Unteo ^i.u^s Environmental Sciences Research > PA G(H> 2 79 ()??
?Mvirorwe'H-i Pro'ertior' Laboratory f-hTii,i:\ l^/'l
Agency Research Triangle Park NC 27711
Reseaict*
M u Iti wavelength
Transmissometer for
Measuring Mass
Concentration of
Particulate
Emissions
^ •*
<•!? 600/2
-------
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-600/2-79-022
February 1979
MULTIWAVELENGTH TRANSMISSOMETER FOR MEASURING
MASS CONCENTRATION OF PARTICIPATE EMISSIONS
by
E. Reisman
Ford Aerospace & Communications Corporation
Aeronutronic Division
Newport Beach, California 92663
Contract No. 68-02-2209
Project Officer
W. D. C onner
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.
ii
-------
ABSTRACT
A field-worthy multiwavelength transmissometer potentially capable of
making near-real-time measurements of particulate mass concentration in
industrial stacks was fabricated and delivered. A computer program is
employed to interpret the transmissometer data and translate the results
into mass concentration. The transmissometer utilizes four different wave-
lengths and records the opacity of the particulate emissions at each wave-
length. Since the response at each wavelength depends on the size of the
particles, the relative values of the opacity provide the computer with
information on particle sizes. If the computer is also given the wave-
length dependence of the optical indices of refraction and guidelines as
to the most probable distribution forms, the computer can adjust the mean
and spread of the distribution to find a best fit to the experimental
data. It then uses this information to compute the mass concentration.
The theory behind the measurement technique, and a laboratory demonstra-
tion of the technique are discussed. Also discussed are the optical and
electrical design of the instrument.
iii
-------
CONTENTS
Abstract ..... iii
Figures vi
Tables vii
1. Introduction 1
2. Conclusions ..... 2
3. Recommendations 4
4. Fundamental Considerations and General Background .... 5
Scattering Theory 5
Related Problems 7
5. The Transmissometer. 15
General Description 15
Electrical System 27
6. Laboratory Studies 41
Dust Chamber 41
Particulates ............ 43
Analysis. ... ......... 44
Test Results 46
References 54
-------
FIGURES
Number
1 Size distribution of cement as determined by Cascade
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Experimental values of extinction coefficient, K, for 1 and
2 micron water droplets
Dependence of scattering efficiency on size parameter for
water droplets
Experimental values of K versus 1/X fitted to Q curve . . .
Transmitter /receiver unit
Transmitter /receiver unit (cover removed)
Retro-unit
Retro-unit (cover removed) .,
Schematic illustration of two spectral channels of the
Multiwave length Transmissometer
Split-cats-eye retroref lector unit
Schematic diagram of the transmitter /receiver unit
Electrical interconnections
Sync pick-up schematic
Preamplifier schematics
Electrical units
Interior of Nernst power supply unit
Nernst power supply schematic
Reference voltage schematic .
b
8
9
10
17
18
19
20
22
23
25
28
29
30
31
32
34
35
36
VI
-------
FIGURES (Continued)
Number Page
20 Sync amplifier schematic 37
21 Sync timing sequence 38
22 AGC amplifier schematic 39
23 Smoke stack simulator for laborator tests 42
24 Mass concentration correlation between optical measurement
and sampling measurement 47
25 Particle size distribution function 48
26 Particle size distribution function 49
27 Particle size distribution function 50
28 Mass concentration correlation between optical measurement
and sampling measurement 51
TABLES
1 Calculated absorptance by gases in the optical windows. ... 13
2 Conjugate points for the multiwavelength transmissometer
optics 24
3 Summary of test results 52
vii
-------
SECTION 1
INTRODUCTION
As techniques to monitor and control pollution evolve with time, new
requirements and specifications were generated that gave rise to new and im-
proved specialized types of instrumentation. The earliest forms of regula-
tion concerned plume opacity, since it was the most obvious indication of par-
ticulate pollution. Because opacity is by its very nature an optical phen-
omena, there are now a number of optical systems both manual and automatic
that do an excellent monitoring job. The more physical properties of parti-
culate emissions, such as mass concentration and size, are still primarily
measured by sampling techniques. In the case of mass concentration, gravi-
metric analysis of the samples is generally used. In the latter case, some
form of aerodynamic sizing or microscopic examination is used.
There are several optical approaches currently under development that
will yield a measurement of mass concentration and size in real time (or near
real time). These approaches generally rely on a prior knowledge of the op-
tical properties of the particulates so that the observed interaction can be
both measured and determined from scattering theory. Since scattering is a
size dependent phenomenon, a suitable number of observations can be reduced
to yield information about the size and, hence, about the mass concentration
of the particles. The section that follows covers the features of scatter-
ing theory that constitute the underlying principles upon which such systems
are based. One system in particular, a multiwavelength transmissometer, is
described in detail in Section 3. Test performance data is presented in
Section 4.
-------
SECTION 2
CONCLUSIONS
Since the task of the program was to build and deliver a field-worthy
multiwavelength transmissometer, it is difficult to reach any conclusions
within the framework of the program beyond the obvious one; that the task
was accomplished. If the discussion is allowed to encompass the associated
independent research that has been done over a number of years, it: is then
possible to draw a variety of conclusions. These conclusions will be made
without detailing the programs since the relevance of the conclusion to the
subject is readily apparent. They are as follows:
• Each different type of industrial stack effluent studied appears
to have an optical signature commensurate with aerodynamic
size measurement.
• Proper interpretation of signatures requires information on
the wavelength dependence of the refractive indices of the
particulates. This can be obtained by techniques described in
Section 4 or by obtaining presized monodispersed samples and
fitting to scattering theory.
• Independent determination of the distributional form should be
obtained by sampling the type of stack to be optically moni-
tored.
• Observational wavelengths should be selected at windows in
the absorption spectra of the stack gases and should encompass
the size distribution. It is best not to work at wavelengths
where the index is varying rapidly.
• Since an observation at a given wavelength is not completely
independent of a second observation at a second wavelength,
it is better to over determine and use a "least squares" fit
rather than seek a precise solution. This can be done by
hand through use of nomographs, or automatically with a data
processor.
-------
A system equipped with automatic data processing can provide
an accurate, continuous readout of particle size and mass
concentration.
A system built for one type of effluent can be used for
another by making minor software modifications.
-------
SECTION 3
RECOMMENDATIONS
It is recommended that the Environmental Protection Agency:
• Support a program to measure optical indices of common stack
effluents. Data for a given type of effluent should be accu-
mulated by a number of observers working at sites in differ-
ent geographical locations to test the repeatability and uni-
versality of the measured indices. The best approach to in-
dex measurement may be to work with a classifier whose sized
output is fed into an optical tester and monitored by a
multiwavelength transmissometer. With the size known, the
indices can be adjusted to produce the observed size de-
pendent results,
• Develop a compact, "on-board" minicomputer to do real time data
processing. The system can be functionally quite simple but
should be capable of storing and rapidly addressing tables of
data. The readout rate should be about I/sec.
-------
SECTION 4
FUNDAMENTAL CONSIDERATIONS AND GENERAL BACKGROUND
SCATTERING THEORY
Since the advent of the Mie scattering theory, it has been known that
the scattering properties of a uniform spherical particle could be completely
determined by specifying the complex optical index of the particle, the in-
dex of the surrounding medium, the radius of the particle, and the wavelength
of the incident radiation. The extension to a size distribution of such par-
ticles is immediately obvious if each particle is treated as an independent
scatterer. Recent experimental work (1,2) has shown that the volume (or
mass) scattering coefficient for a polydisperse system of irregular but ran-
domly oriented particles shows a remarkable similarity to the corresponding
scattering coefficient for spherical particles. These results strongly sug-
gest that even the irregularly shaped somewhat indescribable conglomerations
normally found in stacks can be treated as if they were spheres because of
the simplifying effects brought about by the statistical nature of the prob-
lem. Thus, if an index of refraction can be determined for the system and
the particle size can be expressed by a reasonable distributional form, the
optical effects caused by the system can be calculated.
