EPA-600/2-77-022
January 1977
Environmental Protection Technology Series
A REAL-TIME MEASURING DEVICE FOR DENSE
PARTICULATE SYSTEMS
Industrial Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL PROTECTION
TECHNOLOGY series. This series describes research performed to develop and
demonstrate instrumentation, equipment, and methodology to repair or prevent
environmental degradation from point and non-point sources of pollution. This
work provides the new or improved technology required for the control and
treatment of pollution sources to meet environmental quality standards.
EPA REVIEW NOTICE
This report has been reviewed by the U. S. Environmental
Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the
views and policy of the Agency, nor does mention of trade
names or commercial products constitute endorsement or
recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-77-022
January 1977
A REAL-TIME
MEASURING DEVICE FOR
DENSE PARTICULATE SYSTEMS
by
P.W. Chan, C.Y. She, C.W. Ho, and A. Tueton
Colorado State University
Fort Collins, Colorado 80523
Grant No. R803532-01-0
ROAPNo. 21ADL-018
Program Element No. 1AB012
EPA Project Officer: William B. Kuykendal
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
Foreword
Pollution in the atmosphere is now recognized as a serious problem.
In a varying degree it affects all living things. Since man is the major
polluter, he must accept his responsibility to find ways to measure and
control air quality. The Physics Department of Colorado State University
has, as part of its research program, projects to develop methods of
monitoring the quantity and the movement of air-borne particles.
This report describes the design and the testing of a device which
counts and determines the size distribution of particles in the air passing
through a small volume. This is done be detecting and analyzing the light
scattered by the particles. The range of particle sizes that can be measured
by this method is important since it corresponds closely to the particle
sizes that are retained in the lungs. In situ measurements are possible
with this instrument, because the particles are not removed from the normal
air flow for the measurement. In addition, the size distribution and total
number of particles are obtained as an accumulation with time so the
measurements are in real time.
John Raich, Chairman
Department of Physics
Colorado State University
Fort Collins, Colorado
11
-------
CONTENTS
Figures IV
Tables Vi
Acknowledgment Vil
I. Introduction, Conclusions, and Recommendations ... l
II. Theoretical Background 3
2-1 Mie scattering theory 3
2-2 Application to particle sizing 12
III. Instrumentation 18
3-1 General 18
3-2 The optical module 18
3-3 Electronic module 20
IV. Experimental Methods, Results and Analysis 29
4-1 Experiments using artificially generated
pulses 29
4-2 Tests made on real particles 36
V. Summary and Future Work 56
References 58
iii
-------
FIGURES
Number Page
2-1 Scattering diagram of plane electromagnetic wave
diffracted by a sphere 6
2-2 Variation of scattered lobes with size 6
2-3 Detector output for different concentrations .... 6
2-4 Polar plot of Mie scattering functions for
a=0.2 7
2-5 Polar plot of Mie scattering functions for
a=1.0 8
2-6 Polar plot of Mie scattering functions for
a=2.0 9
2-7 Polar plot of Mie scattering functions for
a=5.0 10
2-8 Intensity ratio against size for wavelength =
0.6328 ym 11
2 2
2-9 Plots of y=e~X +e"(x~t) 15
2 2
2-10 Plots of y=e~x +0.5e~^x~t) 16
3-1 The optical module 19
3-2 Block diagram of processing circuit 21
3-3 Circuit diagram for particle sizing instrument ... 22
3-4 Input and output waveforms on comparator A-8 .... 24
3-5 a) Timing diagram for discrete pulses 24
b) Timing diagram for two overlapping pulses .... 25
c) Timing diagram for pulses too close together . . 25
4-1 Typical distribution for artificial pulses of
fixed ratio 30
IV
-------
Number page
4-2 Distribution for fixed ratio pulses of very low
amplitude .................... 30
4-3 Output channel number plotted against input ratio
'W ..................... 33
4-4 Output channel number plotted against input ratio
4-5 Response of processing circuit to two overlapping
pulses ...................... 35
4-6 Calibration curves using methylene blue and DOP
particles .................... 39
4-7 Intensity ratio against size for different refractive
indices ..................... 40
4-8 a) Distribution for methylene blue particles of
diameter 2.12pm ................. 41
b) Distribution for methylene blue particles of
diameter 2.67ym ................. 41
c) The two above distributions superimposed on
each other ................... 42
4-9 Distribution obtained using one photomultiplier . . 42
4-10 Relative response profiles for the 5° and 10°
collection systems ................ 46
4-11 Simplified optical collection system for theoretical
analysis ..................... 47
4-12 Theoretical plots of relative power intercepted
against displacement for different values of
iris diameters .................. 49
4-13 Typical distribution for DOP smoke of high
concentration .................. 55
-------
TABLES
Number Page
2-1 Probabilities of n particles in focal volume for
various N 13
4-1 Results of calibration using artificial sources ... 32
4-2 Tabulation of Ic/I,n against output channel
number 32
4-3 Results of calibration using the Berglund-Liu
generator 37
4-4 Theoretical values of intercepted power against
position 50
4-5 Experimental results for DOP smoke 52
4-6 Comparison of theoretical and experimental count
rates 54
vi
-------
ACKNOWLEDGMENT
We wish to thank Mr. W- B. Kuykendal of the Industrial Environmental
Research Laboratory, Research Triangle Park, for many helpful discussions
and Mr. J. Abbott for lending us the Berglund-Liu aerosol generator for
use in this project.
Vll
-------
Chapter I
INTRODUCTION, CONCLUSIONS, AND RECOMMENDATIONS
Recent concern over the quality of the environment has raised
interest in the field of particle sizing. Particles with diameters
below several microns are of most interest because they can remain
suspended in air for a long time and can be deposited in human lungs.
However, for this range of sizes a successful technique for measuring
the particle size distribution in real time over a wide range of con-
centrations has yet to be developed. At present, the methods for
determining particle size distributions can be classified into two
categories. The first class includes non-optical methods such as
electron microscopy or passing the sample through a series of stages
in an impactor and then weighing them successively. These methods
involve significant perturbation of the particles under investigation.
In this process the particle size distribution may be altered. Besides,
samples have to be taken and process time is slow.
Methods of the second category utilize light sources and can be
described as optical methods. The optical methods are based on the
fact that a particle cloud provides inhomogeneities in the medium
and would therefore scatter light. Different size particles have
different light scattering properties. Thus, by illuminating the particle
cloud with a light source and then observing the transmission, extinc-
tion or scattering of the source, some information about the particle
sizes can be obtained. In the transmission method a light source
illuminates the particle cloud and the resultant transmission is deter-
mined as a function of wavelength. Using this data it is possible to
extract some information about the size distribution of the cloud.2
In the photometric light scattering method, light scattered simultan-
eously from a large number of particles in part of the incident beam
is collected and analyzed. Although they introduce no perturbation
to the system under investigation, these methods have accuracy very
much dependent on the particular size distribution it is measuring.
Besides, quite complicated procedures are involved to obtain the
desired results from experimental data. Usually these have to be
performed on a digital computer. Thus, these methods do not provide
real time, in si-tu monitoring of particle size distributions.
The single particle counter looks at one particle at a time by
making its sensitive focal volume very small. The light scattered
by a particle at a certain angle (or over a range of angles) is collected.3
From the pulse height distribution of the resultant signal the particle
-------
size distribution can be determined. Because the light scattered from
a particle depends not only on its size but also its refractive index,
this technique cannot handle polydispersed systems with unknown (and
often complex) indices of refraction. Due to the fact that this
situation is often encountered in actual practice, better techniques
need to be developed.
Hodkinson4 made an important contribution when he realized that
the shape of the forward lobe scattered from a particle is due primarily
to Fraunhofer diffraction and has therefore a minimum dependence on
refractive index. Thus, by measuring the intensity ratio of scattered
light at two forward angles, the particle size can be determined quite
independent of refractive index. This instantaneous intensity ratio
technique provides a means of monitoring polydispersed systems in real
time. However, because the particles have to be viewed one at a time,
the concentration that can be handled by this technique is quite low.
It is limited by the smallest focal region that can be achieved in practice.
This report deals with the design, building and testing of an
instrument that is capable of handling dense particulate systems.
Chapter II gives the fundamental Mie scattering theory and intensity
ratio technique that form the background for this research. The concepts
for increasing the concentration handling capability are also discussed.
With those concepts in mind, the optical and electronic instrumentation
are designed. Chapter III describes the instrumentation in detail and
explains how the proposed instrument applies the theoretical concepts.
In Chapter IV the performance of this instrument is discussed. Its
limitations are evaluated and the maximum concentration that can be
handled is investigated. Finally, in Chapter V the results obtained
from this study are summarized. Suggestions for improving the instru-
ment are made and possible future work in this area is suggested.
