EPA-650/4-74-030
JULY 1974
Environmental Monitoring Serlies
I
5
v
iu
-------
; i-^iiMiyi
-------
EPA-650/4-74-030
DEVELOPMENT
OF AIR PARTICULATE
MONITORING SYSTEMS
by
F. S. Goulding and J. M. Jaklevic
Lawrence Berkeley Laboratory
Berkeley, California 94720
Interagency Agreement No. EPA-IAG-D4-0377
ROAP No. 56AAI
Program Element No. 1A1103
EPA Project Officer: T.G.Dzubay
Chemistry and Physics Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, B.C. 20460
July 1974
-------
EPA REVIEW NOTICE
This report has been reviewed by the National Environmental Research
Center - Research Triangle Park, Office of Research and Development,
EPA, and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
This document is available to the public for sale through the National
Technical Information Service, Springfield, Virginia 22161.
11
-------
TABLE OF CONTENTS
I. INTRODUCTION 1
II. IMPROVEMENTS IN COMPUTER PROCESSING 3
III. PARTICLE SIZING 3
A. Objectives 3
B. Approach 5
C. Progress and Results 6
1. Description of ERC unit 7
2. Loss measurement method 7
3. Experimental procedure in loss measurement ... 8
4. Results on ERC impactor 10
5. Summary of evaluation of ERC unit 14
6. Work on new dichotomous sampler 15
IV. PULSED X-RAY TUBE 18
A. Objectives 18
B. Tube Design 20
1. Cathode-grid assembly 20
2. Electron optics 21
3. Switching characteristics 23
C. Electronic Design 24
1. Grid pulse circuit 24
2. X-ray control loop 25
D. Mechanical Design 27
111
-------
E. Results 28
F. Conclusion 32
V. DIGITAL FILTER MARKING 33
VI. TOTAL PARTICULATE MEASUREMENT 34
References and Footnotes 36
Figure Captions 37
Appendix A 58
iv
-------
-1-
I. INTRODUCTION
This report describes progress achieved during the calender year 1973
on work carried out under an Interagency Agreement between the Environ-
mental Protection Agency and the Atomic Energy Commission. The program
concerns several aspects of air particulate analysis using energy-disper-
sive X-ray fluorescence elemental analysis techniques. The 1973 project
is a direct continuation of the 1972 program during which a particulate col-
lection and analysis system was developed and delivered to the EPA labora-
tories at Research Triangle Park, North Carolina. This system provides for
automatic sample collection, and also for automatic elemental analysis using
an X-ray fluorescence analyzer under the control of a Texas Instrument 960A
computer. The 1973 program has been directed toward a number of improve-
ments and additions to the monitoring system with particular emphasis on
developing the concepts, hardware and computer programs required for the
large-scale St. Louis RAPS program to be carried out toward the end of 1974.
Aspects requiring particular attention included:
i) Improvements in the X-ray spectral analysis program to facilitate
its use and to improve the accuracy of results. This has required continu-
ous contact with EPA personnel using the system at Research Triangle Park,
and the changes made, reflect experience in use of this system.
ii) The importance of separating small particles from large ones, both
for medical reasons and as an aid in pinpointing origins of various ele-
ments, has led to the requirement for collecting two parallel samples
-------
-2-
per collecting station, each representing a different size range (less
and greater than 2 micron diameter). To satisfy this requirement a
dichotomous sampler has been developed which deposits particles from the
two size ranges on separate filters which may then be analyzed by the
X-ray system.
iii) In order to speed up the X-ray analysis process, work has been
carried out on a pulsed X-ray system1 which should reduce the analysis
time by approximately a factor of four compared with the earlier EPA
system.
iv) The handling of very large numbers of samples will be facilitated
by digital marking of all filters prior to use and subsequently reading of
this marking by the computer during elemental analysis. Various methods
of marking have been studied and one has been chosen for use in the St.
Louis RAPS program.
v) Since total particulate mass measurements are required in the
St. Louis RAPS program, a strategy has been developed for using a beta-
gauging technique on the same filters as used for elemental analysis. The
digital marking facilitates keeping records of the filter mass before and
after exposure.
The following detailed discussion will be broken down into these
five sections.
-------
-3-
II. IMPROVEMENTS IN COMPUTER PROCESSING
A number of changes in the hardware and software associated with
the TI-960 computer were implemented during the contract period. These
improvements were concerned both with expanding the data handling capabil-
ities and increasing the convenience and reliability of the existing sys-
tem. These changes are detailed in Appendix A.
III. PARTICLE SIZING
A. Objectives
Interest in size separation of atmosphere particles results from
the desire to obtain size vs. elemental concentration correlations as part
of a complete aerosol monitoring and characterization program. The size
distribution of suspended particulates and the related elemental distribu-
tions are important both in determining the origin and history of the
atmospheric aerosol, and in the assessment of respiratory health effects.
Most existing particle size separators depend upon the impaction of
particles on surfaces placed in a rapidly moving air stream. The change
in direction as the air stream is deflected by the surface causes parti-
cles above a certain critical size to be impacted onto it. By cascading
a number of these impactor stages with successively smaller critical size
2, 3
cut-offs, a series of size fractions can be collected.