The problem that is encountered in optical diagnostics is the inverse of
the problem solved by Mie in that one or more characteristics of the scatter-
ing function are measured and the input parameters leading to them must be
deduced. The question therefore is, that given the index of refraction, can
the particle size be determined from measurements of such optical effects as
the wavelength dependence of opacity. It is relatively easy to see that this
is true for a single monodisperse distribution as a subsequent example will
demonstrate. It is equally true, but not as obvious, for a size distribution
that can be completely specified with a few parameters such as the mean and
the standard deviation. It is probably untrue if the distribution is complete-
ly unspecified since there will be an infinite number of ways that a finite
number of observed results can be created. It is most fortunate that natural
processes tend to give rise to simple results. Many of the particulate sys-
tems can be described by the exceedingly simple log normal distribution while
others may have a simple bimodal form. We obtained the data shown in Figure 1
by sampling the stack effluents of a cement plant with a cascade impacter.
As plotted, a log normal distribution produces a straight line. More complex
distributions would have to be approximated by a simplifying mathematical form.
-------
.
[
1
* •<
c
M
1
1
s
0
•
•
*
(swas-pn
)
J9
}Wa
wia
O
o
M
Oi
g:
S w
* 01
4-1
O
1A CO
a! CU
» 6
•H
e.
5 9
cascade
e -g
* c
•H
0 01
» 2
0)
0 ^
g to
0 «
« u
O m
«« r^
a,
o
^ 4J
C
O (U
0 "
•^ III
o
c
e o
*™ •!-(
U
D
u» ^
to
•H
0)
•H
in OT
i— <
0)
3
>rt 00
9 -H
-------
The principle on which an optical 6ydten might operate can be illustrated
through a very simple example using a monodisperse distribution of water drop-
lets and making the further simplification that the refractive index does not
vary with wavelength. The relations of interest are as follows:
K(X) - NQ Q(r, X)TT r2 (1)
M - N P 4/3 TT r3 (2)
where k(X) is the wavelength dependence of the extinction, No is the number
density of the particles, rrr^ is the geometric area of particles of radius r,
and Q (r, X) is the extinction efficiency which can be determined from Mie
theory when the complex index of refraction is known. The second equation re-
lates MC, the mass concentration, to the density of the particulate material,
(p), and the volume of a particle (4/3 rrr^). if, for example, the particles
were 2 microns in radius and No » 10*Vm, the resulting values of K(X) which
would be experimentally measured are as shown as open circles in Figure 2
where the K's have been plotted as functions of 1/X. The blackened circles
show the different signature generated by 1 micron particles (where N0 has
been set at 4 x 10**/m to keep the total geometric area the same. Figure 3
shows the Q function plotted as a function of r/X. Figure 4 shows the data
sets from Figure 2 fitted to the curve of Figure 3. From the amount of
translation that is required to achieve each fit, r and NQ can be determined,
the abscissa displacement is r, the ordinate displacement is NQnr .
Normally the particles are not all the same size but have to be des-
cribed by a distribution. In that case, NQ is replaced by n(r)dr and the
right-hand side of Equations (1) and (2) have to be integrated over the dis-
tribution. The best fit is then determined through use of a computer and an
algorithm by adjusting the key parameters of the distribution (generally the
mean and the spread). Since the index of refraction varies with X, each K(X)
must be fitted to its appropriate Q function instead of using a single func-
tion as above.
RELATED PROBLEMS
This general concept has been studied for a number of years at Aeronu-
tronic. Specific problem areas covering the nature of the particle size dis-
tribution found at typical industrial sites, the optical indices of the mat-
erial, problems with gaseous components, and data processing have been in-
vestigated in great detail along with the engineering problems that had to be
met in order to build a practical optical instrument. Because the nature of
the problem does not lend itself to simple analysis, most of the answers have
to be determined experimentally.
Field work has been done at more than eight different industrial sites
where the particle size distributions were monitored by cascade impactor
studies and samples extracted for further laboratory analysis. The most com-
mon distributional form encountered was the log normal distribution, although
it sometimes showed signs of truncation due to the pollution control systems
that were employed. When the effluent contains two different species, such
-------
10.0
1.0
0.1
0.01
.11111 1 1
-
I i i 1 l 1 [ I l l 1 I i L
43
: /X?V~V * :
c/
/
x~» '
1-1 ,
r- S / /
•u II
: !• / '4'
u »
- S / /
~* o ;
- s r
1
- e '
0 /
TJ /
o /
•H /
4J .
I" ta /
/
/
/
/
/
/
/
/
/
J
/
1 1 1 1 1 1 1
/? \ f*\ /r(*\ f\j
/*
—
2 \ - .65^
^ \= 1.25U
4 X - 1.75H
5 X » 2.20u
6 X - 3 . 80a
7 \ ~ 9.10^
—
• r * IH. N0 » 4x10 /m
O r . 2U N = lO11^3 ~I
° —
_
_
—
•—
-
1/X (microns )
1 L 1 1 ! 1 1 1 1 l 1 1 1 1
0.01 0.1 10.0
8-7-2-2
Figure 2. Experimental values of extinction coefficient, K, for 1 and
2 micron water droplets.
-------
10.0
1.0
0.1
0.01
I I
1 I I
I [III
Cf
u
c
-------
10. Oi
1.0
0.1
0.01
I I IIT
I IT
5s
u
c
-------
as soot and fly ash, there is generally a biraodal distribution. Whatever the
form actually is, an approximation must be made that can be expressed mathe-
matically so that it can be put into the algorithm. The computer program
must then be able to fit the experimental data by adjusting size and shape
parameters, defining the mathematical form of the distribution. Calculations
of this type have been made with some of the data taken at the different sites.
A breadboard version of a four wavelength single pass transmissometer was
used to acquire this data.
It is unfortunate, but there is very little data available on optical
indices. This appears to be true not only of the exotic stack particulates
but also of the simplest materials we try to use as calibration particles.
While the real part of the index may be known for many materials in the visible
and the imaginary portions may be known for many materials in the infrared
(IR), only a few materials have been completely studied. There are excellent
data on water (3,4), ice (4), and quartz (5), and, more recently, data have
been published on carbon (6)(results depend heavily on the type of source).
Volz (7,8) has published data on atmospheric debris and esoterics such as
ammonium sulfate, Sahara dust, volcanic dust, and debris from a coal-burning
power plant. In prior study programs, if we categorized our sample as high
in Si02 and resembling sand, we would use the index of quartz. If it was car-
bonaceous in form, one of the carbon curves would be used. Better data is
needed to interpret results of the more sophisticated instruments that we are
now constructing. Our plans for all current and future work include taking
bulk samples of the particulates in question and making direct measurements
of the indices of refraction using the technique presently employed by Volz.
The material being investigated is pressed into KBr wafers and the
opacity of a known concentration is measured as a function of wavelength.
This yields the imaginary part of the refractive index, n1. The material is
also pressed into pellets which are polished for use in spectral reflection
measurements. The real part of the refractive index, n, is calculated from
the reflectivity equation.
o ?
r - fa-*) +n''
" 9
+ n1
The technique appears to be limited by the quality of the surface ob-
tainable in the pellets, particularly for the shorter wavelengths. An alter-
nate approach in the visible and near IR is to directly index match the par-
ticles with a working fluid. The fluid should be a mixture of two miscibie
liquids, one whose indices are lower than the particulates and one whose in-
dices are higher. The indices of the liquids are determined by measuring the
refraction at a given wavelength, using a thin liquid prism and Snell's law.