A real-time device having the potential of increasing the partic-
ulate concentration handling capability by one order of magnitude has
been constructed and tested. It is based on analyzing the intensity
ratio of scattered light at two angles. Contrasted to present particle
sizing devices the device allows more than one particle to be viewed in
the focal volume at any instant and selects the scattered light pulses
randomly for data processing. The device can handle an average of N =
2.5 particles in the focal volume (compared to N = 0.2 of other similar
devices) for particle size from 0.6 to 3.5 ym in real-time with good
accuracy and resolution.
Immediate improvement on the optical module to increase the con-
centration handling capability by one order of magnitude to 107 par-
ticles per c.c. is of utmost importance. The system's resolution should
be improved and the size range handling capability should be enlarged to
both smaller and larger particles. A rugged device could then be built
for field testing and use.
-------
Chapter II
THEORETICAL BACKGROUND
The discussion below is based on spherical particles. For non-
spherical particles, which are often encountered in actual practice,
theoretical solutions have been solved only for certain particular cases
(e.g., cylinders, ellipsoids). It is, however, noticed that a poly-
dispersed, randomly oriented nonspherical system of particles has the
same gross light scattering characteristics as that for spherical
particles if the scattering angles are in the forward direction.
Hence the results below are applicable generally.
2-1 Mie Scattering Theory
The problem of scattering of a plane monochromatic electromagnetic
wave by a sphere of arbitrary size and refractive index has been solved
by Mie in 1908.6 In short, this is a problem of solving Maxwell's
Equations with the appropriate boundary conditions imposed by the
presence of the sphere in the field.
Consider a plane wave travelling along the positive z axis and
incident on the sphere as shown in Figure 2-1.
Here the origin is taken at the center of the sphere, the positive
z axis along the direction of incident wave propagation and the polar-
ization of the wave is in the direction of the x axis. The incident
field has unit amplitude for simplicity.
In spherical coordinates, the resultant wave equation is variable
separable. The scattered components of the electric field EG and E
are spherical outgoing waves given by:
E6 = i- e"1 r ia)~cos c|> S2(6,a,m)
nj. i -ikr+iut . , _ ,Q .
-E S..(9,ot,m)
K.1T 1
where r is the distance from the origin to point of observation
TTd
a = size parameter = -—
A
-------
m = refractive index of sphere relative to the medium
d = diameter of sphere
k = wave number = -—
A
Si, 82 are scattering functions, each being a sum of infinite
series of Legendre Polynomials and spherical Bessel functions.
The radial component Er may also be derived from Mie's solution but
it tends to zero with a higher power of 1/r.
In general, the scattered light is elliptically polarized even in
this case, when the incident radiation is linearly polarized. This is
because the functions Si and Sa are complex numbers with different
phases.
The intensity at the point of observation can thus be calculated.
i = ie + i
= |EO| + |E<|>|
cos |so(a,m,6)| + —-—— sin S..(a,m,) .
,J ~ ^^.J lf> | u_ ^u,!!*, v/ | I « O
kY" \f T"
L K. L
Putting
i _ s . -i. i 2
=i1(a,m,e); S2(a,m,8)
The equation becomes
X2 2 2
I = —2~2 '•sln (f)i1(a>m»9) + c°s i2(a,m,6)J
Air r
For an incident wave of intensity lo,
2
I = —^y [sin cj>i (a,m,6) + cos (f>i (a,m,6) ] (2-1)
4u r l
Equation 2-1 gives the expression for linearly polarized incident
light of intensity lo. If the incident light is randomly polarized,
as in the case of natural light, an averaging over the angle has to
be performed to get the intensity scattered. Since cos2(|> = sin2(j> = 1/2,
the expression for scattered intensity becomes:
-------
Sir r
i (a,m,9) + 12 (a,m,6)] (2-2)
Simplified theoretical solutions exist for various classes of
particle sizes and refractive index.7 For example, for small particles
(a«l , Jm a«l), the first term in the expressions for ij and ±2 is
predominant and scattering is due primarily to electric dipole oscil-
lations. For large particles (an order of magnitude greater than the
wavelength) , the scattering pattern can be found by superposition of
Fraunhofer diffraction and reflection and refraction. For refractive
index close to 1, the Rayleigh-Gans theory may be applied to give a
simpler solution. However, for the particle size range of interest
(0.05 - lOura) , no simplifications exist and the full Mie solution must
be used.
From equations 2-1 and 2-2 one can observe that the scattered
intensity does depend on the size parameter a. Better physical insight
and ways to utilize the Mie scattering theory for practical applications
is not easy to achieve because of the complexity of the functions i]
and ±2. Fortunately, during the past decade many numerical tabulations
of the scattering functions (i^ and ±2) 8»9 for spheres with different
sizes and refractive indices are available. With the help of these
tables, full potential of using the Mie theory in practical applications
can be appreciated. Figure 2-2 shows qualitatively the variations of
scattering pattern with size.
It is observed that, for small particles, the forward lobe is
wide with a small backward lobe. As the particle size is increased,
the forward scattering lobe narrows and becomes strongly enhanced,
the backward scattering lobe becomes somewhat enhanced, and a number
of weak side lobes develop. The number of side lobes increases with
increasing particle size. Hence, by detecting the intensity variation
scattered from the forward lobe, information about the size of the particle
can be obtained. Figures 2-4 to 2-7 show polar plots of the two
components of intensity as a function of scattering angle (0 - 90°)
for different sizes and refractive indices. From these figures, one
notices that the scattered intensity also depends strongly on the
refractive index. Therefore, to size a polydispersed system of particles
with unknown index of refraction, the scattered intensity at one angle
does not give sufficient data. However, one important observation is
that the shape of the forward lobe is by and large independent of
refractive index and varies only with particle size. Hodkinson noticed
this and proposed that, by measuring the intensity scattered at two
different forward angles and then taking a ratio of the two, the effect
of refractive index can be minimized. To illustrate the variation of
intensity ratio with size, curves of intensity ratio vs. size within
the forward lobe are reproduced in Figure 2-8. This figure is reproduced
after Kreikebaum and Shofner10 and is calculated assuming spherical
particles with imaginary component of refractive index > 1. Depending
on the range of particle sizes to be covered, a pair of angles can be
selected. The 10°/5° is used as an example for discussion. From this
figure several characteristics of this technique can be observed:
-------
Scattered
Wave
Figure 2-1.
Incident
Wave
Scattering diagram of plane electromagnetic wave diffracted
by a sphere.
o
Increasing a
Figure 2-2. Variation of scattered lobes with size (a= ird/X).
Non overlapping Pulse
(a)
Overlapping Pulse
(b)
Figure 2-3. Detector output for different concentrations:
a) Low concentrations b) Higher concentrations
6
-------
60C
Figure 2-4. Polar plots of Mie scattering functions for size parameter
a » ird/A = 0-2 of different polarization and refractive index.
-------
60
oo
Figure 2-5. Polar plots a - ird/X • 1.0.
-------
90
30°
i,! /as x I02
i,/as x 10°
Figure 2-6. Polar plots a - ird/X - 2.0.
-------
90
12
14
Figure 2-7. Polar plots a = ird/X - 5.0.
-------
X = 0.6328/xm
0.7 1.4 2.1 2.8
Particle Diameter , img(n)>l ,(/zm)
Figure 2-8. Intensity ratio vs. size for wavelength = 0.6328 vim.
3.5
-------
1) The curve is decreasing with increasing particle size inside
the forward lobe. Hence a one to one relationship between particle
size and intensity ratio is obtained.
2) The curve is relatively flat for small a, i.e., the intensity
ratio is insensitive to particle size variations. This imposes a limit
on the smallest size that can be measured.
3) When the particle size becomes large, the scattered intensity
at the outer angle (10° in this case) becomes smaller because it is
approaching its first minimum. The upper limit to particle sizing is
thus set by the sensitivity of the scattered intensity detector and
its background noise level. To extend the upper limit to a larger size,
a smaller pair of angles may be used (e.g., 2.5°/5°). For the 10°/5°
case, using a visible light source, the practical range of size detection
is ~0.1-4um.
4) Figure 2-8 is plotted using an imaginary component of index
> 1. Particles with large absorption actually give a smoother intensity
ratio -a curve. For real indices, small oscillations can be observed
in the curve (Figure 4.7) and can cause errors. In any case the maximum
error due to refractive index is estimated to be 15-20 percent.
2-2 Application to Particle Sizing
Two techniques applying this principle may be used for measuring
particle size. They are described below:
1) Time averaged measurements: If two slow detectors are used
to detect the scattered intensity from a cloud of particles, a time-
average will be measured. The resultant intensity ratio will be given
by the following formula:
/i(0)f(a)da
/i(82)£(a)da
where f(a) represents the particle size distribution function.
Theoretical calculations of p(e1/92) for various bell-shaped
distributions with mean a and deviation p can be made. By comparing
the measured results with those calculated, the parameters of the size
distribution (a and p) can be deduced. This method is, however, time
consuming. Furthermore, only certain distributions can be handled
and it cannot be considered as a real-time method to monitor size
distributions. Hence this topic will not be discussed further.