-------
-4-
Existing cascade impactors have certain disadvantages from the point
of veiw of compatibility with the X-ray fluorescence analysis measure-
ments. One problem is particle blow-off resulting from the rapidly moving
air stream; as the particle deposition on the substrate increases, par-
ticles tend to be blown off the impaction surface and deposited on succeed-
ing stages. Another problem is the finite probability for a particle to
bounce off the surface on impact. Remedies, such as coating the substrate
with a sticky substance, are only partially successful in preventing
bounce-off and might interfere with subsequent elemental analysis. The
particle deposits in these impactors are necessarily in the form of an
image or images of a small slit or circular holes which are not ideal for
subsequent elemental analysis of the samples by X-ray fluorescence.
The objective of the present program is to design an air sampling
system which can automatically acquire size-fractionated air particulate
samples in a form compatible with X-ray fluorescence analysis, while at
the same time avoiding particle bounce-off difficulties. Initial specifica-
tions call for the ability to separate particles into two size ranges cor-
responding to particles above and below a cut-off at 2 micron particle
diameter. This represents the minimum in the distribution of total volume
vs. particle size curve observed in the so called bimodal distribution of
lj
urban atmospheric aerosols. The design will permit later extension to
three or more size ranges. Measurements performed on the prototype size
fractionator include collection efficiency and loss as a function of
particle size.
-------
-5-
The complete sampling system will be capable of automatic acquisition
of a series of samples with adjustable sampling periods of 1 to 24 hours.
The calibration and reliability of the prototype device will be suitably
tested. Since it is likely that a number of duplicate systems will be
required for the St. Louis RAPS program, extensive documentation will be
required to allow large scale production of the final model.
B. Approach
A brief survey of particle sizing methods was necessarily a preface
to the design phases of the program. Various mechanisms for the physical
separation of particles according to size were studied and their advan-
tages and limitations outlined. Possibilities included electrostatic,
thermal, acoustical and optical methods in addition to the aerodynamic
methods discussed earlier. These studies eventually related to the more
fundamental problem of the definition of particle size. Since, in gen-
eral, the shape, density, dielectric properties, etc. of the particles in
the atmospheric aerosol are unpredictible, the interpretation of the term
"particle size" depends on a large degree to the method of measurement.
However, in any foreseeable atmospheric aerosol studies, the questions of
most interest concern effects such as the mobility of the particulates in
the atmosphere and their uptake and retention through respiration. In
these contexts thejnost relevant definition of particle size relates to
the aerodynamic size of the particle which governs its motion in these
-------
-6-
situations. These considerations lead to the conclusion that separation
methods based on aerodynamic effects most nearly measure the important
parameters. Present methods of particle size measurement and the associ-
ated terminology reflect this bias. Since none of the alternative methods
of particle sizing exhibits unique advantages over the aerodynamic method,
we conclude that the adaptation of the latter would be appropriate.
C. Progress and Results
At the start of this program it was obvious that use of a normal
impactor, in which the heavier particles impact and collect on a surface,
cannot conveniently provide samples suitable for X-ray fluorescence
analysis on large numbers of samples. Consequently emphasis was imme-
diately placed on a "virtual" impactor scheme which results in separating
the particles into two airstreams according to their size. Each air-
stream could then be passed through a filter to provide particulate sam-
ples in the ideal form for X-ray analysis. Since Environmental Research
Corporation (ERG) had supplied a particle-size fractionator to EPA
based on these general principles, the evaluation of this unit, described
in the following sections, was carried out prior to developing a better
unit. The study was designed to determine the separation characteristics
and losses of the unit as a function of particle size and to relate these
characteristics to critical design parameters. This information could
then be used to assess the practicability of large-scale use of this
principle in air samplers.
-------
-7-
1. Description of ERG unit
Figure 1 is a phtotgraph o£ a disassembled ERG virtual impactor show-
ing the various components, and Fig. 2 is a simplified assembly drawing.
The aerosol is drawn through the three entrance holes in the part
labeled B (in Fig. 2) at a total rate of 50 Jl/min. The diameter of the
holes in parts B and F are 3.96 mm and 2.18 mm while the "impaction dis-
tance" to jet diameter ratios are 0.45 and 0.75 respectively for the two
separating stages. The corresponding Reynolds numbers for the jets are
5900 and 1500. The combination of the hole in part F and those around the
flange on part E distributes the internal flow such that only one seventh
of the total flow is allowed through the inner tube E. Particles which
cannot negotiate the turns around the lips of the protruding tubes in
part D will be carried down into the inner tube. The process is repeated
in a similar fashion at the lower section of the apparatus where the flow
through tube K is maintained at 1 £/min. Thus filter A will collect
large particles above the 2 micron size cut together with about 21 of
the small particles. Most of the small particles will be carried by the
main flow at 49 &/min and be collected on filter B.
2. Loss measurement method
Due to the close approximation of their geometric diameters to unit
density Stoke's diameters, dioctyl phthalate (DOP) droplets, having a
specific gravity of 0.98, are well suited to serve as test particles.
They are generated by a Berglund-Liu monodisperse aerosol generator.
Uranine (fluorescein sodium) is used as a tracer for quantitative measure-
ments, particle losses at various points within the impactor being
-------
-8-
determined by washing off the aerosol deposition from a particular
region of interest. The uranine content of the resulting solution is
then measured by UV fluorescence techniques.