The index of the particulate is determined by finding the liquid ratio
that minimizes the opacity of a liquid cell in which a specific concentration
of the particulate has been suspended. The index matching must be done at all
wavelengths of interest. Obviously, the particles must not react either
physically or chemically with the immersion fluid.
11
-------
The gaseous constituents in the stacks can also contribute to the mea-
sured opacity. If each operational wavelength is not selected with extreme
care, a gas species might provide a major share of the opacity. It is not
enough to know where the major windows are in the spectral absorption bands
of the dominant gas species. In each window there are residual absorptions
that arise from the tails of the major bands. Their contribution is a func-
tion of the partial pressure of the gas, the total pressure, and the tempera-
ture. An analysis of the residual absorption of CC>2, S02, and l^O in the
spectral window regions has been completed for two sites of interest. The
work utilized an Aeronutronic computer program that, for a given wavelength,
sums the absorption in the tails of all known absorption lines. The program
then integrates these results for the filter bandwidth under consideration.
The calculations were done using the gas parameters determined at two sites.
The results are summarized in Table 1.
The residual absorption must be treated as corrections to the total mea-
sured opacities and the fluctuation in the residuals accepted as errors un-
less an independent effort is made to monitor the gas characteristics.
The problem of data processing is far more complicated than the sample
curve fitting example given earlier. As previously indicated, there is gen-
erally a separate Q-curve for each wavelength and, therefore, completely
separate integrals for each wavelength. The integrals are compared to the
extinction coefficients being monitored to obtain estimators of N0 (one at
each wavelength). If everything worked perfectly, the proper distribution
function would result in all values of N0 being the same. In practice, a
least square fit is done on N and the search is for the distributional para-
meters that minimize the residuals of the least square's fit.
The program can be made to minimize the influence of residual opacity due
to gaseous absorption. This is done by subtracting out the average gaseous
opacity from that measured. The effect of the variations in gaseous absorp-
tion are not important if the particulate concentration is high but can pro-
duce errors at very low levels of particulate concentration.
The earliest versions of a multiwavelength transmissometer employed
single-pass optics with the detectors on the opposite side of the stack from
the light source. While a system of this type can monitor either particle
size or mass concentration, it is subject to a number of operational compli-
cations in that it cannot handle gain changes or zero drifts in the electron-
ics and has no way of sensing changes in source brightness or the increased
opacity caused by degradation of the optical components. The simplest way to
overcome these shortcomings is to utilize a retroreflector-system on the far
side of the stack and keep all other system components on one side. It is
then simple to create a reference beam that can be distinguished from the pri-
mary beam in a number of ways (i.e., chopping frequency, time of occurrence,
etc.). If this reference beam utilizes the same detectors and amplifiers that
the primary beam does and most of the optics are common, then the system can
be made to stabilize the output of the reference channel, thereby ensuring
stability to the signal channel.
12
-------
tn
£2
o
p
M
hJ
0
M
H
(In
O
W
H
I i
C/3
^
p
H
3
U
)— J
<^
U
r-l
w
I-I
m
H
4J
^ i *g£ Q
^ t o
w 00 O
ti m r*
a * *
a a H
5 H .J
g~* 0
^ +1
*en3
M CM >n
O II 1
a.
H -J
B
o.
a- r- CM -4 -^
t c \/ V V
m CM v* **
• • • •
p o o o o o o
7 V V V V
ON CM O O O O O
O CM e>l ^ in O l»«
1-4
in
vo oo in o P» ^ en
oo oo CM r» CM vo -a-
• ••••••
oo en CM "* r* O O
13
-------
The last problem to be considered is the mechanism by which the data is
processed,, The simplest system is to compare the incoming signal with a pre-
pared nomograph that yields the size information. This technique was used
during the early development of the system. With a little experience, it is
possible to make useful deductions as to particle size by merely watching the
output of the four channels. Our most recent system forms the data into a
pulse code modulation for storage on audio tape. The tape is played into the
main computer via a ground station normally used to introduce telemetry data
to the computer., This was done because of the difficulty in introducing
data into a major computer except via punched cards, etc. The problem is
one of incompatible formats. With this arrangement, we were able to record
10 data sets per second which is far more than would normally be needed.
Recently, there have appeared on the market a number of data converters
reported to have the ability of linking laboratory generated analog data
directly to a major computer. The simplest solution to data processing is
to build a microprocessor directly into the system. The output could be ob-
tained in near real time in the form of a display or a printed record.
14
-------
SECTION 5
THE TRANSMISSOMETER
GENERAL DESCRIPTION
The multiwavelength transmissometer is in reality four separate trans-
missometers, each operating at its own wavelength. The four channels use
common mirror optics over most of the system and share a common path through
the stack. Once the beam has returned to the receiver, it is broken down in-
to four wavelength domains by a series of three beamsplitters which define
four channels: the visible channel X.., the near IR, X~, the mid IR, X 3 and
the far IR, X/. The main splitter transmits in excess of 70 percent of the
incident radiation throughout most of the visible and near IR. It reflects
better than 80 percent of the incident radiation in the 4 to 13 micrometer
range. In between, the splitting properties are not as good but, since the
output of the source peaks in this range, there is more than enough energy.
The transmitted light is further wavelength divided into visible and near IR
by a splitter that reflects 95 percent of the incident radiation below 0.65 M-
and transmits 90 percent above 0.75 P1 while the reflected light is divided
into mid-and far-IR by a splitter that reflects about 90 percent of the ra-
diation below 4 \i and transmits better than 90 percent between 8 and 13 n .
The operational wavelength of each channel is precisely selected by means of
a narrow band filter located within each of the four chromatic ranges. Each
filter has been carefully selected to operate at a window of maximum trans-
mission through the gas species normally found to stacks (i.e. FLO, CO, C0?,
S0_, etc.). Unfortunately, even in a window there is considerable residual
absorption at elevated stack temperatures (i.e., a few percent). These re-
siduals should be considered when interpreting transmissometer data, parti-
cularly when a particle concentration is low. Channel 1 operates at about
0.46 M- , Channel 2 at about 1.26 M-, Channel 3 at about 3.9 M" and Channel 4,
a composite filter, at about 11.3 M-.
At the heart of the system are four detectors selected for their per-
formance at the four wavelength regions: Channel 1 uses a silicon photodiode,
Channel 2 uses a lead sulfide cell, Channel 3 is a lead selenide cell, while
Channel 4 uses a broadband pyroelectric detector. The light that is measured
by each detector is chopped at the source at 90 Hertz. A sync pickup is pro-
vided at the 90 Hz chopper in case it is desirable to monitor the signal us-
15
-------
ing a phase locked amplifier. This sync signal is not currently used in
data processing.
For about 90 seconds out of each 100 seconds, the beam passing through
the stack returns to the detectors. For the remaining 10 seconds, a refer-
ence beam that never leaves the main unit impinges on the detectors. These
two events are controlled by a chopper making one revolution every 100 seconds.
The position of the chopper is monitored by two sync units, one which puts
out a side pulse beginning before the beams switch from signal to reference
and ending after they have switched back and the signal beam is restored.
Thus no valid data is to be taken when the wide pulse is on. During the middle
of the wide pulse, a narrow pulse is generated. It begins after the refer-
ence beam is fully on and ends before it begins to fade. Calibration is in
effect only when the narrow pulse is on. The system is housed in four units.