2) Instantaneous measurements: The advantages of intensity
ratio method can be fully utilized if the particles can somehow be
viewed one at a time and the intensity ratio measured and recorded.
The scheme is as follows:
12
-------
A light source (laser for example) is focused down to a very
small volume and the particle cloud under investigation is allowed to
flow through the focal volume. When there is no particle flowing
through the focal volume at a particular instant, the detectors for the
two angles give no signal other than background noise fluctuations.
Whenever a particle passes through the focal volume, light is scattered
to yield outputs on the two detectors. The pulses are immediately
processed and the ratio of the intensities taken. By means of a pulse
height analyzer, the frequency of occurrence of intensity ratio current
pulses as a function of pulse height (and hence particle size) can be
obtained. If the pulse height analyzer is calibrated beforehand in
terms of particle size, the distribution can be obtained immediately.
By taking a ratio of the scattered intensities, experimental errors due
to variations in source output power can be eliminated. The fact that
this technique does not assume anything about the shape of particle size
distribution means that it can handle nearly any type of particle
distributions. The desirability of the method over method (1) is thus
evident.
One flaw in this technique is that only one particle at a time is
allowed in the focal volume. This would set an upper limit on the
concentration that can be correctly detected. Consider a unit focal
volume through which a particle system of average concentration N/unit
volume is flowing. Assuming that the particles are distributed randomly
in space at a particular instant, the probability of n particles present
in the unit volume at any given instant is given by the Poission
distribution
P(n) =
-Nrr
e N
n!
(2-3)
The following table gives S£me values of probability for different
values of average concentration N:
Table 2-1. Probabilities of n particles in focal volume for various N.
N P(0)P(l) P(2) P(3)
1
0.8
0.6
0.3
0.2
0.1
_i
3.68x10 7
4.49x10
5.49xlO~f
7.41x10":
8.19x10":
9.05x10"
_i
3.68x10 J
3.59x10 :
3.29x10 :
2.22x10
1.64x10
9.05x10
_i
1.84x10 :
1.44x10
9.88x10 y
3.33x10 2
1.64x10"
4.52x10
_2
6.13x10
3.83x10
1.98x10
3.33x10
1.09x10 .
1.51x10
The above figures indicate that in order to keep the probability of
more than one particle in the volume at a given instant quite low, the
average concentration that can be handled for a unit volume is much less
than 1. For example, for an error of ~ 1.64 % maximum, maximum N that
can be handled is 0.2 (in which case P(2) = 0.0164). For a volume as
small as 10~7 cm3, which can be achieved by using a laser as light
13
-------
source and good focusing lens, the maximum concentration detectable is
about 2>cl06/cm3. Dense particle systems have to be diluted before
being counted; and in the dilution process the size distribution may be
altered.
It is realized that by committing oneself to just one particle in
the focal volume at a time, the measuring system is not very efficient11.
Larger concentrations can be handled if the number of scatterers in the
volume is greater than unity.
When there is just one particle traversing the focal volume at a
particular instant, the detector outputs record non-overlapping pulses
as shown in Figure 2-3a.
If there is more than one particle in the scattering volume, the
detector outputs will record overlapping pulses like one shown in
Figure 2-3b. In this case one assumes that the particles enter the
focal volume at slightly different instants so that two peaks can be
observed. The intensities are also assumed to add. If one can somehow
identify the two pulses and process them independently, an improvement
in the detectable concentration can be achieved. Referring to Table 2-1,
if two overlapping pulses can be identified, the maximum concentration is
N ~ 0.6 (in which case P(3) = 1.98*10~2) for the same degree of accuracy.
In principle, if more overlapping pulses can be handled, the manageable
concentration is increased even more. However, there is a rigid limit
to this and the case is discussed below.
Consider a TEM beam as light source. The intensity distribution
at the cross section is then Gaussian. Therefore the scattered light
pulses will also be Gaussian. In Figures 2-9 and 2-10, some theoretical
curves for two Gaussians with different degrees of overlap are shown.
In Figure 2-9, the two pulses are of same peak amplitude, while in
Figure 2-10, one amplitude is half the other.
It is seen that as the two pulses come closer to each other, the
dip in the middle becomes less and less. Then, at some degree of
overlap dependent on their relative amplitude, the two pulses merge
into a bigger pulse (Figure 2-9a and 2-10a). Experimentally the above
case cannot be resolved and erroneous results will be obtained.
Besides, when the concentration is too high, the pulses tend to become
small ripples riding on a d.c. level and their intensity ratios do not
necessarily represent the particle size.
One way of solving the above problem is to place a limit on the
degree of overlap the measuring system will process. If one can, by
proper instrumentation, arrange the data processing system such that
pulses too close together are discarded, the error due to the above
effect can be minimized. In doing this one assumes that the inter-
arrival time of particles is random and independent of its size. In
other words, two big particles have the same probability of arriving
at the focal volume at nearly the same time as that of two small
particles. Thus the measuring system would be sampling randomly a
14
-------
Figure 2-9,
Plots of y - e
15
x
+
-------
t =
(a)
(b)
t = 3
-2
0
(O
Figure 2-10.
2 .2
Plots of y - e"X + 0.5e~Cx~t} .
16
-------
distribution of particles. By the theory of statistics, the samples
taken should have the same size distribution as that of the main stream.
Bearing these concepts in mind, the design of the optical and
electronic system to achieve the above objectives is discussed in the
next chapter.
17
-------
Chapter III
INSTRUMENTATION
3-1 General
The optical system for this instrument is required to achieve the
following:
a) Define a finite sampling volume through which the system of
particles under investigation traverses.
b) Collect the light scattered from those particles at two forward
angles, 5° and 10° in this case.
Main design considerations are the minimization of stray light and
the minimization of focal volume. The first one determines the signal
to noise ratio of the instrument, while the second one determine the
maximum concentration the system can handle.
The electronic module does the processing of the light pulses
collected. It is required to take the ratio of the peak value of the
pulses scattered at the two angles and accumulate those values to yield
the appropriate size distribution. These two modules are discussed in
more detail in the following sections.
3-2 The Optical Module
Figure 3-1 shows the optical setup for this instrument. A He-Ne
laser (A=0.6328y) is focused down by a lens (LI). The plnhole P is
made out of flat black paper and it serves to restrict and clean up
the edges of the laser beam and also reduce reflections from the lens
L£. The light scattered at the correct angles (5° and 10°) is selected
by concentric rings of appropriate diameter. The use of three of these
ensures that stray light coming in at an unwanted angle is blocked out.
At the other side of the rings convex lenses L2 and 1,3 are used to
focus the scattered light to one point to be picked up by optical fibres.
In the original arrangement, the collection slits for the 5° and 10°
scattering angles were parallel. However, measurements of the focal
volume revealed that the region seen by the 5° collection system did
not exactly coincide with the region seen by the 10° collection system.
This is due largely to spherical aberration of the lenses used. The
origin of the fault is that for an uncompensated lens, parallel light
that comes in near the edge of the lens has a shorter focal length
18
-------
Beam Stop
VD
Laser
Source
-ff-
To Phototube 2
Figure 3-1. The optical module.
-------
than light that comes in closer to the axis. If parallel light were
collected for both angles, the 5° would see a region which is further
away from the lens than the 10°. To correct for this, the slits for
the 10° collection are made such that they collect light which is
slightly converging. The inner rings (5°) collects light that comes
in essentially parallel. By this method good overlap of the two focal
regions can be achieved.
The entire collection system is housed in an aluminum tube. This
makes the system rigid and easy to align. Stray light is also mini-
mized to the extent that the system can operate in ambient light con-
ditions without stray noise becoming too excessive. Its operation can
be improved if appropriate spike filters are inserted in front of the
collection lens to allow only light of the correct wavelength (wave-
length of laser used) to pass through.
3-3 Electronic Module
The electronic module processes the photomultiplier outputs and
converts them to a particular size distribution. It is designed to
achieve the following objectives:
1) Detecting the peak value of the scattered pulses and taking
the value of their ratio. This minimizes the error in the ratio due
to stray background noise.
2) To identify two overlapping pulses due to two particles
traversing the focal volume at slightly different instants and obtain
their intensity ratio.
3) To neglect pulses that are so close to each other that the
value of their ratio is no longer correctly related to their size.