Figure 3 is a schematic of the UV fluorescence analyzer. A 76-mm
diameter pyrex dish holds the sample solution. The central portion of
the solution is viewed by a photomultiplier via a mirror so that edge
effects and UV attenuation in glasswares are avoided. The signal-to-
background ratio is greatly enhanced by the use of complementary filters,
one at the mercury light source with a transmission band from 2500A to
4000 A and the other (green XI) in front of the photomultiplier. This
scheme also renders measurements practically immune to UV scattering by
dust or DOP itself. The filter combination is not optimized for maximun
sensitivity but rather it yields an adequate and convenient operational
level. The stability of the analyzer is monitored with the aid of fiber
optics in order to maintain a fixed source-detector transfer function
which is checked before and after each measurement by means of a sliding
shutter.
3. Experimental procedure in loss measurement
The virtual impactor must be operated vertically since gravitational
effects are not negligible for large particles in regions of low flow
rate. A carbon vane pump (Cast #0522-103-G18D) operating at 1725 rpm is
used. The total flow rate of 50 Jl/min is monitored with a manometer which
is calibrated with a temperature-compensated integrating flow meter. The
rate through filter A is maintained at 1 H/min by an inline flow meter.
-------
-9-
DOP particles o£ 1 to 10 micron diameters are produced by a Berglund-
Liu monodisperse aerosol generator in which a liquid jet is produced by
forcing a solution of the particulate matter in a volatile solvent tnrough
an orifice with a syringe pump. A periodic disturbance is imparted to the
liquid jet by vibrating the orifice with a piezoelectric ceramic. Operat-
ing in the proper region of instability of a liquid jet, the eventual size
of the particle depends on four influencing factors:
i) The feed rate of the syringe pump.
ii) The vibrational frequency of the orifice.
iii) The concentration of non-volatile solute.
iv) Any voids occurring in the residual particle when the solvent
in the original droplet evaporates.
For consistant operation, a single calibrated feed rate of 1.3307 x
10"3 ml/s and a vibration frequency of 144.60 KHz are used to generate
initial droplets of 26 microns with a 10 micron orifice. The solutions
used consist of one part by volume of demineralized water to 14 parts of
isopropryl alcohol with 50 yg/ml of uranine and selected amounts of DOP
according to the final particle size desired. The droplets are dispers-
ed and dried during mixing with a dry airstream. They are charge neu-
tralized by exposure to a 10 me Krypton-85 source. Uniformity of the
aerosol is checked by visual observation of the initial droplet stream
as deflected by a transverse air jet. With the aid of a scanning elec-
tron microscope, samples of individual particles are also examined occas-
ionally for size and void fraction estimations.
Since the sizes of small particles are significantly affected by
the amount of impurities present in the solution, the syringes are
-------
-10-
regularly ultrasonically cleaned with detergents followed by water and
alcohol rinses. Impurities in the DOP solution are thereby limited to
about 10 yg/ml of alcohol residue and 50 yg/ml of uranine.
For each test run, the loss profile is determined from washings
of depositions from six designated regions: (1) parts B and C, (2) part
D, (3) the inside wall of tube E, (4) parts F and G, (5) part H, and
(6) part K. A very slight loss also is known to occur in connecting tube
to filter B. Uranine, being very soluble in water, is retrieved readily
from the depositions on 0.8 micron Millipore filters. For DOP droplets
greater than 5 microns, a few seconds of ultrasonic agitation will facil-
itate complete retrieval.
4. Results on ERG impactor
A number of calibration and monitoring functions were checked to
minimize instrumental errors. The average air sampling rate as determined
from total integrated flow for all the test runs is 49.75 £/min with an
rms deviation of 1.2%. Note that this represents the stability of the
sampling rate with no feedback control, following initial adjustment to
allow for different filters. The average feed rate of the syringe pump
is 1.3307 x 10"3 ml/s with an rms deviation of 0.3% for different syringes.
The orifice vibration frequency is continuously monitored and exhibits
less than 0.1% deviation from its mean value of 144.60 KHz. With careful
preparation of solutions used for generating aerosols, the overall error
on the particle size is probably 1% in all cases except for the rare case
where the impurities become a dominating factor.
-------
-11-
Since the absorption and emission bands of fluorescein sodium
partially overlap each other, the effect of self-absorption becomes
significant and must be corrected for when using concentrated solu-
tions. Figure 4 shows a plot of detector output current vs. uranine con-
centration obtained with seven dilutions of a conmon sample. A least-
squares fit to the calibration data gives
I = 0.4723 (1-exp (-42678x))
where I is the photomultiplier anode current in microampere and x is the
uranine concentration in g/cm2. A constant background of 5.6 x 10"10
amp is subtracted from each measurement. This yields an effective attanua-
tion length of 58.43 yg/cm2 defined as the concentration at which 63% of
the radiations is absorbed. The detection limit at 95% confidence level
is 1 ng/cm2, corresponding to approximately 10 ng in the sample. A typical
test run collects about 10 micrograms of the tracer; at such level self-
absorption has no significant effect.
Table 1 is a summary of the percent deposition of DOP on various
parts of the sampler and filters for incident particle sizes between 1 to
10 microns. The particle sizes are expressed in terms of unit density
Stoke*s diameters. They include corrections for 50 yg/ml of uranine and
10 yg/ml of solvent residue which are taken to have a bulk density of
1.7 g/ml, and also for an estimated void fraction of 48% as determined by scan-
ning electron microscope observations. The 1.06 micron particles consist
of mainly uranine with no DOP.