The largest unit is the transmitter/receiver package which contains most of
the optics, the light source, the aforementioned beamsplitter and filters,
the detectors, and their preamplifiers. The unit is 67.3 cm high, 52.1 cm
wide and 14.6 cm deep (26k in. x 20% in. 5 3/4 in.) and weighs about. 30.9 kg
(67.9 Ibs). It is designed for stack mounting and has been weatherproofed and
ruggedized. The second unit, a retrosystem, is much smaller but similarly
packaged. It is 49.5 cm high 26 cm wide and 13.3 cm deep (19% in, x 10% in.
x 5 % in.) and weighs about 9.1 kg (20.1 Ibs).
Figure 5 is a photograph showing main unit completely assembled while
Figure 6 is a photograph of a unit with the cover removed showing the in-
terior. Similar views of the retro-unit are shown in Figure 7 and Figure 8.
The optical system for the multiwavelength transmissometer was designed to
maximize system efficiency and stability while minimizing system complexity
and cost. Common optics were used wherever possible for both the signal and
and reference beams. Since photodetectors, narrow bandpass filters, and
dichroic beamsplitters are especially prone to variations in performance as
a function of temperature, careful attention was given to making these com-
ponents common to both beams to minimize thermally induced drift. The optical
throughput of a beam is limited by two apertures. All other apertures are
sufficiently overfilled or underfilled so that modulation will not occur due to
a combination of edge effects and vibration or relative motion between con-
jugate or adjacent apertures. Mirror optics were employed throughout to elim-
inate variations between channels due to chromatic effects and to simplify
the task of optical alignment. Wherever possible, stock components were
utilized.
The final design was selected from a number of candidate systems. The
transmitter/receiver units considered included using beamsplitters to combine
and separate transmitter and receiver beams that share a common aperture and
those that use essentially different portions of an aperture for the two
functions. Retrosystems considered included a single element corner cube, a
corner cube array, on and off axis cats-eye retroreflectors and on and off
axis split-cats-eye retroreflectors. Since, under certain applications, the
system might be energy limited, the most efficient design was selected, that
using separate apertures and a split cats-eye retroreflector. The functional
16
-------
O
(U
CO
n)
H
«
in
0)
17
-------
T)
0)
I
o
o
-r-l
fi
3
•r-t
0)
o
0)
S-i
0)
in
c
cfl
CD
S-i
3
18
-------
Figure 7. Retro-unit,
19
-------
Figure 8. Retro-unit (cover removed)
20
-------
details of the optical system are best demonstrated with the aid of an
abridged schematic (Figure 9), illustrating only two channels. The trans-
mitter and receiver beam have been deliberately crossed to minimize the ef-
fects of turbulence within the stack.
A Nernst glower source is imaged on apertures Al and A8 by mirrors Tl
and Rl, respectively. Al and A8 are similar in size and are overfilled by
the images for stability. For the transmitted beam, Al is imaged on Cl at the
retro by mirrors T3 in combination with T4. For stability, Cl is smaller than
the size of the image of Al. Modulation of the signal as a function of any
angular motion or vibration of the transmitter is minimal and is limited to
that and resulted from nonuniformities in the image that are caused by non-
uniformities in the source radiant emittance and/or aberrations within the
transmitter optics. Aperture A2 within the transmitter is imaged onto C2 with-
in the retro by Cl, and Cl is imaged onto C3 by C2. The image of A2 at C2 is
imaged onto apertures A4 and A5 within the receiver by C3. Aperture A3 within
the retro is almost conjugate to Cl arid serves as a limit to the portion of Cl
that is used. A4 is imaged onto photodetector Dl by Si. Similarly, A5 is
imaged onto photodetector D2 by S4. The photodetectors are larger than the
images. Apertures A4 and A5 are of equal size and are equidistant from dichro-
ic beamsplitter F2. Narrow bandpass filters F4 and F5 are larger than A4 and
A5 and are located near A4 and A5. Since A2 is conjugate to A4 and A5, and
A2, A4, and A5 are all within the transmitter/receiver unit, there is no motion
of the beam on A4 or A5 due to angular vibration of either the transmitter/
receiver unit or of the retro unit. Placing the bandpass filters at these lo-
cations and making the photodetectors conjugate to them minimizes modulation
of the signal that could occur due to vibrating these typically nonuniform com-
ponents about within a probable nonuniform beam. Apertures A4 and A5 are
slightly overfilled by the image of A2. The limiting apertures that define the
throughput of the signal beam are A4 (A5) and A3. Within limits, angular mo-
tion or vibration of the retro unit will not modulate the beam as long as the
image of A2 remains on C2 and as long as C2 is kept free of dirt or scratches.
The angular motion of the retro must be restricted such that vignetting does
not occur from the retro window W2 or from the sides of the path in the stack
at either A3 or at the image of A3 or Cl. A more detailed schematic of the
retroreflector unit is presented in Figure 10.
For the reference beam, the source is imaged on and overfills aperture
A8. Spherical mirrors R3 in combination with R4 form virtual image of A8
approximately the same distance from R4 as the retro is from T3. Radiant
energy from R4 impinges on A4 and A5 with the same divergence as radiant
energy from the retro unit. The beam is almost collimated. Aperture A9 is
smaller than apertures A4 and A5 and the reference beam slightly underfills
these apertures. Apertures A8 and A9 define the throughput of the reference
beam. The diameter of aperture A9 was selected to make the outputs of the
photodetectors about equal from the signal and the reference beams when the
system was used on a 6 to 10 meter diameter stack. The reference beam is
made coaxial to and coincident with the signal beam by beamsplitter R5. This
beamsplitter is an uncoated BaF2 window that results in little attenuation of
the signal beam. The calibration chopper alternately transmits the signal
and the reference beams to R5. Common areas of common optics are used for
21
-------
OS
d
>
CO
c
CO
CO
3 co
rH IS
^-1 •!-(
•H 4J
t-l
O 3
CO
-------
BOX
w.
BOX
-BaF2 WINDOW
SIDE VIEW
ri Hi
BaF2 WINDOW
45 MIRROR
8-7-2-6
Figure 10. Split-cats-eye retroreflector unit.
23
-------
both the reference and the signal beams after the two beams are combined at
R5. This minimizes system errors that could arise from occurrences such as
variations in the spectral transmittances of the filters or variations in
photodetector outputs as the system temperature changes.
A detailed schematic diagram of the optics of the transmitter/receiver
unit is presented in Figure 11. The nomenclature is the same as was used in
Figure 9. Table 2 presents a list of conjugate points for the system. The
general design and alignment philosophy was covered in the above discussion
of Figures 9 and 10. Only deviations or additions are covered here. Photo-
detectors Dl and D2 have actually been made conjugate to a completely arbi-
trary point, i, set halfway between R5 and Fl rather than their respective
apertures A4 and A5. This achieves the proper spot size at the detector
while using off-the-shelf optical components and makes no significant differ-
ence in performance. Point i is essentially conjugate to A2 as are A4 and
A5 because A4, A5 and point i are all almost equidistant from C3. F5 was
placed at D2 rather than at A5 to help prevent stray light from saturating
the detector. There is no significant image motion on F5 because it is
essentially conjugate to point i. The spectral shift and bandpass broadening
that results from placing F5 in convergent light does not adversely affect
this channel.