Figure 3-2 shows a block diagram of the electronic processing
circuit and Figure 3-3 shows a complete circuit diagram. The operation
of the circuit is as follows:
The current pulses from the two photomultipliers, each terminated
in lOkft resistors, are fed into operational amplifiers Al and A2, which
are voltage followers acting essentially as current to voltage con-
verters. The outputs from Al and A2 go to two sample and hold modules
(Analog Devices AD583) for data processing. These voltage pulses are
also fed into the timing and control unit which generates the
appropriate control and coincidence signals for accurate data acquisi-
tion. Amplifiers A3 and A4 serve as isolation amplifiers for the
triggering and incoming circuits. The output of A3 goes into the
peak detector trigger circuit. This consists of amplifiers A5, A6
and comparator A8. Amplifier A5 is connected as an ordinary voltage
inverter. A6 has a parallel R-C combination connected in its input.
This has two effects: it attenuates the incoming signal a little and
also tends to delay it by a small amount. However, this amplifier has
a slightly bigger gain so that the output is an inverted and also
20
-------
Current
Voltage
Converters
Sample a Hold
Command
Photomultiplier
Output 5°
Photomultiplier
Output 10°
N>
Sample
a
Hold
Switch Control
Sample 8 Hold
Command
J_
Sample
8
Hold
Log
Ratio
Module
(756P)
Antilog
Module
(755N)
CMOS
Switch
Pre-
amp
Pulse
Height
Analyzer
Inv.
amp
Level
Discriminator
Trigger
Inv.
amp.
Peak
Detector
Trigger
Timing
a
Control
Sample a Hold
Command
Switch
Control
Digital Counter
Figure 3-2. Block diagram of processing circuit.
-------
KJ
LMI458
IIOK
LMI458
IIOK
IIOK"
Figure 3-3. Circuit diagram for particle sizing instrument.
-------
slightly delayed version of the input signal. This pulse and the
output from A5 form the input of comparator A8. This comparator
detects the crossover points and would therefore trigger whenever there
is a change of slope. This is shown in Figure 3-4.
By this method the peak of the pulses can be detected with good
accuracy. The amount of delay depends on the rise time of the input
pulse. Hence the actual trigger point will also vary a little as the
overall pulse width is varied. However, this arrangement works fine for
a large range of pulse widths that are of interest.
The output of amplifier A4 is connected to comparator A7. The
other input is connected to an adjustable d.c. This comparator triggers
whenever the input is above the preset d.c. level. The function of
this is to provide a discrimination level. Input pulses which are
below this d.c. level are neglected by the system. Normally this d.c.
level is set so that it is bigger than random photomultiplier fluctua-
tions but smaller than the smallest pulse that originates from light
scattered by particles in the sensitive volume.
The outputs from the two comparators (TTL level) form the basic
signals for timing of the system. Figure 3-5 shows timing diagrams
for the circuit. Three cases are to be considered: 1) nonoverlapping
pulses, b) two overlapping pulses, and c) a series of overlapping
pulses so close together that the values of their ratio no longer give
a fair representation of their sizes.
First of all consider case a). The relative timing is shown in
Figure 3-5a. When there is no input at all the outputs of both A7 and
A8 are high. The output of A13 is therefore low. This output provides
the control signal for the two sample and hold modules Bl and B2 and
causes them to stay in the hold mode. When a pulse is detected com-
parator A7 triggers first and goes low. The output of All goes high.
On the other hand, the output of A10 is still clamped high because
peak triggering has not yet occurred. The sample and hold modules
switch to the sample mode. The outputs of Bl and B2 then follow the
values of the input. When the input pulse reaches its peak, com-
parator A8 triggers and goes low. This forces the output of A10 low and
hence output of A12 high. The sample and hold control signal goes low
and switches the system to hold mode, holding the peak values of the
input pulses. The outputs of Bl and B2 are connected to a logarithmic
module (AD756P). This module takes a ratio of the two inputs accord-
ing to the following transfer function
ii
e = -k log -^ (3-1)
gi2
where i = signal current
= 10° output in this case
23
-------
Input
Figure 3-4. Input and output waveforms on A8.
Output
Input
A7
A8
AI3
BI.B2
B3
AI5
AI6
AI4
Figure 3-5a. Timing diagram for discrete pulses,
24
-------
Input
+5V
+5V
A7
A3
+5V
AI3
A7
A8
+5V
A14
AI5
_n
AI3
AI6
AI4
Figure 3-5b.
Timing diagram for
two overlapping
pulses.
Figure 3-5c,
Timing diagram for
pulses too close
together.
25
-------
i = reference current
= 5° output in this case
k = -1 .
This is followed by an antilog module with transfer equation
e' = E ref l(f e ±n/k (3-2)
o
where E ref = reference voltage
= 0.1V
e in = input voltage
k = 1 .
The overall transfer equation of the two modules is given by
e' = 0.1 -r^ (3-3)
There are several advantages in using a logarithmic setup. Firstly,
it allows a large range of input current values. Secondly, a ratio
range of a million to one can be handled. Although in the present
system the antilog module is used to give a linear output, slight modi-
fications can be made to give a logarithmic output ratio and therefore
a logarithmic distribution on the pulse height analyzer. One disadvan-
tage of these modules is that they have a relatively slow frequency
response. The use of sample and hold modules for input eliminates
this problem since they hold the peak value of the input pulses. Thus
the logarithmic ratio circuits see essentially d.c. inputs.
The trailing edge of the sample and hold pulse triggers monostable
multivibrator A15. The output of A15 is about 20ysec wide. It acts
as a delay one-shot, allowing sufficient time after the hold command
for the logarithmic modules to reach their steady state output before
data is actually taken. It rising edge (since output is Q this occurs
20ysec after hold command) triggers single-shot A16. A16 has an out-
put pulse of approximately 2ysec and it controls the opening and
closing of a CMOS switch (CD4016). Hence at this instant the switch
closes for 2psec, samples the value of the ratio and passes it into the
pulse height analyzer. This switch is required to provide fast rise
time pulses acceptable to the pulse height analyzer. The PHA will
detect the value of the voltage pulse and store it in the appropriate
channel. Single shot A16 also forms the input to a digital counter and
this gives the total number of pulses counted. From this number and
the time for which data is taken the average concentration of the
particle stream can be determined.
26
-------
The falling edge of the peak detecting pulse (A8) also triggers
another monostable A14. The period of this is made longer than the
total time required for a data point to be taken (~ 23usec). This
pulse is fed into the input of nor gate A13, and once triggered, it
clamps the output of the sample and hold command in the hold mode
for a time equal to its pulse width.
When the pulse from A14 goes away, the measurement system is
completely reset and is ready to accept new data.
Next, case b) is considered. This occurs when two particles
arrive at the focal volume at slightly different instants so that there
is a finite time during which both particles are scattering light from
the laser beam. The result is an overlapping pulse like one shown in
Figure 3-5b. The operation and timing of the circuit is as follows:
At first when the input exceeds the background discrimination
level comparator A7 triggers as in the single pulse case. The circuit
functions in exactly the same way, taking the ratio of the first peak
and passing it to the pulse height analyzer. However, at the first
dip, output of comparator A7 does not reset because the voltage level
is still above the d.c. noise discrimination level. The comparator A8
will, however, reset to high because of its peak detecting character-
istic. The output of nor gate A10 therefore goes high. Because the
other input to A12 is held high by comparator A7, the output of A12
goes low. If by this time the preceding measuring cycle due to the
first peak has terminated, output of monostable multivibrator A14 will
be in low state. This allows the output of A13 to go high again.
The sample and hold modules switch to the sample mode. When the peak
of the second pulse is detected, the sample and hold modules change
to hold mode and a new measuring cycle is initiated. It is seen that
overlapping pulses which are not too close are recognized and counted
by the circuit.
Finally, case c) is considered. This occurs when a chain of
particles very close in space flows through the focal volume. The
output will consist of a series of rippling pulses overriding on a
certain d.c. level. Under such circumstances the ratio of the pulses
are no longer accurate representation of their size. The circuit is
designed to ignore most of these. Figure 3-5c shows the timing involved.
When the first pulse comes in the circuit operates in the same
way as a) and b). The ratio of the first peak will be recorded. When
the first dip comes output of comparator A8 will change state. However,
because the ripples are close together, the output of monostable A14 is
still high from the previous measuring cycle. This inhibits the sample
command and clamps the sample and hold modules in the hold mode. When
the next peak is detected the output of A8 goes low again. This high
to low transition retriggers monostable A14 and keeps it in its high
state. Thus no measurement cycle is initiated as long as the peaks
are too close.
27
-------
The point at which the circuit ceases to recognize overlapping
pulses depends on the period of single shot A14. This can be adjusted
to suit the flow rate of the particle stream through the focal volume.
According to the concept outlined in Chapter II, this design should
allow much higher concentrations to be detected without appreciable
degradation in accuracy.
28
-------
Chapter IV
EXPERIMENTAL METHODS, RESULTS AND ANALYSIS
This chapter describes the experiments performed on the system and
discusses the results obtained. The response to artificially generated
light pulses under various conditions were first analyzed. Then the
system was tested and calibrated by monodisperse aerosols of known
diameter and relatively low concentration. Finally, particle streams of
high concentration were fed into the system to determine its concentra-
tion handling capacity.