-------
-12-
PC!
o
EH
O
EH
M
O
O
O
M
EH
Hi
CO
EH
Hi
Hi
P4
co
1
FH
CO
H
P
O
|
f=-t
O
EH
M
co
o
PH
P
EH
O
PH
o
o^~^
a
0 3
i -
VD a
co*
ola
* -j_
CO
ola
3
o'a
^s3
o^a
to
LA- v
LA a
3.
CO'--.
0 6
"v-3
to^
VD S
3-
VD^-~.
CM a
3-
T~ ^-^
VO--^
o a
3-
V
g
M PC;
CO COW
S - EH
EH EHH
H COP
VD
^-
VD
CM
O
CM
CO
O
CM
O
^
0
CM
CT>
CO
T-
CM
0
T
**
c-
*=i-
T
^
O
M
cb
CO
o
^
CD
O
*
O
CM
«-
0>
,-!
CM
to
LA
v~
CM
to
to
*
C\'
l
0
O
CM
&
O
M
s
K
CM
T
VD
O
t-
O
CM
*
O
_
0
^_
o
CM
0
0
CM
«
O
T-
O
0
to
fe
0
M
s
rt
LA
«
to
^
CM
T~
CM
V
to
^
T~
^
0
o>
0
*
o
o
9
CM
en
«
CM
CO
t
^J-
&
o
M
pc!
^j-
*
CM
fO
O
CO*
LA
CM
Q
O
o
T_
o
CO
*
o
en
CO*
o
T
^^
sj-
0
,
o
LA
|25
o
M
C±>
§
f^
m
T
VD
T-*
T~
CM
^
^~
LA
O
=*
o
LA
O
CM
»
O
CM
»
0
c\
»
c
*
o
VD
fe
O
M
CJ3
w
P5
CM
*j-
"^
VD
LA
^
CT\
CM
,
vo
00
t^
VD
*
CM
cn
«
VD
CM
t
LA
CM
en
-d-
CM
en
^
CTi
VO
co
co
o
l_q
EH
O
EH
to
t
LA
LA
O
,-£
CO
CM
0
LA
*
fO\
f-
LA
_
₯^
f-
LA
9
CT\
^
to
VD
CM
LA
,
r
"*"
r-
^
CM
^j-
-P
O
rt
EH 0
Hi £>
M
CM
A
CO
"^
LA
to
VD
CM
[
CO
,
CO
co
-p
o
W 0
EH H
h^ CO
M pq
Hi
9
EH
-------
-13-
Figure 5 is a graphical illustration of the overall results.
The structure in the loss spectrum is derived from the composite of its
various smoothed components. The loss spectrum due to a single contrib-
uting mechanism is expected to have a smooth variation with particle size.
Figures 6 through 11 are the plots for the six loss regions. It
is observed that for region 1, the peak at small particle sizes is due
to the ring shaped deposition around the holes on the underside of part B,
as can be seen in Fig. 1. Such depositions are believed to be the results
of direct impaction from the radial spatial oscillation of the stream-
lines rather than from eddy depositions. The oscillation is characteris-
tic of the local geometry and flow rate. Similar phenomena have been
observed in rectangular impactors where a succession of line deposits
appear about an impaction slit.5 We note that one of the three support-
ing legs on part D in Fig. 1 is narrower than the others. The presence
of these legs in the most direct line of flow has a definite bearing on
the balance of flow rate in the individual channels. Consequently, there
is usually no ring deposit around the hole opposite to the narrow support-
ing leg, but a higher loss is found on the edge of the corresponding
opening in part C. The losses at larger particle sizes are from the
depositions on the upper lips of the entrance holes.
Depositions in region 2 are found mostly on the lips of the pro-
truding tubes of part D. The impaction depositions show distinct azimuthal
variations, reflecting the sensitivity of flow patterns to structural
asymmetries. Part C is used to reduce the eddy losses around the tubes
but some deposition is found on its circular openings.
-------
-14-
For region 3, there is evidence that the expanding jets touch upon
the adjacent walls on part E. The slightly higher loss toward larger
particle sizes probably reflect the higher penetrating power of the more
#
massive particles through the boundary layers near the walls.
Losses in region 4 have the same nature as those in region 1. The
shifting of the loss peak to a smaller size indicates that the characteris-
tic size cut of the second stage of separation is somewhat lower than that
of the first stage. The data point at 6 ym is believed to be an anomaly;
this is also obvious from its deviation from the expected smooth collection
curve for filter A.
The loss spectrum for region 5 should behave similarly to region 2.
Besides the slight shift of the loss peak to the smaller size, which is
consistant with the parallel comparison of regions 1 and 4, there is an
overwhelmingly high turbulance loss component at large particle sizes when
the flow through filter A is restricted to 1 £/min.
There are no distinct patterns of deposition in region 6; most of
the losses probably occur in the slight mismatch to the filter holder.
5. Summary of evaluation of ERG unit
It is obvious that the virtual impactor can successfully separate
two size ranges of particles and collect them on filters.