TABLE 2. CONJUGATE POINTS FOR THE MULTIWAVELENGTH
TRANSMISSOMETER OPTICS
Signal Beam Reference Beam
Source Source
Al A8
Cl Virtual Image at Selected
Distance
C3
S8, S12, Crosshairs S8, S12, Crosshairs
Signal Beam
A2
C2
A4, A5, A6, A7
D3, D4
Point i (Almost Conjugate to C2)
Dl, D2
24
-------
G
S-i
0)
•H
0)
a
0)
S-i
•H
6
en
c
M-i
O
cd
S-i
60
O
•H
4-J
n)
OJ
o
60
•H
fn
25
-------
Because the 11 >m channel was severely detector-noise limited, its aper-
ture A7 was first made oversize then eventually eliminated and detector D4
serves as one of the two limiting apertures that defines the throughput for
this channel. D4 is conjugate to the plane of A7. Filter F7 was moved rela-
tively close to S12 to help block stray radiation to D4. These changes sig-
nificantly improved the signal-to-noise ratio and resulted in no measurable
degradation in the stability of this channel.
The built-in alignment telescope is comprised of lenses Ll, L2, and a
set of crosshairs. Either the signal or the reference beam can be viewed
through the telescope while the system is in operation. Figure 11 includes
a sketch of the reference beam image and the signal beam image as seen through
the eyepiece when the system is properly aligned. Mirror T7 is adjusted to
locate correctly the signal beam image.
It is necessary to adjust T6 and C3 when the instrument is shifted to
stacks of different diameters. Mirror T6 is adjusted to center the trans-
mitted beam on Cl and then C3 is adjusted to center the retroreflected beam
on R5. These adjustments should be made before mounting the instrument on
the stack.
While a complete alignment of all components is very time consuming, it
need be done only once during the original assembly. Once the components in
the receiver channel are located and aligned along with the telescope, they
are permanently locked in place.
The following description of the alignment is an ideal case, perhaps
best done only once prior to installation permanently on one stack.
. T3 and T4 focus Al on Cl.
. R3 and R4 create virtual image of A8
back the distance of Cl.
• A2 is imaged on C2 by Cl.
. Image of A2 at C2 is re-imaged by C3 to A4, etc. Some
longitudinal motion of C2 will be required.
• Reference beam axis is reestablished by adjustment of T5 to
center beam on telescope objective and crosshairs.
• Transmitter beam axes are reestablished by adjustments of T6 and
C3 (perhaps Cl and C2) to restore all indicated ray paths and
center beam on telescope objective (or A4) and crosshairs.
• T7 is used to re-aim output beam to accommodate flange inaccuracies
on day-to-day deformations of stack and duct mounting configuration.
If an application calls for installing the system at many different
locations, a series of compromises can be effected. This may cause minor
reductions in signal but will not otherwise limit the performance of the in-
strument. Since typical stack diameters run from 10 to 30 feet, some median
distance can be selected (i.e., 17 feet), thus,
26
-------
• The reference channel can be set at -17 feet. R3, R4, and T5R
can be permanently locked.
• The signal channel can be set at +17 feet. T3, T4, and T5 can be
permanently locked.
• If the turbulence is not a limiting factor (it frequently is not),
the system can be operated without a crossover, the outgoing beam
striking C3 and then relayed back by Cl. This enables T6, Cl, C2,
and C3 to be locked. The only adjustment will then be made by T7.
Under the above three simplifications, multistack operation is very
straightforward. The complete adjustments will only need to be done when
a stack is encountered with a diameter much smaller than the typical range
(e.g., less than 5 feet).
ELECTRICAL SYSTEM
Of the four units comprising the transmissometer, one, the retroreflec-
tor, is completely passive in that it contains nothing electrical. The re-
maining three units are shown in Figure 12. The largest of these, the trans-
missometer itself, is internally wired in a manner illustrated in Figure 13.
The 110 volt lines and similar power line are fed in from the Nernst Power
Supply Unit by means of S04. A regulated -15 volts and +15 volts are fed in
from the signal processing unit via S03. A relay actuated battery pack pro-
vides 135 volts dc for the ZnS and ZnSe detectors. The condition of the
battery can be checked by pressing switch Si and reading the voltage on
meter, Ml.
The outputs of the three sync pickups go to the sync pickup card mounted
within the transmissometer box. The card circuitry is shown in Figure 14.
The signals are clipped and then amplified. The wide sync signal is inverted
by means of a second stage. The sync outputs terminate at BNC connectors
on the connector panel of the transmitter box.
The outputs of the four detectors go to four preamplifier cards that
are essentially the same for all channels. The differences that occur are
in the input circuitry and the gain settings. The schematics of the preamp
cards is shown in Figure 15; Channels 2 and 3 use battery operated photo-
conductive detectors that create a driving signal across Rll. Channels 1
and 4 have their own preamplifiers with outputs directly across terminals 9
and 10 (Rll, R12, and C7 are not needed). The potentiometer, R5, is used to
balance the DC offset of the preamp. The gain of each channel is adjusted
by selecting the value of R3 to produce a reference signal output of about
1.75 volts peak-to-peak. The chopped signals appear as square waves. The
output of each channel terminates on the connector panel of the transmisso-
meter box.
The remaining electrical units, the Nernst power supply box and the sig-
nal processing box, are designed for use in a control room environment and
have not been ruggedized as have the optical units. They are shown in the
form of a table mounted chassis in Figure 16 but could easily be adapted
27
-------
TO HSVAC
sol
so4
-Cfl
sea
out}
D
v_
S CABLES
TO U5VAC
T"*UI-T OUTPUTS
8-7-2-8
Figure 12. Electrical interconnections.
28
-------
P is
cfl
S-i
00
03
•H
00
C
•H
VJ
•H
15
•H
CU
O
0)
S-i
0)
•H
CO
c
cfl
M
(11
(-1
bO
•H
29
-------
-•—4-ISV
2) '".OHs"
g-7-2-10
Figure 14. Sync pick-up schematic.
30
-------
CHANNEL 1 AND 4
MO
•
J7
T
Cl (M
per IN 0 1( wv-
04
Ci
T
u
QT) +isv
I—(ji) COM
™
Z2 >a » gi) OUTPUT
Mj
*
C£,
RS
V
-H5v
CHANNEL 2 AND 3
MO
Cl P.I
DET IN ©
09)—I
V
wa
i
G4
C5
T
&AT IN
u
-vw
| © COM
V
P.5
OUTPUT
^
C«o ~
^ C7
T;
V
(13) -I5V
8-7-2-11
Figure 15. Preamplifier schematics.
31
-------
TO
4-1
•H
C
O
•H
M
4-1
O
0)
rH
W
o>v'*?*w
-------
for rack mounting. The internal view of the Nernst power supply box is
shown in Figure 17 while the schematic is shown in Figure 18. A wiring dia-
gram of this box is given in Figure 18. If the unit is to be inoperative for
several days, it is desirable to keep the Nernst operating at a reduced power
level; a standby mode is available. This minimizes both the thermal shock to
the unit and prolongs its operational lifetime.
The main electronic functions are conducted in the signal processing box.
It contains four AGC amplifier cards, a reference voltage card, and one + 15
volt commercial dc supply. The reference voltage card, (Figure 19) has a
+ 15 volt input, a regulated + 12 volts output, and a 6.2 volt reference out-
put.
The wide and narrow sync signals are amplified by the sync amplifier
card - the circuitry of which is shown in Figure 20. The wide sync output is
normally +15 volts swinging to -15 just prior to the switchover from signal
to reference beam and returning to +15 volts soon after the return to the
signal beam, an interval of about 10 seconds. The narrow sync signal is
normally -15 volts and swings to +15 volts during the interval that the ref-
erence beam is on and steady. Calibration takes place during the interval
when the narrow sync is positive, a period of about 5 seconds. A third volt-
age, picked off of the wide sync channel, is normally +5 volts but goes to
zero volts when the wide sync goes to -15. This voltage appears at the event
terminal and is used to tell a computer to ignore data during the 14-second
interval when the signal beam will be interrupted. The timing sequence of
the sync probes is illustrated in Figure 21.