4-1 Experiments Using Artificially Generated Pulses
In order to determine the response of the electronic system, an
ideal source is required. This was implemented by an LED driven by a
sinusoidal oscillator. The amplitude and pulse width of the output can
be conveniently adjusted by changing the amplitude and frequency of the
sinusoidal input. The optical module was not used and the LED was
placed directly in front of the collection photomultiplier tubes. In
this way the electronic circuit detected pulses whose ratio was a
constant. Theoretically the pulse height analyzer should record a
single channel distribution. The actual distribution obtained was about
10-15 channels wide and is shown in Figure 4-1. This broadening of the
distribution could amount to a 5% uncertainty in particle sizes and is
due to the following factors:
a) There is an inherent electronic spread in the circuit. This
means that input pulses of a single ratio are recorded as pulses of
slightly different ratios.
b) Optical spread due to the photomultipliers also causes broaden-
ing. This means that given a definite input intensity, the output
current from each phototube would actually vary a little. This varia-
tion is due to differences in the photomultiplier characteristic and
dark current.
The input pulse width was varied by adjusting the input sine wave
frequency. It was found that no appreciable degradation of the output
response occurred for a pulse width range of ~ SOOysec to SOysec. The
limit is set by the response of the peak-detector trigger circuit. If
the pulse is too wide, it tends to rise very slowly and erratic peak
triggering results. If the pulse is too narrow, the peak trigger tends
to occur at a time when the input has long passed the peak value. A
shift in the output distribution would result.
29
-------
Figure 4-1. Typical distribution for artificial pulses of fixed ratio,
Figure 4-2. Distribution for fixed ratio pulses of very low amplitude,
30
-------
This range of acceptable pulse widths limits the particle flow
rate this instrument can measure correctly. For example, 100pm beam
gives a pulse width of lOOysec for a flow rate of 100 cm/sec. However,
by varying the time constant in the peak detector delay circuit, the
system can be made to handle very slow or very fast flow rates.
The input intensity was varied by adjusting the amplitude. Theoreti-
cally no change in output ratio should occur. Experimentally the
distribution obtained stayed practically constant for a large range of
input pulse amplitudes unless the input amplitude is too low.
Signal to Noise Ratio Considerations
As the amplitude of the input pulses is reduced, noise inherent in
the system becomes more and more prominent. The result is that the peak
of the distribution stay constant but as a whole it became broader
corresponding to a size uncertainty of a few percent. This is shown in
Figure 4-2, where the input pulse amplitude is about 0.03V, correspond-
ing to an input current of 3uA. This becomes important when measuring
small particles since they scatter much less light. Big particles near
the minimum in the forward lobe are also affected. There are two
contributions to the noise in the system. One contribution is due to
the dark current of the photomultipliers and the baseline noise of
electronics. It is independent of the light source used. By using a
laser source with big enough power, even very small particles can
scatter enough light to make the noise unimportant. There is, however,
another noise contribution which is due to stray light from the laser
source that gets scattered off various parts of the optical collection
module. The magnitude of this is proportional to the source laser
power. Thus this signal to noise ratio cannot be improved by using a
higher power laser. This part of the noise can be minimized by using
antireflection coated lenses and better shielding of the collection
system.
The response of the instrument to input pulses of different ratios
was also investigated. A laser was chopped by a rotating plate with a
small hole in it. The light that came through was then scattered off a
translucent piece of plastic and collected by the two photomultiplier
tubes. In this manner Gaussian shaped pulses very similar to those
encountered in real particles were generated. The pulse width was about
ISOpsec. The ratio seen by the instrument was varied by inserting
different neutral density filters in the path of either one of the
phototubes. The maximum values of the input pulses were read from an
oscilloscope and the corresponding distribution of the pulse height
analyzer was recorded. Table 4-1 shows the results.
31
-------
Table 4-1. Results of calibration using artificial sources.
Input on
10° PMT
2.2
2.2
2.2
2.2
1.6
1.6
Input on
5° PMT
1.4
1.95
3.2
3.6
3.4
3.9
Ratio
1.56
1.12
0.69
0.60
0.47
0.41
PHA Output
(Channel No.)
40
70
140
180
220
260
A graph of input ratio (10°/5°) against output channel number is
plotted in Figure 4-3. From this curve two points can be observed:
1) A larger input ratio results in a smaller channel number. This
was intentionally done using a combination of a P-type module for the
logarithmic ratio circuit and a N-type for the antilogarithmic module.
If Figure 2-7 is referred to, one sees that as the particle size in-
creases, the intensity ratio (10°/5°) decreases. Using this transfer
characteristic, larger particles are placed to the right and smaller
particles to the left in the output distribution.
2) The price paid for making the above adjustment is that the
input output relationship is no longer linear. In fact if one refers to
equations 3-1 and 3-3, it is seen that if the input to the antilog
modules is k log -.-'- , the output is not proportional to v*- but rather
io l2 l2
-r2- . The table below gives the values of 5°/10° against output channel
number for the same set of data:
Table 4-2. Tabulation of I /I., against output channel number.
^O^S
1.56
1.12
0.69
0.60
0.47
0.41
Vho
0.64
0.89
1.45
1.68
2.13
2.44
Output Channel
No.
40
70
140
180
220
260
A plot of 15/1^0 against output channel number is shown in Figure
4-4. It is linear, showing that the circuit performs as expected.
In order to check the ability of the instrument to detect two
particles, an artificial source which generated two overlapping pulses
was exposed to the photomultiplier tubes. The results are summarized
in Figure 4-5. In Figure 4-5a the upper trace shows the input waveform
as detected by the photomultipliers. The lower trace shows the output
32
-------
CO
300r
250-
200
0>
o 150
JC
O
3
O.
100
50
10 0.2 0.4 0.6 0.8 1.0
Input Ratio , IIO/I5
1.2 1.4
1.6
Figure 4-3., Output channel number plotted against input ratio (I- /I ).
-------
UJ
-P-
3.0
2.5
2.0
o>
c
Q.
3
O
1.0
0.5
40 80 120 160 200 240
Input Ratio , I5/IIO
Figure 4-4. Output channel number plotted against input ratio (I /I ).
280
-------
(a)
(b)
•igure 4-5. Response of processing circuit to two overlapping pulses.
.,
-------
of gate A13 i.e., the sample and hold command. It is seen that the
circuit successfully detects the peak values of the input and holds
them for processing. In Figure 4-5b, the upper trace shows the same
sample and hold command while the lower trace shows the output of the
analog switch which is fed into the pulse height analyzer. One observes
that the data is sampled a little while after the input hold command
(~20usec) to allow time for the ratio circuit to settle down. The
switch period is about 2ysec. This demonstrates that the proposed
circuitry can identify two pulses overlapping to about half the full
width.
4-2 Tests Made on Real Particles
A) Calibration with Monodisperse Aerosols of Low Concentration
In order to calibrate this particle detecting system, a source
providing uniform aerosols of known size is required. For this purpose
the Berglund-Liu Monodisperse Aerosol Generator (Model 50A) was used.12
In this particle generator, the desired aerosol is first dissolved in
a volatile solvent in certain known concentrations. The resultant
solution is then placed in a syringe. It is then pushed through a
small orifice at a very steady rate by an infusion pump. The orifice
is vibrating at a frequency which is controllable by an external sinusoidal
oscillator. Under such conditions the solution stream is broken up into
uniform droplets, one being generated for each cycle of disturbance.
These droplets are then injected into a dry air flow which disperses
them and dilutes them by a large amount. This large volume of air
also allows the volatile solvent to evaporate, leaving the nonvolatile
solute which then comes out as uniform particles. The diameter of the
particles thus formed can be calculated as follows:
Let the volumetric concentration of the nonvolatile solute in the
solvent be C.
The orifice vibration frequency be f Hz.
The liquid flow rate through the orifice be Qcm3/sec.
Then, for each cycle of disturbance, the liquid drop generated
would have a volume -^ cm .
After the volatile solvent has all evaporated, volume of the
particle = -jf—.
Assuming the aerosol is a sphere, the diameter d is given by
d =1^1 1/3 (4-1)
36
-------
Since Q, C and f are all known quantities in an experiment, the
diameter of the output aerosol need not be measured by another method
(e.g., microscopy).
In the experiment methylene blue in isopropyl alcohol and water
(50:50) and also Dioctyl Phthalate (DOP) in isopropyl alcohol were used
as particle sources. These two types of particles have different
refractive indices and can reveal the dependence of this ratio technique
on refractive index. The particle diameter was varied by changing the
frequency of the sinusoidal oscillator. The range thus obtained was
not large since the particle generator can only operate within a limited
range of frequencies. Using a 10pm orifice the operable frequency
range was found experimentally to be 70KHz - 190KHz. To obtain a
wider range of particle sizes, different concentrations were also
used. The particles generated were allowed to flow through the focal
volume and measurements taken. The particle diameter was calculated
using equation 4-1. The table below shows the results of particle
size with the channel number at the peak of the output distribution:
Table 4-3. Results of calibration using the Berglund-Liu generator.