High losses near the cut region are probably intrinsic to any vir-
tual impaction scheme since a deflecting surface will behave as an effic-
ient impacting surface at the critical particle size. Furthermore, the
boundary condition which occurs at the rigid surface of a real impactor
-------
-15-
cannot be fully realized by a collection hole. The Gaussian-like pressure
profile across the impinging jet cannot be balanced by a uniform static
pressure in the collection volume. Streamlines will therefore penetrate
into the hole and be accompanied by a counter flow near the edges when the
net flow into the (large) particle collection volume is restrained. We
suspect excessive turbulance will be developed in a very confined region
where such counter flow cannot be maintained. The development of such
turbulance is probably responsible for the high losses at large particle
sizes. On the other hand, a significant amount of loss is found to depend
on details rather than on basic principles of a virtual impactor and these
can possibly be eliminated by suitable attention to design and fabrication.
6. Work on new dichotomous sampler
These tests on the ERG unit show that the virtual impaction scheme
is workable for a dichotomous aerosol sampler, assuming that a peak in
the particle loss spectrum near the cut point is tolerable. We have pin-
pointed the sources of losses in the ERG model and expect to be able to
minimize losses in the new design for the St. Louis RAPS study.
We have decided against the alternative sampling scheme of using a
total filter together with an impactor after-filter for the following
reasons: (1) it is much more complicated to provide twice the pumping
capacity and to match the flow rates in the filter pairs, (2) the bounce-
off and blow-off problems associated with impactors would require the
added complexity of sticky and possibly rotating impaction surfaces, thus
making the servicing of the instrument more difficult, (3) the virtual
-------
-16-
impactor has two intrinsic advantages: namely, losses are reduced whenever
bounce-off occurs and any blow-off phenomenon is of a second order effect
in that only loss build ups are subjected to blow-off.
Based on the results on the ERG unit, we have designed and tested an
improved version of the dichotomous sampler for use in the automatic air
sampling station being designed for the St. Louis RAPS study. Improvements
incorporated into the new design are as follows:
a) The sampler is designed for easy dismantlement; all sliding seals
have been eliminated or replaced by compression seals.
b) Stainless steel construction is employed for mechanical integrity.
c) Inlet holes are streamlined to eliminate large particle losses
in region 1. At 10 micron particle diameter, the losses in this region were
reduced from 4.6% to only 0.1%.
d) Ring deposits around the back side of the inlet holes have been
eliminated by recessing the surface where normal depositions occur. Losses
have been reduced from 11.8% to only 0.1% for 2 micron particles.
e) The square lips of the protruding tubes are rounded to somewhat
reduce the impaction losses on the lips.
f) The geometry in region 2 has been redesigned to minimize asym-
metries in the flow patterns.
g) The internal flow division at region 2 has been measured to be
6.14:1. This ratio may still need to be optimized.
h) Region 3 is designed to have a conical flow section to minimize
large particle settling at low flow rates. The inlet hole in region 4 is
also streamlined. Losses at 10 micron are thus reduced from 3.5% to only
0.1%.
-------
-17-
i) The ring deposit on the back side of the hole in region 4 is
eliminated by recessing the surface. This means the elimination of the
2.9% loss peak at small particle size.
j) The size and shape of the hole in region 5 is optimized to re-
duce turbulent deposition of large particles which result in a loss reduc-
tion from 32.4% to 6.2% at 10 microns.
Preliminary measures indicate that total loss has been reduced from
42.5% to 11.0% at 10 microns, and a reduction of about 1/3 has been
achieved in the small particle loss peak. A preliminary test of blow-off
effect has been done on the new virtual impactor. We introduced 2 micron
uranine particles into the airstream after the unit had sampled 234 m3 of
room air. By measuring any uranine collected on clean filters from sub-
sequent pumping, we found the upper limit for blow-off to be 10"1*. We
shall perform more thorough measurements when we finish all the detail
changes on the prototype.
Further studies of the basic design of the virtual impactor are pro-
ceeding in parallel with design and construction of a prototype air-sampler,
The final design of the virtual impactor will then be incorporated into
the construction of the twelve automatic dichotomous air sampling stations
required for the St. Louis RAPS program.
-------
-18-
IV. PULSED X-RAY TUBE
A. Objectives
Previous investigations have indicated that substantial improve-
ments in system performance can be achieved by employing a pulsed excita-
tion source operated synchronously with the X-ray spectrometer. By puls-
ing the X-ray tube on only at those times when the spectrometer is available
for analysis of detected X-rays, pile-up effects are eliminated and a
considerable increase in effective counting rate is achieved. Figure 12
is a schematic of the earlier pulsed system showing the operation of the
prototype tube. A discriminator operating on the output of the X-ray
detector triggers a grid pulse which shuts the X-ray source off for the
duration of the analysis cycle; associated electronics turn the tube back
on immediately after the preceeding event has been stored. In addition
to eliminating dead-time and pile-up problems in the associated electronics,
the method reduces the total average power requirements of the X-ray tube
since the tube is operated only when the spectrometer system is sensitive.
No current flows into the anode during the pulse processing time so instan-
taneous power levels many times higher than the average anode dissipation
can be used when the tube is "ON".
Figure 13 is a plot of input-output counting rate characteristics,
comparing the conventional excitation methods to the pulsed tube. Earlier
data have indicated that effective counting rates of up to five times
greater can be achieved by the pulsed method compared with conventional
operation.
-------
-19-
Although the prototype system was adequate for demonstration of
the basic method, the X-ray tube used in these preliminary experiments
was unsuitable for extensive X-ray fluorescence analysis mainly due to
inadequate electron emission from the directly heated tungsten emitter.
