*Monitors
*0ptical equipment
Flue gases
Evaluation
13B
20F
21B
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport)
UNCLASSTFTET1
21. NO. OF PAGES
' 77
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
67
------- |