(a) Dioctyl Phthalate (DOP).
Particle Diameter Channel No. at Peak
(pm) of Distribution
3.23
2.82
2.57
2.04
1.62
1.28
1.01
270
160
130
96
105
78
66
(b) Methylene Blue.
Particle Diameter Channel No. at Peak
(ym) of Distribution
3.37 290
2.67 170
2.12 122
1.68 95
1.47 90
1.33 80
1.16 75
37
-------
If the concentration of the solution is too low, soluble impurities
present in the solvent becomes a significant part. The diameter of the
resultant particle can no longer be accurately determined by Equation
4-1. For this reason no quantitative data was taken for particles much
smaller than 1 ym.
Figure 4-6 shows a plot of the particle size against output position
on the pulse height analyzer for both DOP and methylene blue. Several
characteristics of the two curves need to be discussed.
(1) First consider the curve for methylene blue particles. The
theoretical curve (Figure 2-7) shows that the relationship of intensity
ratio (10°/5°) to particle diameter should be fairly linear between
3 ym and 1.5 urn. The experimental curve is not linear but convex.
This can be attributed to the fact that the circuit, instead of taking
a ratio of 10°/5°, actually takes the ratio 5°/10° as explained in
section 4-1, where the same effect is observed when calibrating the
instrument with known ratio pulses.
(2) If the intensity ratio method were perfect, the two curves
shown in Figure 4-6 would overlap exactly i.e., particles of the same
size would always have the same intensity ratio. This is obviously
not true. To make it worse, the calibration curve for DOP shows a
dip in the size range of 1.5-2.0 u.m, making the instrument insensitive
when sizing particles in that range. This dip is not observed for
methylene blue. This discrepancy can be explained by the difference
in refractive index of the two types of particles. Being transparent,
DOP should have a refractive index which is almost real and its value
is given to be 1.49. No specific information can be obtained concerning
the refractive index of methylene blue. However, judging from its
color, it should have a high imaginary component. Figure 4-7 shows
theoretical curves of intensity ratio against particle sizes for various
imaginary components of refractive index. This figure is reproduced
after Kreikebaum and Shofner.10 The curves show that the intensity
ratio is strictly decreasing with increasing particle size for large
imaginary component of refractive index (n2 = 1.0 and 0.2). These two
curves are almost exactly the same. As the imaginary component becomes
smaller, there is an increasing departure from the above characteristic
and a flattening or dip can be observed for n2 = 0.05, 0.01 and 0.
These curves are plotted for angles of 14° and 7° so that no quantitative
comparison with the experimental results can be made. However, the
ratio 10°/5° should follow the same general trend and one can qualita-
tively see that the experimental curves follow closely the theoretical
predictions.
It would thus appear that the instrument gives much more accurate
and consistent results when sizing highly absorbing particles. From
the experimental curves, the maximum error incurred due to refractive
index difference is around 20 percent.
(3) Another important criterion in evaluating the performance of
this instrument is its resolution. Figure 4-8a and 4-8b show the actual
38
-------
co
VD
50
w 100
-O
E
3
0)
c
c
o
x:
O
150
^•200
3
O
250
300
0.5 1.0
Particle Diameter
1.5 2.0 2.5
3.0
4.0
Methylene Blue
Dioctyl - Phthalate
Figure 4-6. Calibration curves using methylene blue and DOP particles.
-------
0
\ = 0.6328
0,= 7°
02 = 14°
n = 1.6-i n,
0.7
1.4
Particle Diameter
Figure 4-7. Intensity ratio against size for different refractive indices. Following Ref. 10.
-------
Figure A-8a. Distribution for methylene blue particles of diameter 2.12 pm.
Figure 4-8b. Distribution for methylene blue particles of diameter 2.67 pm.
-------
Figure 4-8c. The two above distributions superimposed on each other,
Figure 4-9. Distribution obtained using one photomultipiier,
...
-------
distributions obtained on the pulse height analyzer for methylene blue
particles of diameter 2.12 and 2.67 urn respectively. Figure 4-8c
shows the two distributions superimposed on each other. The full width
at half maximum of these distributions is approximately 20 percent of
the maximum value. This is larger than that expected from a mono-
disperse aerosol. This broadening is not due to malfunctioning of the
processing circuit. To prove this the following experiment was performed.
The particle stream was allowed to pass through the focal volume. Instead
of using two photomultiplier inputs, one photomultiplier was disconnected
and the other photomultiplier output was connected to both inputs of the
processing circuit. In this manner the circuit detected pulses of various
amplitudes, but their ratio was constant. The distribution obtained is
shown in Figure 4-9. It is only a few channels wide and it was discovered
that variations in particle size did not change its position.
By using only one photomultiplier tube the variations due to dif-
ferent photomultipliers is eliminated and the width of the distribution
is a measure of electronic spread only. The distribution shown in
Figure 4-1 is broader than that in Figure 4-9. This indicates that
phototube variations and dark current do contribute to some variations
in ratio.
It was observed that the scattered light pulses were much bigger
in amplitude than the background noise, so that the broadening cannot
be explained on the basis of poor signal to noise ratio.
The spread can be accounted for if the following factors are
considered:
(a) Imperfections of the particle generator—For ideal operation
the particle generator should have a radioactive source in its drying
column. This source ionizes the dilution air and the ionized air in
turn neutralizes the charge on the aerosol particles. The above experi-
ments were run without this source. The main effect is to cause
deposition of the particles onto the walls of the drying column and
the outlet, but there may be other undesirable effects resulting in
a less uniform particle size. Another flaw is that the flow rate was
not very uniform. This was determined by the steadiness of the infusion
pump feed rate. During the experiment the clutch in the infusion pump
was observed to slip occasionally. This would also give rise to non-
uniformity in particle size. Therefore the standard deviation of the
particles is expected to be higher than that claimed by the manufacturer
(2-3 percent).
(b) Error in the optical collection system—The most important
cause of broadening is the imperfection of the optical system. In order
to give a constant ratio independent of position at which the particle
passes through the beam, the relative response profile for the 5° and
10° collection should be exactly the same across the entire length of
the focal volume. This was experimentally investigated. An intense,
uniform jet of particles was blown through the laser beam. The diameter
of the jet was much smaller than the length of the focal volume so that
43
-------
relatively speaking, it could be considered as a point. The scattered
pulses from the phototubes were fed directly into a pulse height analyzer.
The average value of those pulses was determined by reading the peak of
the resultant distribution. The jet was moved longitudinally across
the length of the focal volume and the above data was taken at various
points. From the data curves of the relative response of the 5° and
10° angles can be plotted. This is shown in Figure 4-10. One observes
that the positions of the peaks are slightly different. This is due
to spherical aberration which has not been fully compensated. The
important thing is that even if the peak were exactly in the same position,
the two profiles would still be different. The 10° profile has a larger
value at the peak but it falls off much faster than the 5° profile.
This means that a particle passing through the middle of the volume
will give a different ratio from the same one passing through the ends.
In the experiment a pair of parallel plates was used to direct the
particle stream. Hence the effective focal volume was actually smaller,
having a length of approximately 0.06 in (1.5 mm). With this length,
the ratio spread is estimated from Figure 4-10 to be about 1:1.3.
Obviously, this would result in a broader distribution than one expects.
The difference in the response profiles is due primarily to
difference in geometry of the 5° and the 10° collection optics. A
simplified theoretical analysis is given below.
Consider the case shown in Figure 4-11. A lens of focal length
f is illuminated by an object placed at a distance u from its plane.
The object has a finite height b which represents the finite diameter
of the laser beam. Its intensity distribution is assumed to be Gaussian.
An iris of diameter D limits the light and allows only light at a certain
angle to pass through and get focused. The finite width of the slit is
neglected. The light that comes through focuses at a distance v away
from the lens. At this point a pinhole of diameter d collects the light.
This represents the size of the optical fibre. This arrangement simulates
the optical collection system used in this experiment. The pinhole is
a two-dimensional circle, but a one-dimensional analysis with constant
intensity shows approximately the same results.
From the basic lens equation
I + I-l/f
u v
fu
v = —,. .
u-f
Taking the differential on both sides
dv = "f du (4-2)
(u-f)Z
44
-------
The magnification of the image is given by
m = J (4-3)
If the object is at exactly u from the lens plane, the intensity
distribution on the image plane will be Gaussian centered on the axis
as shown in Figure 4-lib.
If the object is at a distance u + du from the lens, the image
will be focused closer to the lens. The light will diverge out after
they focus at the point v + dv. At the position of the pinhole the
intensity distribution will be two Gaussian shapes, each displaced
from the axis by an amount 6 and having peak intensity half the original
value.