The problem is further compounded when operating with a secondary fluores-
cence geometry such as is used in recent systems. A new design and
development study is being carried out under the current contract to make
the pulsed tube a practical approach for large-scale fluorescence analysis.
The X-ray fluorescence system delivered to the EPA as a result of
last year's work exhibited typical detector counting rates of 5 to 10 Kc/s
with its X-ray tube operated at 40 watts anode dissipation. This produced
output rates of processed (i.e. free of pile-up) signals of only 4 to 6 Kc/s
as a result of pile-up rejection. Using the same system in the pulsed mode
would allow counting rates of 24 Kc/s of processed events with no pile-up
losses, with the average X-ray tube power level below 100 watts.
Development of a new pulsed X-ray tube is being undertaken with the
restriction that resulting design should be compatible with operation in
the earlier EPA system. This boundary condition has a large effect on the
design of the tube geometry and also affects the associated electronics.
In particular, the electron optics must be designed to project a >100 mA
beam of 30 to 80 keV electrons, into a 6 cm long 1/2 cm diameter anode
tube to permit use of the close-coupled X-ray geometry as in the existing
instrument. This beam must be turned "OFF1 and "ON" with a timing uncer-
tainty of less than 200 ns for full advantage of the pulsed mode to be
realized. Since the close-coupled geometry necessitates a grounded-anode
-------
-20-
tube design, pulsing must be performed at the negative high-voltage elec-
tron source, and must preferably be accomplished without fast switching
of high voltages due to the proximity of the low-noise X-ray spectrometer
electronics.
B. Tube Design
Although a number of methods of pulsing X-ray tubes have been con-
sidered, the decision has been reached to design the tube in the form of
a planar triode structure with a thermionic cathode, low-voltage control
grid, and high-voltage X-ray anode. The planar electrode design permits
the generation of a uniform electron beam which can then be projected
down the anode tube. One of the more important considerations is to avoid
space charge limitations in the amount of current which can. be transported
in a narrow beam. In the following sections the design of the X-ray tube
will be discussed in detail.
1. Cathode-grid assembly
The design of the high-current pulsed X-ray tube has been facilitated
by the commercial availability of cathode-grid structures produced for
microwave triodes. These large-area planar cathodes (typically 0.5 to
0.8 cm2) are capable of peak emission current densities of 20 amps/cm2.
The grid-cathode spacings are 0.15 mm or less, allowing high-current switch-
ing with moderate grid control voltages. We have evaluated units from two
manufacturers--one using an 0.8 cm2 oxide-coated cathode, the other with a
-------
-21-
0.5 cm2 dispenser type cathode. Both cathodes need to be activated after
pump-down of the tube and are "poisoned" by any subsequent exposure to
atmosphere. The dispenser cathode can be reactivated, and has the addi-
tional advantage that it is less susceptible to damage from positive ion
bombardment--an important consideration in X-ray tubes.
Although the difficulties attendant to using activated cathodes are
serious, the alternative method of a directly-heated tungsten filament is
out of the question due to power considerations. The activated cathodes
operate at emission current densities of 10 amp/cm2 with a filament power
input of less than 10 watts; by contrast this emission level would require
a power of several hundred watts in tungsten emitters. Based upon our
earlier experience, the fact that the cathode is operated at a high nega-
tive voltage relative to the grounded anode imposes a practical limit of
approximately 10 watts heater dissipation.
2. Electron-optics
Having defined the grid-cathode structure of the tube, we next con-
sider the electron beam requirements. The principal question is whether
adequate current can be projected within the beam dimensions in spite of
space-charge effects. Referring to Fig. 14 we can consider the design
problem in the following steps:
For a tube length of 10 cm from anode aperture to the X-ray
target, and a maximum beam envelope radius, space-charge effects
establish a maximum current which can be projected through this
region at a given voltage.6 This relationship can be expressed as
I = PV
3/2
-------
-22-
where V is the potential of the beam relative to the tube
and P is the perveance. Assuming a beam radius at entry
of 0.25 on and a tube length of 10 cm, then PSIO"7 if
V is in volts and I in amps. Thus the maximum currents which
can be passed through the tube are:
0.48 amps at 30 KV
1.06 amps at 50 KV
6.82 amps at 80 KV
Since a peak current of 200 mA is sufficient for our
application, these parameters are quite adequate.
The boundary conditions for maximum current flow
through the tube define the focal length of the anode-
cathode lens structure. The necessary focusing, cor-
recting for the anode aperture lens effect, can be
achieved by an arrangement of two concentric spheres
of radius 2 on and 4 cm respectively. The beam enve-
lope is similar to the one illustrated in Fig. 14, and
remains more or less constant over a wide range of cur-
rent.. The perveance of the focussing part of the lens
structure is also approximately 10"7 so no space-charge
limit to the current is imposed at this point. These
electron optics appear to be capable of maintaining
steady-state currents of 500 mA, or greater, over the
entire operating range of the tube although in prac-
tice anode dissipation provides the limit to continuous
operation.