The four AGC-amplifier cards are essentially identical. A typical cir-
cuit diagram is shown in Figure 22. The quasi-square wave signal from a
given preamp card (i.e., channel 1) is fed into AGC - amplifier card number
1. The first stage, Z3A, and its associated components constitute a bandpass
filter that convert the input square wave into a sine wave. The signal is
attenuated by a factor of approximately 50/1 at the input of ZlA. The gain
of ZlA can be varied between 50 and 300 depending on the state of Ql. When
Ql is open (in a high impedance state), the gain of ZlA is controlled by R6
and R24 and is about 50. When Ql is closed (i.e., in a low impedance state),
the gain is controlled by R6 in series with R23 and R24 in parallel. The
gain is then about 300. Between the attenuator and the variable gain, the
system produces an effective variable gain ranging between unity and six.
The output of ZlA is fed into a precision full wave averaging circuit
composed of Z4A, Z3B, and their associated electronics. This output (Z4 pin
12) is a dc voltage equal to the average of the rectified sine wave at Zl
pin 10. This dc signal is fed back to the AGC loop through R32 to ZlB pin 1
where it is compared to an AGC reference voltage level set by potentiometer
R20 and appearing on ZlB pin 2. The combination of ZlB, R32, and Cll serve
as an integrator. Q3 is a switch feeding a sample and hold circuit composed
of Z2, Q2, and Cl. Since Q2 has a very high input impedance and Z2 has a very
high output impedance, ClO is essentially isolated until Z2 is switched to a
sample state by the narrow sync pulse. The output of Q2 which is coincident
with the junction of R27 and R28 will reset itself to an updated value, hope-
33
-------
w
•H
C
o.
Cu
3
CO
O
ft
en
fl
S-i
o
34
-------
CM
n
U
•H
4J
§
(U
,C
CJ
to
ex
ex
CD
0)
§
ex
w
c
00
QJ
S-i
to
•H
35
-------
n
N
O
•t-(
4J
O
w
0)
00
to
4-1
O
c
(U
)-<
QJ
M-l
-------
o
•H
O
to
a)
•H
IH
•H
c
to
O
OJ
•H
fn
37
-------
WIDE SYNC (SSW)
NARRCW SYNC (SSN)
SIGNAL BEAM
REFERENCE BEAM
\
5 SEC.,
\
8-7-2-15
Figure 21. Sync timing sequence.
38
-------
u
•H
4J
CO
I
U
CO
S
TO
CM
CN
00
•rH
39
-------
fully close to its previous value. This signal drives Ql and sets its inter-
nal impedance somewhere between near zero to near infinite. This in turn sets
the gain of Z1A between 300 and 50 (i.e., overall stage gain of 6 to 1). Thus,
the AGC loop tries to set the gain of ZlA so that the dc signal caused by
the reference beam (i.e., ZlB - pin 1) is equal to a preset precision voltage
set by R20 at feed to ZlB - pin 2.
The signal output from Z4A - pin 12 goes to Z5A when Q4 is closed as it
normally is during the operate portion of the 100-second duty cycle. The
gain of Z5A has been adjusted with potentiometer R14 so that at 100 percent
transmission, the signal level is exactly 5 volts at the "transmission out"
terminal (i.e., "T" out). With "T" out set at 5 volts, R19 is adjusted so
that "l-T" out is 10 volts. Thus, as "T" swings from 5 volts to 0 volt, the
"1-T" output will swing from 0 volts to 10 volts.
A summary of the events that are involved calibration is as follows:
• Wide sync open Q4 ... Output "T" floats.
• Beam undergoes transition from signal to reference.
... reference signal established.
• Narrow sync on, closes Q3, allows AGC to reset.
• Narrow sync off, Q3 opens, AGC fixed.
• Reference beam off, signal beam on, signal beam stabilizes.
• Wide sync off, Q4 closes, Q5 momentarily closes for quick reset
of voltages on C7, C8, and C9, the output time constant capacitors.
• Q5 opens restoring RC time constant (i.e., R13 and C7, C8, C'9).
40
-------
SECTION 6
LABORATORY STUDIES
DUST CHAMBER
The work described in this section was done as part of an ongoing Inde-
pendent Research and Development program conducted by the Advanced Develop-
ment Operation of the Ford Aerospace & Missile Systems Division. The multi-
wavelength transmissometer was conceived and developed during earlier phases
of this technical effort. The current program involves testing the perform-
ance of the instrument in a controlled laboratory situation. The program
and the results have been included in this report for the sake of completeness.
Under the program, it was necessary to develop a dust generating system
that could be used to create particulate environments similar to those found
on industrial stacks. The system, Figure 23, incorporates a system of blowers,
sieves, and sieve shakers to disperse the powdered solid into an air stream.
The particle ladened air has to traverse a series of baffles which serves to
remove the agglomerates. The resulting aerosol is passed into a 2.7 cubic
meter holding volume or dust chamber where any remaining oversize particles
settle out during the more than 20-second residence time. The aerosol is
diluted with clean air and then passed down a 0.2 meter diameter duct which
acts as the optical path and, finally, the aerosol passes up and out through
a roof exhaust fan. The optical path length can be adjusted from 1 to 10
meters. The flows in the system are governed by a series of dampers: one at
the exit of the dust chamber which determines the amount of dust-laden air
entering the optical path, one which regulates the amount of dilutent air,
and one at the exhaust end which regulates the gas velocity in the optical
chamber.
The concentration of the dust is controlled by varying the number of
sieves, the mesh size, and the sieve area. It is also controlled by the
amount of dilutent air that is mixed with dust-ladened air. Once the parti-
culate has reached the optical chamber, it undergoes continuous optical moni-
toring and can also be sampled isokinetically as desired. One of the best
sources of particulates are the effluents arising from the industrial sites
of interest. The material is generally collected from the pollution control
systems. The indices of refraction at the wavelengths of interest are mea-
sured using the techniques described in Section 4. This information serves
both the lab experiments and possible future site experiments.
41
-------
co
0)
t-l
O
4-1
«
(-1
O
O
4-J
CO
O
CO
4-1
CO
CU
3
t>0
•H
42
-------
PARTICULATES
A raultiwavelength transmissometer was used in earlier studies performed
at a number of different industrial sites. The data obtained indicated that
the particle size of the effluents in the stack was not constant with time
but would change as plant operations changed (i.e., starting and stopping of
the plant process, changing the feed materials, etc.), or when conditions in
the pollution control system were altered (i.e., blowing or shaking bags, etc.)
When samples were taken into the laboratory and redeployed using the dust
chamber, the size distribution was generally constant and repeatable. This
is not surprising since the observed distribution should be that of the feed
sample appropriately corrected for the efficiencies and prejudices of the
extraction system.
The material that was studied the most was the effluent from the kiln of
a dry process cement plant. Initially, index measurements were made on both
baghouse samples and chimney samples. When the optical properties proved to
be the same, the more readily available baghouse samples were used. The
material was taken from the freshly shaken bags while still hot (300 C) and
stored in sealed containers. As long as the sample was not exposed to labor-
atory air for undue lengths of time (i.e., overnight), it would flow freely
through the feed sieves (67 micrometers to 125 micrometers). After exposure,
the particulate would not shake through the sieves but could be brushed
through. When sieved a second time, the results were the same. Apparently
the material is hygroscopic and the particles develop a strong affinity for
each other and would rather roll into balls than fall through the sieves.