The same result is obtained if the object is at a distance u - du
from the lens. This time the image is formed further away from the
lens, and the pinhole sees the converging beam. The displacements are
assumed to be small so that the quantity 6 is the same for same change
in positive and negative directions.
The value of 6 is related to the geometry of the arrangement.
Referring to Figure 4-llc and assuming small displacements, similar
triangles give
6 dv
D/2 dv+v
= dv
v
Hence 6 = |^ dv (4-4)
At any position of the object the total power falling on the
collection pinhole is given by
,2 (4-5)
J dx
where I = peak intensity and 1 = d/2.
a' is related to the input Gaussian constant a by a' = ma.
Consider the term
,-[<*-«>/«f]2dx
45
-------
600
500
400
o
Q.
tn
DC
300
3
a.
3
O
200
100
|0°
0.2 0.3 0.4
Displacement Along Focal Volume (in.)
0.5
Figure 4-10. Relative response profiles for the 5° and 10° collection systems.
-------
Limiting
Objecj,^
T
Exactly in Focus Slightly Out of Focus
(b)
jdv I—
Figure 4-11. Simplified optical collection system for theoretical analysis.
47
-------
Putting y = -x, the integral becomes
.+1
e-[(y-6)/a']2dy
Hence the two integrals in equation 4-5 are equal. Total power
= 21
J dx
(4-6)
-1
If the diameter of the beam is taken as the —j points, the
quantity a is given by
, 2
Hence a
'
2 = b_2
2 ,2
(4-7)
The total power intercepted for various values of du can be
numerically calculated. The only difference between the 5° and 10°
arrangement is that the diameter of the iris (D) is different. Table
4-4 shows the theoretical results for the following typical values
of u, v, f, D and d.
f = 4 cm
u = 6 cm
v = 12 cm
d = 0.12 cm
48
-------
VO
-1.0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.0
Displacement,du, From Perfect Position (mm)
Figure 4-12. Theoretical plots of relative power intercepted against displacement for different
values of iris diameter.
-------
Table 4-4. Theoretical values of intercepted power against position.
du (mm) Intercepted Power (unnormalized)
D=3cm, b=0.12cm D=6cm, b=0.12cm D=3cm, b=0.085cm
-1.0
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1.0
4.907
8.143
12.072
15.905
18.907
19.976
18.907
15.985
12.072
8.143
4.907
2.641
8.143
15.985
19.976
15.985
8.143
2.641
1.194
3.296
7.260
12.755
17.878
19.908
17.878
12.755
7.260
3.296
1.194
Curves of intercepted power against du are plotted for the three
cases in Figure 4-12. The first and second correspond to the same
beam diameter but two different angles, while the third one corresponds
to a smaller beam diameter.
From the curves 1 and 2 it is seen that a difference in collection
angle results in a difference in response profile. The one with larger
angle falls off much faster. These two curves have very similar shape
to the experimental ones shown in Figure 4-10.
Case three is a case with a beam diameter /2/2 times smaller.
When compared with case one (since they have the same collection angle),
it has a smaller spread. This indicates that the diameter of the beam
has a significant effect on the length of the focal volume.
The above theoretical analysis shows that the difference in
response profile is inherent in the optical system. If the two angles
are to respond in exactly the same way, the quantity & must be the
same for a change in object distance u. From equation 4-4, since D
is not the same for the two angles, it requires that dv be different
to compensate for this. With these equations and the required conditions,
it is possible to design an optical collection system with exactly
the same profile for both angles.
The above discussion suggests that the resolution ambiguity due
mainly to the poor collection system is 10 percent or ±0.05 urn at
1.0 ym. This represents a major problem for the current optical module.
50
-------
(4) Range of Sizing—One theoretical limit for the intensity ratio
technique is that the curve of intensity ratio against size is flat
at small sizes. A practical limitation is that the scattered signal
becomes small and signal to noise ratio degrades as size is decreased.
For a 10°/5° and laser wavelength of 0.6 ym, this theoretical limit is
around 0.3 ym. The laser power of 5 mw is sufficient to give a reason-
able signal. However, because of the inherent optical spread the size
resolution goes down as particle size is decreased. No quantitative
data has been taken, but the lower limit is estimated to be around
0.6 ym.
For bigger particles the limit is set by the point at which the
scattered intensity goes out of the forward lobe. The maximum size
that can be handled is around 3.5 ym for 10°/5°. One disadvantage
of the ratio technique is that the relationship of particle size to
intensity ratio is not single valued. Error due to this becomes large
when sizing a distribution of particles which has a lot of big ones.
Methods such as an upper intensity cutoff (suggested by Gravatt)13
or the use of a third angle (suggested by Hirleman and Witting)ltf
may be used to reduce sizing error.
B) Tests on Particles of High Concentration
The Berglund-Liu particle generator can only provide particles
of relatively low concentration (~ a few hundred/cc). The scattered
light pulses observed on an oscilloscope were widely separated. It
was not possible to get quantitative data concerning the maximum
concentration that can be handled by the instrument. A source which
generates a particle stream of high concentration with a time invariant
particle size distribution is required. It would be advantageous to
be able to vary the particle stream concentration. A DOP smoke gen-
erator was used.15 This generator is a collision type atomizer.
Basically the operation of this atomizer is as follows: A jet of com-
pressed air is blasted into the liquid DOP solution. The air stream
interacts with the liquid, generating small disturbances on the liquid
surface. The turbulent jet drags out fine ligaments of liquid from
the bulk liquid stream. Finally, these ligaments break up into small
droplets because of surface tension effects.
There is no specific information on the size of the particles
generated this way. However, they are estimated to be around the range
of 1-5 ym and are quite uniform. The concentration that can be achieved
is very large.
The experimental setup consisted of applying a jet of compressed
air to the atomizer. The output aerosol was then directed to a T,
where it was allowed to mix with a large volume flow of dry air.
By adjusting the flow rate of the dilution air, the concentration of
DOP particles could be controlled. The diluted particle flow was then
coupled to a pair of parallel plates which directed the particle stream
to the focal volume, limiting the effective focal volume to be about
51
-------
1.5 mm. Along one side of the coupling there was an adjustable bypass.
This was used to divert a part of the particle stream, regulating the
flow rate through the focal volume. This prevented the stream from
flowing too fast when the dilution was high.
The experiment was started with a very large flow of dilution air
so that the concentration was low. The distribution obtained was
recorded for comparison. Then the dilution air was reduced step by
step. The output distribution was again recorded and compared with
the one obtained previously. The amount of flow diverted from the main
stream was accordingly varied to keep the output flow rate quite
constant.
When the particle concentration is large, the count rate on the
digital counter does not truly represent the total number of particles
passing through the focal volume per unit time since the more closely
overlapped ones are neglected. The output of comparator A8 is a
convenient point to monitor. The reason is that it triggers on every
peak and would count all the particles.
The count rate at output of A8 and the actual rate of data acquisi-
tion (at the digital counter) was recorded for each value of concentration.
The width of the pulses scattered from the focal volume was monitored
on the oscilloscope. During the entire process the pressure of com-
pressed air applied to the atomizer was kept constant.
It was found that the output distribution remained fairly constant
as the concentration was increased. At an input count rate of 21KHz,
the output distribution was just a little broader than the dilute case,
and the peak of the distribution was in the same position. Even at
32KHz input rate the peak stayed in approximately the same place although
the distribution was considerably broader. Further increase in concen-
tration resulted in an almost steady d.c. offset in the photomultiplier
outputs and the output count rate dropped drastically. The count rates
for higher concentrations are given in the table below:
Diameter of beam = 0.2 mm
Effective length of focal volume = 1.5 mm
Output pulse width = 100 y sec
Period on monostable A14 = 60 ysec.
Table 4-5. Experimental results for OOP smoke.
Input count rate f (at A8) Output count rate f (at A16)
9.5KHz
12.0KHz
21.0KHz
32 . 0KHz
7.0KHz
7.7KHz
8.5KHz
8 . 0KHz
52
-------
The concentration of the particle stream at different input
count rates can be calculated as follows:
The velocity of the particles through the laser beam
Diameter of beam
transit time
= 200 cm/sec.
If the laser beam has a diameter b and effective length of focal
volume is 1, in one second the particle stream would sweep out a volume
Ibv.
2
Substitute in actual numbers, volume swept out = 0.6 cm .
If the input count rate is f/sec, the average concentration of the
particles will be ,,/cm3.
u. o
The focal volume seen by the particle stream is calculated assum-
ing a circular laser beam and hence a cylindrical volume.
This is given by tr^°'"2^ x 0.15 cm3
= 4.7xlO~5cm3.
The average number of particles present in the focal volume at any
instant (N) is equal to average concentration multiplied by the focal
volume.