-------
-23-
3. Switching characteristics
Since the interelectrode distances, grid wire spacing, and cathode
characteristics are known, it is possible to estimate the triode characteris-
tics of the tube. The most important parameter initially is the amplifica-
tion factor y, defined as (VANf)DE/VGRID^ at tJie cut"off point. A low value
of y implies that a high negative grid voltage is necessary to cut off the
anode current, necessitating the introduction of additional electrodes to
reduce the field gradient at the grid. Calculations of y appropriate for
the triode structure of Fig. 14 yield an approximate value of y = 3000
corresponding to negative grid cut-off voltages of 12, 15 and 24 volts for
plate voltages of 30, 50 and 80 KV respectively.
Calculations of plate current vs. grid voltage have been made for
each of the above plate voltages assuming y = 3000. Results show that beam
currents exceeding the space charge limits of the anode tube can be achieved
and that these currents can be controlled at grid voltages well within the
capabilities of transistor circuits. The switching speed of the tube is
limited by the grid capacitance; electron transit times through the tube
are of the order of a few nanoseconds. Assuming a grid circuit capacitance
of 20 pF (the measured grid capacitance of the electron gun is 8 pF), then
a pulse rise of 100 volts in 100 ns requires a current switching capability
in the pulse driver of
I = CV = (20X1Q-12) (IP2) = 20mA
At 10-7
which is well within the operating characteristics of typical transistors
that might be used.
-------
-24-
C. Electronic Design
The design of the associated electronics must contend with an assort-
ment of problems not previously encountered in X-ray tube operation. For
convenience we will consider the grid-pulse circuit separately from the
X-ray controller. Although the X-ray tube characteristics outlined above
appear to make the design of the grid circuit fairly straightforward, the
problem is complicated by the necessity of maintaining the entire pulser
circuit at the high negative cathode potential. The switching-speed require-
ment precludes the use of the high-voltage X-ray cable as a signal transmit-
ting device since its capacity is excessive. Therefore it is necessary to
attach the grid-drive circuits to the tube at its high voltage terminal.
1. Grid pulse circuit
Figure 15 shows such an arrangement whereby the dc power supply for
the grid-drive circuit operates from the 6.3 V filament supply which is
floating at voltages ranging from 0 to 100 KV dc. Control signals for the
grid-circuit are optically coupled across the high-potential difference
using LED and photodiode pairs. The components contained within the dashed
region are all operated at the negative cathode potential. Signals are
provided via the optical couplings to the control grid bias level, which
determines the operating point according to the curves of Fig.16, and also
an on/off control to determine the timing and duration of the beam current
cut-off. Beam-current level in this system is determined by the grid bias
instead of by the filament power control as in the earlier EPA system.
The present design calls for the grid bias control amplitude to be transmitted
-------
-25-
as a pulse duration rather than a variable voltage since the latter approach
is sensitive to non-linearities in the optical isolation system. The on/off
control signal can be utilized directly.
2. X-ray tube control loop
The full advantages of the pulsed mode of operation are only realized
when the associated control circuitry maintains an optimum operating point
for the system. The various operating conditions can best be discussed by
considering the behavior of the system as the sample mass is increased,
resulting in a corresponding increase of counting rate per unit of X-ray
tube power. When the sample mass is small it is not possible to maintain
high counting rates and the tube is operated at its maximum average power
dissipation. For reasons to be discussed later, this is regulated by limit-
ing the duty cycle of the tube and operating at maximum emission current
during the ON period. This is effectively the same as operating continu-
ously at maximum power since the probability of detecting one event per
cycle is assumed small. In this first region of operation the power level
is fixed by the maximum pulse beam current and maximum duty factor, and
the counting rate varies linearly with sample mass. As the sample mass
increases, the probability of an event during the ON period is increased
to the point where a sizeable reduction in the average duty cycle would
normally occur. However, this would decrease the average power dissipa-
tion so, in this second region, the instantaneous tube current can be in-
creased to maintain the maximum allowable average power; the current
increase compensates for the reduction in duty factor due to higher count
rates. The magnitude of this effect increases with increasing sample
-------
-26-
mass until a point is reached where the average ON time becomes comparable
to the time required by the pulse circuitry to shut off the tube. The
probability of additional pile-up events occurring during this time inter-
val then tends to become significant. This constitutes the maximum count-
ing rate limit for the system, further increases in sample mass in this
region must be accompanied by a reduction in instantaneous X-ray tube power
so that the pile-up probability in the tube shut off time does not become
excessive.
The block diagram of Fig. 15 illustrates how these control features
will be accomplished. The maximum ON time is determined by a preset one-
shot which triggers a dead time interval when the preset ON time has been
exceeded. This dead time interval is chosen according to the amplifier
shaping times as defined in Fig. 13. The end of the dead time waveform
then triggers the next ON period. This is the mode of operation at very
low counting rates. As the counting rate is increased, the ON time one-
shot is reset more frequently by pulses detected by the X-ray system. Thus
the average ON time is reduced. This is the second region of operation
during which the auxiliary feedback loop increases the tube current to
maintain its average power constant. Meanwhile, a separate circuit is con-
tinuously monitoring the average duty factor and the desired minimum duty
cycle is reached, this circuit causes an appropriate reduction in the pulse
beam current as discussed for operation in the third region. Characteris-
tic times for a typical system might be:
Dead time 50 ys
Maximum ON time 10 ys
Minimum ON time 2 ys
-------
-27-
Since this feedback arrangement causes the pulsed tube system to
determine its own operating point, normalizing data from one sample to
the next is a different problem from that in the previous (conventional)
system. The pulsed method of excitation used in this manner reduces
dead time losses essentially to zero. Normalization of the counts can
therefore be accomplished by feeding information on the integrated beam
current during a count into the computer where the program performs the
normalizing process.