In order to make measurements on cement dust samples of different sizes,
it was decided to use fractionated cement dust as the raw feed material. Ap-
proximately 50 kg (110 Ibs) of the virgin dust was shipped to a company that
builds instruments that classify particles by size. When the samples were
returned and did not flow through the feed sieves, it was obvious that they
had been severely moisture contaminated. The unit that had been used in per-
forming the classification suspended the particles in laboratory air which
ran from 40 to 60 percent relative humidity.
A second attempt at classification was made with the unit completely en-
closed in a polyethylene house into which flowed 3.4 x 10 liters (12,000
cubic feet) of dry nitrogen. The sieving properties of most of the samples
were improved, however, the small fractions (1 micrometer and less) would
still not pass the sieves. Either the precautions taken were not sufficient
or the ultrafine particles may have this inherent characteristic.
In order to obtain effluents with reduced mean radii, an alternative ap-
proach was adopted. The feed material used was the virgin dust but the flow
characteristics of the chambers were altered to discriminate against large
particles. The air flow in the stack simulator, normally 3 meters/second (10
feet/second), was reduced to 1.5 meters/second (5 feet/second). The feed
damper from the dust chamber was replaced with a flow limiting screen while
the fresh air damper was fully opened. This resulted in good log normal dis-
tributions with mean radii of 1.7 to 1.8 micrometers. In order to obtain a
43
-------
larger mean radii, the air flow was set at 7.5 meters/second (25 feet/second)
and the feed damper opened wide while the fresh air damper was replaced by a
flow limiting screen. The sample used came from the classifier and had some
of the fine material removed. This resulted in a slightly altered distribu-
tion with a mean radius of 3.0 to 3.1 micrometers.
ANALYSIS
The equations governing the key properties of a distribution of particles
can be written
CO
K(X) = NQ J p(r)dr nr2 Q(n,X) (3)
o
03
M = N J p(r)dr 4/3 rrr3 p (4)
o
Where K(X) is the wavelength dependence of the extinction, p(r) dr is a
dimensionless probability function (i.e., the probability that the size of
a particle will fall between r and r + dr), NQ is the number of particles
per unit volume, Q is the extinction efficiency, M the mass concentration,
and p is the density of the material.
If the mass concentration is log normally distributed:
~(ln r - In ro)2
r --5 dr
P(r) =* ==-.[ e 2CT r (5)
/2TT a o
then p(r) dr is given by:
-( In r - In rQ)2
7~2dr
p(r) dr = Ae 2cr ~ (6)
where A is a normalizing constant such that:
00
J p(r)dr = 1 (7)
o
It can be shown that:
3 e - 9/2 d2
A „ !o_J (8)
44
-------
If it is necessary to truncate the integral of Equation (3) or (4) for com-
putational purposes or for physical reasons (e.g., the particles have been
subjected to the same size-limiting processes such as a cyclone) , Equations
(3) or (4) can be restated so as to be self-normalizing:
B
N JA p(r) dr TT r2 Q(r,X)
K(X) = -a-^5 - (9)
J P(r)dr
A
T>
N f. p(r)dr 4/3 rrr3 p
-2-±| -
J p(r) dr
A
(10)
While p(r) is a function of r, it is also a function of the distributional
parameters and is more correctly
P(r, rQ, cr)
showing the dependence on the mean, r , and the spread, O" . If, for a given
system of particles, the indices are Known, then the Q functions are known.
If p(r) is known to have a characteristic distributional form, then K( X) can
be calculated once rn, CT and N are known.
Lo»
o
The value of K( X ) can also be calculated from the measured transmission
values
where L is the optical path length through the particulates and T is the
transmission at a given wavelength.
The experimental data is processed in a way analagous to the curve fit-
ting described in Section 4 (Figures 2, 3, and 4) . For a given set of distri-
butional parameters, the computer calculates the value of N needed to pro-
duce agreement between Equation (9) and Equation (11) at each wavelength. If
the distribution selected was absolutely correct and the experimental data
perfect, the calculated value of N would be independent of wavelength. Gen-
erally, this is not so and it is necessary to use a least squares procedure on
NO to evaluate the quality of the fit of Equation (9) to Equation (11) . When
the values of ro and a are found that give this best fit, they are used to
calculate M£ using Equation (10) . Since the computer time required to analyze
the data is generally shorter than the time taken to acquire the data (for
computer sampling rates of 1 set of observations per second) , it is reason-
able to predict that onboard minicomputers could be designed to produce near
real-time monitoring systems.
45
-------
TEST RESULTS
An extensive series of tests using cement particulates were undertaken
to compare the mass concentration as measured with method 5 techniques to
values calculated from the optical data. On occasion, the probe was fitted
with an impactor so that the calculated size distribution could be directly
compared with those sampled. Since the runs were typically 20 to 30 minutes
long and the data acquisition system sampled the data about 10 times a second,
approximately 15,000 sample sets were available for analysis. In order to
conserve computer costs, every 10th set or every 60th set was analyzed, pro-
ducing 250 to 1500 readouts per run. Each readout would include the best
values of ro for a given cr, the M£ associated with that r, and the residuals
of the least squares fit. The average mass concentration for the entire run
was also presented.
The value of rQ that occurred most frequently was 1.9 micrometers, with
2.0 micrometers a close second. In some runs the value stayed constantly at
1.9, in others it varied with time. Most variations occurred towards the end
of a run and even then seldom fell outside of the range 1.9 to 2.2 micrometers.
The values of o" generally ranged between 0.45 and 0.60. One run, number 21,
appeared to have different properties. In this single instance ro ranged be-
tween 2.4 and 2.7 micrometers and cr ranged between 0.85 and 0.95. Apparent-
ly the sample used for this run differed in some way from the others but still
produced acceptable data.
The excellent agreement of the mass concentration data is shown in Figure
24. A perfect fit would require data points to fall along the indicated
straight line. The data from a typical impactor run is shown in Figure 25.
The data is best fit by a value of ro of 1.9 micrometers and a a of 0.60.
An experiment of more limited scope was attempted while trying to vary
the particle size distribution in the manner described above. Because an
electrical problem made it impossible to use the automatic data processing
system, each run was limited to 10 or 15 data sets fed to the computer by
hand. The results are summarized in Table 3.
Figures 26 and 27 show the log normal plots of the impactor runs while
Figure 28 illustrates the agreement between experimentally determined and
calculated values of the mass concentration.
In run number 8, virgin baghouse dust was used and the rate of extraction
of particulate air reduced to enhance the relative abundance of small parti-
cles. The distribution still appears to be log normal over most of the range
(i.e., 2 percent to 98 percent). The experimental data yields ro = 1.7 micro-
meters and CT = 0.45, agreeing well with the optical data for runs 8 and 9.
Runs 10 and 11 were performed in a manner similar to runs 8 and 9 except that
the feed sieves were different. The optical data implies both a smaller mean
and spread.
46
-------
tfl
g
O
-r-l
4J
P.
O
c
0)
0)
c
o
•r-l
CU 4J
M C
s-i cu
o E
O QJ
d D
O CO
•c-l CO
jj
-------
'
\
1
\
r-.
O.
t-i
r-
1
\
s
H
CO
o
H
55
CJ
>
*V
\
V
2 s.
ii ii
O Q
\.
\
\
\
\
\
^
t
»
IA
*
ft* W
t-H
CO
* w
H
M
* M
z
SKi
nJ
S<
§
CO
?M
«
S^
ptj
SS
w
.
•&.
(SNOHOIK)
48
-------
y
*
\
\
\
f-
ON
00
od
\
\
H
to
ii n
„» 0
V
\
A
\
\
\
\
* ?