Corresponding to a certain N, there is a certain value of proba-
bility of 1,2—etc. particles in the focal volume according to the
Poission distribution outlined in Chapter II. If the instrument is
handling single particles only, then the fraction of particles that
gets counted would be given approximately by
+ P(2) + P(3) hP(n)
= P(D
l-P(O) '
The approximate fraction of particles counted if both single and
double pulses are handled is
P(D + P(2)
l-P(O)
Using the above formulae the theoretical values of y_p/0, and
1 +2^ can be calculated for the different concentrations. This
53
-------
can be compared with the experimental count rates. The following table
summarizes the results:
Table 4-6. Comparison of Theoretical and Experimental Count Rates.
f (KHz)
9.5
12
21
32
f '(KHz)
7
7.7
8,5
8
Average
concentration N
(particles/cm3)
1.58X101* 0.74
2xlOt+ 0.94
3.5X101* 1.65
S.SxlO4 2.51
f'/f
0.74
0.66
0.40
0.25
P(D
l-P(O)
0.66
0.61
0.39
0.22
P(D + P(2)
l-P(O)
0.88
0-91
0.70
0.50
The above table shows that the experimental fractional count rate
P (1) P(l) 4- P(2)
is always slightly bigger than .j .' but smaller than — _p , . .
This shows that the instrument is counting single pulses and also
double ones that are not closely spaced.
Figure 4-13 shows a typical distribution obtained on the pulse
height analyzer. From its position the size of the particles would be
from 1.5-2.Sum. There is no accurate way to check the size by another
method. A rough estimate was made by collecting some of the particles
on a slide and looking at them under a microscope. They appear to be
larger than the instrument reading. This is probably due to coagulation
and flattening of the oil droplets on the slide.
In summary, the above experimental results indicate that_the present
instrument can size particles with reasonable accuracy up_ to N = 1.65.
If _larger errors can be tolerated, it can operate up to N = 2.5. Even
if N = 1-65, it means an improvement of a factor of 8 in concentration
handling capability over the single particle counter.
The actual concentration that can be handled is only 5.3xloVcm3 •
This is because the focal volume in this case is quite large (4.7xlO~5cm3).
With better laser focusing and better optical collection, the focal
volume can be reduced to around 10 cm . With such volume concentra-
tions as high as 2x10 /cm can be sized by this instrument.
54
-------
.
Figure 4-13. Typical distribution for OOP smoke of high concentration.
-------
Chapter V
SUMMARY AND FUTURE WORK
A light scattering instrument for measuring size distributions of
dense particulate systems in real time has been developed. It is based
on the concept of allowing more than one particle in the focal volume
at the same time and selectively processing the scattered light pulses
by proper electronic circuitry. This selection process is entirely
random as far as particle size is concerned. Hence the size distribu-
tion is not altered.
This instrument was calibrated with methylene blue and DOP particles.
The calibration reveals that an error of ~20 percent is incurred due to
difference in refractive index. If the index of refraction of the
particle cloud is known, the intensity ratio technique may not be the
most accurate one for sizing. However, for particles with unknown
refractive indices, this method is superior compared to other optical
methods.
The theoretical limits for the size range is from 0.3 to 3.5ym using
the present configuration. Experimentally, the range was found to be 0.6
to 3.5ym due to various imperfections. As pointed out in Chapter I,
this range of sizes is of high concern in air pollution studies.
The resolution of the instrument was found to be not as good as
expected. This was due to imperfect design of the optical collection
system resulting in different response profiles for the 5° and 10°
detectors. With a better designed collection system, it is possible
to improve the resolution by a factor of two or three.
The instrument was tested with DOP smoke and its performance was
consistent with predictions by probability theory. Concentration up
to N = 2.5 has been tested and the results were satisfactory. This
represents an improvement of an order of magnitude over previous
optical sizing devices using an intensity-ratio technique. It was
shown that, with tighter focusing, say, a focal volume of 2 x 10~7 cm
used by previous investigators, concentrations as high as 10 particles/
cm3 can be sized in real-time by our technique.
The ultimate goal of this work is to develop a practical, rugged
real-time particle size monitoring system for source evaluations. The
validity of the concepts outlined in Chapter II have already been
56
-------
demonstrated by the experimental results. Hence, future work would
involve the refinement of the present system, especially the optical
module, to improve both the actual concentration handling capability
and resolution of this instrument. Since the range of particle size
is limited, efforts should also be directed towards increasing the
sizable range. In this connection, a shorter-wavelength laser with a
different choice of collection angles might be used and the possibility
of using back scattering should be investigated.
57
-------
REFERENCES
1. Belden, L. H., and Penney, D. M., "Optical Measurement of
Particle Size Distribution and Concentration." General Electric
technical information series (February 1972).
2. Apichatanon, 0., "Measuring Aerosol Size Distributions," M. S.
Thesis, Colorado State University (1974).
3. Liu, Y. H., Berglund, R. N., and Agarwal, J. K., "Experimental
Studies of Optical Particle Counters." Atmospheric Environment
Vol. 8 (1974).
4. Hodkinson, J. R., "Particle Sizing by Means of the Forward
Scattering Lobe." Applied Optics 5:5 (May 1966).
5. Hodkinson, J. R., Electromagnetic-Scattering, edited by Milton
Kerker (Pergammon Press, London, 1963).
6. Mie, G., Ann. Physik 35, 377 (1908).
7. Van De Hulst, H. S., Light Scattering by Small Particles.
(Wiley and Sons, New York, 1957).
8. Born, M., and Wolf, E., Principles of Optics. (McGraw-Hill,
New York, 1941).
9. Demnan, H. D., Heller, W., and Pangonis, W. J., Angular Scattering
Functions for Spheres. (Wayne State University Press, 1966).
10. Kreikebaum, G., and Shofner, F. M., "Design Considerations and
Field Performance for an in situ, Continuous Fine Particulate
Monitor Based on Ratio-Type Light Scattering." Presented at the
International Conference of Environmental Sensing and Assessment
(1975).
11. She, C. Y., and Chan, P. W., "Real Time Particle Sizing: Increasing
the capability of the Instantaneous Scattered Intensity Ratio
Technique." Applied Optics, Vol. 14. p. 1767 (August 1975).
12. Berglund-Liu Monodisperse Aerosol Generator Operation Manual.
13. Gravatt, C. C., "Li°;ht Scattering Methods for the Characterization
of Particulate Matter in Real Time."
14. Hirleman, E. D., Jr., and Witting, S. L. K., "In Situ Optical
Measurement of Automobile Exhaust Gas Particulate Size Distri-
butions: Regular Fuel and Methanol Mixtures." Presented at the
Symposium on Combustion, Boston (1976).
15. Yang, B. T., and Meroney, R. N., "On Diffusion from an Instantaneous
Point Source in a Neutrally Stratified Turbulent Boundary Layer
with a Laser Light Scattering Probe." Technical Report No. 20,
Office of Naval Research.
58
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
. REPORT NO.
EPA-600/2-77-022
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
A Real-Time Measuring Device for Dense Particulate
Systems
5. REPORT DATE
January 1977
6. PERFORMING ORGANIZATION CODE
.AUTHOR(S)
P.W. Chan, C.Y. She, C.W. Ho, and A. Tueton
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Colorado State University
Fort Collins, Colorado 80523
10. PROGRAM ELEMENT NO.
1AB012; ROAP 21ADL-018
11. CONTRACT/GRANT NO.
Grant R803532-01-0
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 6/75-8/76
14. SPONSORING AGENCY CODE
EPA-ORD
15.SUPPLEMENTARY NOTES EERL-RTP Project Officer for this report is W. B. Kuykendal,
919/549-8411 Ext 2557, Mail Drop 62.
16 ABST RACT
The report describes the design and performance of an instrument, based
on the concept of instantaneous intensity ratio, for measuring particle size distribu-
tions of dense particulate matter. The method involves simultaneously measuring
the intensity of light scattered by a particle at two small angles, and then taking their
ratio. The ratio depends on particle size, but has minimal dependence on refractive
index. By using a pulse height analyzer as the display device, particle size distribu-
tion changes can be detected rapidly. Thus in situ, real-time monitoring of size
distributions can be achieved. The instrument allows more than one particle in the
focal volume at any instant and selects the scattered light pulses randomly for proces-
sing, enabling dense particulate matter to be sized accurately. The concept is dis-
cussed. The instrument's detailed design features are presented. Calibration has
been performed using monodisperse aerosols of accurately known diameter. The
effect of refractive index is investigated, and the performance and limitations of the
instrument are discussed. It is shown that, by incorporating the concept of random
selection of input pulses, the concentration handling capacity is improved by an order
of magnitude.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
COS AT I Field/Group
Air Pollution
Aerosols
Dust
Measurement
Refractivity
Air Pollution Control
Stationary Sources
Particulate
Instantaneous Intensity
Ratio
Real Time
13 B
07D
11G
14B
20F
8, DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
66
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
59
------- |