^ M
rj-
00
s
5
t
c
o
•H
K4-;
o
3
_ 4-1
^r TT1
•« NI .-
M C
W O
•H
O j_i
«g 3
*^ ^
O -H
S M
^ M M
•H
• 1 T3
S cu
O N
*8 »
1 ^
a^;
M H
S m
fe P-l
.i vc-
— 2
e S
• H
Eu
M
'
as
(SNOHOIH) naaawvia
49
-------
^
s
I
<1
^s
\
X
^.
N
^-s
l-~
r~
o\
,-j
x^
00
t— 1
tti
\
H
en
^ _
5
g
u
4**
0
PI
II
,,c
^1
r-J
cn
II
lj
S^
*X
r*.
•
ii
b
'ff "
o
o
f~J
II
3 b
V
V
•\
X
N
s
(
S
^
s
S
X
3
s
"S
\
a
•?
l
wt
t
»
*.
Ml
•»
a
K
•
5
«
e
r
a
M
•
*
IA
e
A
w
M
s
3
L>
3
_3
S
[d
O
02
M
fc
50
-------
_l I IIIIII
1 1 I I I I I I
CO
o
CN
§
M
H
3
W
O
§
O
CO
co
p
3
pi
£3
co
>>
iJ
J
H
H
§
c«
O
•i-l
4J
d
(U •
a) w
^ c
JJ 0)
O CO
•i-l CO
JJ OJ
0) bO
M C
i-l -H
O i-l
CJ &•
d I
O CO
CO C
}-i CO
4J
C 4J
O 0)
O SJ
CO CO
co cO
CO
CM
i i i i
I I I I I I
o
o
ro
O
CN
O
NOIIVHIN30NOD SSW
D
oo
•H
51
-------
TABLE 3. SUMMARY OF TEST RESULTS
Experiment
No.
8
9
10
11
14
15
16
17
Monitored Data
r M
Type o a C
Imp. 1.7 0.45
Method 5 - - 67
Method 5 - 68
Method 5 - 157
Method 5 - 240
Method 5 - 220
Method 5 - 94
Imp 2.9 0.60
Calculated Data
ro
1
1
1
1
2
2
2
2
.7-1.8
.7-1.8
.5
.5
.0-2.9
.4-2.9
.5-2.9
.5-2.9
4
0
0
0
0
0
0
0
.5-. 5
.4-0.45
.3
.3
.5-0.6
.5-0.6
.5-0.6
.5-0.6
Mc
89
77
84
189
237
216
105
135
.0
.0
.7
Runs 14 to 17 represent attempts to maximize ro by using particles that
were sent through the classifier and collected in the large particle output.
This particular sample had been through the classifier three times and repre-
sented the residuals after a portion of the distribution below 9 micrometers
had been removed. The dust chamber was also operated in a condition designed
to enhance the number of larger particles. Had the original sample not been
truncated by a cyclone operating upstream of the baghouse, the resulting dis-
tribution might have closely resembled a shifted log normal. The impactor run
of Figure 27 shows a value of ro = 3.1 M1 and a cr of 1.00 based on the points
in the 5 to 40 percentile. The 40 to 70 percentile points yield a value of
ro = 3.0 tA and cr =0.76. The impactor was not designed to give data beyond
this point (i.e., d = 8 micrometers). Were such points available, they would
show a gross deviation from log normal, since the dust entering the baghouse
has been passed through the cyclone and is relatively free of particles sig-
nificantly larger than 10 M- . The computer interprets this truncation plus
enrichment (affecting some 30 percent of the mass) as a reduced spread (i.e.,
CT = 0.55 to 0.60) and calculates ro to range between 2.5 and 2.9 micrometers.
The fast extraction rates used to enhance the number of large particles
make the observed results very dependent on feed rate. With slow extraction
rates, irregularities in feed rates are averaged out. The results tend to be
more constant and uniform with time. With fast extraction rates not only
does the instantaneous value of the mass concentration fluctuate with feed
irregularities but the size distribution also fluctuates. When the feed stops
completely, the large particles disappear almost immediately while the fines
can be seen (at reduced concentration) for several minutes. In experiment
52
-------
No. 14, the feed supply began to run out before the end of the run and the
value of ro began to fall. In spite of the fluctuations, the calculated
average values of the mass concentrations are always in close agreement with
the measured values.
53
-------
REFERENCES
1. Holland, A. C., and J. S. Draper. Analytical and Experimental Investi-
gation of Light Scattering from Polydispersions of Mie Particles.
Applied Optics, March, 1967. Vol. 6, No. 3.
2. Holland, A. C., and G. Gagne. The Scattering of Polarized Light: by
Polydisperse Systems of Irregular Particles. Applied Optics, May, 1970.
Vol. 9, No. 5.
3. Querry, M. R., B. Curnutte, and D. Williams. Refractive Index of Water
in the Infrared. Optical Society of America, October, 1969. Vol. 59,
No. 10.
4. Irvine, W. M., and J. B. Pollack. Infrared Optical Properties of Water
and Ice Spheres. ICARUS 8, 324-360 (1968).
5. Peterson, J. 1., and J. A. Weinman. Physical Properties of Quartz Dust
Particles at Infrared Wavelengths. Geophysical Research, December 20,
1969. Vol. 74, No. 28.
6. Twitty, J. T., and J. A. Weinman. Radiative Properties of Carbonaceous
Aerosols. Journal of Applied Meterology, August, 1971, P 725.
7. Volz, F. E. Infrared Refractive Index of Atmospheric Aerosol Substances.
Applied Optics, April, 1972. Vol. 11, No. 4.
8. Volz, F. E. Infrared Optical Constants of Ammonium Sulfate, Sahara Dust,
Volcanic Pumice, and Flyash. Applied Optics, March, 1973. Vol. 12,
No. 3.
54
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-79-022
3. RECIPIENT'S ACCESSION NO.
4 TITLE AND SUBTITLE
MULTIWAVELENGTH TRANSMISSOMETER FOR MEASURING MASS
CONCENTRATION OF PARTICULATE EMISSIONS
5. REPORT DATE
February 1979
7 AUTHOR(S)
Eli Reisman
6. PERFORMING ORGANISATION
8. PEHFOHMING ORGANIZA1 ION HU'OHl NO
9 PERFORMING ORGANIZATION NAME AND ADDRESS
Ford Aerospace and Communications Corporation
Aeronutronic Division
Newport Beach, California 92663
10. PROGRAM ELEMENT NO.
1AA010 (1AD712)
11. CONTRACT/GRANT NO.
68-02-2209
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTP, 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 7/75 - 5/78
14. SPONSORING AGENCY CODE
EPA/60Q/09
15. SUPPLEMENTARY NOTES
16 ABSTRACT
A multiwavelength transmissometer potentially capable of making near-real-time
measurements of particulate mass concentration in industrial stacks was developed.
A computer program is employed to interpret the transmissometer data and translate
the results into mass concentration. The transmissometer utilizes four different
wavelengths and records the opacity of the particulate emissions at each wavelength.
Since the response at each wavelength depends on the size of the particles, the rela-
tive values of opacity provide the computer with information on particle sizes. If
the computer is also given the wavelength dependence of the optical indices of refrac-
tion and guidelines as to the most probable distribution forms, the computer can
adjust the mean and spread of the distribution to find a best fit to the experimental
data. It then uses this information to compute the mass concentration. The theory
behind the measurement technique, a laboratory demonstration of the technique, and
the optical and electrical design of the instrument are discussed.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air pollution
Flue dust
Particles
Weight (mass)
Transmissometers
Computers
13B
21B
14B
09B
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFTFD
21. NO. OF PAGES
63
2O SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
55
------- |