D. Mechanical Design
The exterior design of the basic tube will be similar to the pre-
vious EPA tube design although its size is increased due to the larger
cathode-grid structure used in the pulsed tube. Recent improvements in
technique to be incorporated in the design include a ceramic high-voltage
insulator and an electron-beam welded anode.
The grid-drive circuits will be contained in an oil-filled chamber
attached to the rear of the tube. Although high-voltage insulation
requirements will dictate a length of 25 cm or more for this section--the
penalty of increased size is compensated by the improved cooling capabili-
ties and much higher emission currents than in earlier designs. Optical
isolation of the control signals will be accomplished via light pipes
immersed in the oil bath.
-------
-28-
E. Results
A prototype tube and associated electronic controls have been con-
structed and extensively tested with encouraging results. The basic mechan-
ical design of the tube enclosure and desired lens structure was implemented
with minimal difficulty. The tube is designed to be completely baked at
325°C with a 2 1/s ion pump attached for subsequent pumping of gases released
by materials.
The electron gun used in the first attempt was of the dispenser cathode
type with 0.5 cm2 active area. (Eimac Y646B). The cathode was successfully
activated and subsequent operating experience has not required reopening the
tube. No observable deterioration in emission properties has been observed
after approximately 100 hours of operation, mostly at 40 to 50 kV applied
anode potential. Observations of the electron beam structure using X-ray
imaging techniques indicate that the optics perform almost to the design
specifications. The beam spot was approximately 5 mm diameter and remained
relatively constant over a wide range of anode voltages and emission currents.
Measurements of plate (anode) current/grid voltage characteristics
have been made and are presented in Fig. 17 for plate (anode) voltages of
30, 40 and 50 kV. The prot ;type design does not operate reliably above
50 kV due to marginal spacing within the oil filled tank (it is felt that
the design requirements affecting the problem are well understood and that
the final model will operate successfully at 80 kV) . It is apparent from
the curves that the maximum emission levels are adequate and that they can
be easily controlled with the available grid voltages. A beam current of
100 mA at 51 pulse duty factor and 30 kV represents an average anode power
dissipation of 150 watts.
-------
-29-
The amplification factor (y) based on the measured cut-off voltages
is approximately 1700, compared to the value of 3000 calculated from the
geometric factors. A more detailed analysis of the curves shows that they
do not obey the usual triode current law,
expected from single theoretical considerations. This result can be
explained by assuming a variable y triode. In view of the larger cathode
area and grid to anode spacing, and the overall configuration, it is
reasonable to postulate that the effective area of the cathode increases
with increasing anode voltage penetration through the grid. This would
result in a steeper dependence of Ip on Vg than that of the above equation.
Although there are possible adjustments in the lens configuration which
could reduce this effect, no changes in design are anticipated since the
departure from ideal triode behavior does not materially effect the pulse
tube performance.
The switching characteristics of the triode have been determined and
appear adequate. The grid circuit capacitance (including the high-voltage
connector) were measured to be < 15 pF. The turn-off time for the system
including detector response was < 300 ns. The ultimate limit on time
response is set by the signal/noise ratio in the detector system and its
effect on time resolution. Improvements in design of the pulser circuit
could reduce the timing uncertainty to as low as 150 ns which would result
in a further reduction in pile-up effects at the higher counting rates.
-------
-30-
The total system performance has been measured in the feedback mode
as a function of increasing sample mass and tube power. The basic control
circuitry operated in the mode described previously with the exception
that the maximum duty factor was determined manually instead of via feed-
back circuitry. One of the goals of the measurements was to determine
the optimum ON time; the OFF period was 50 ys for all the data to be dis-
cussed.
Figures 18 and 19 are calculated curves based on the observed data
from the rate measurements. Figure 18 is a plot of input rate vs. non-
pile-up output rate for the case of two choices of maximum ON times; the
turn-off time was assumed to be 300 ns. The performance for the 10 ys
ON time (and larger times) is adequate; however, for the case of 2 ys ON,
the pile-up probability is excessive even at moderate counting rates.
From these type of data we can obtain a maximum allowable input counting
rate based on a maximum acceptable pile-up probability. In the following
discussion, we will assume a limit of 5% pile-up as maximum for useful
analytical data. For 10 ys or longer ON times this occurs at approximately
17 kc assuming 50 ys OFF. If the timing uncertainty were reduced, the
maximum input rate would more closely approach the theoretical imit of
20 kc/s for 50 ys OFF time. Further increases in counting rate limit can
then be accomplished by going to shorter amplifier shaping times corres-
ponding to shorter OFF periods for the X-ray tube. This can only be
accomplished with some sacrifice in X-ray detector energy resolution.
Figure 19 is a plot of average and peak X-ray tube power vs. sample
mass using the X-ray yield measured with the prototype tube. Also included
is a plot of the output counting rate. The performance is as described
-------
-31-
previously. In the region of small sample mass, the tube is operated at
its maximum duty factor and corresponding maximum average power level. As
the sample mass increases, the peak power increases to compensate for the
increased dead time while maintaining the maximum average power. Finally
the instantaneous counting rate is high enough for there to be a 51 prob-
ability of an event during the shut-off time, i.e. the 5% pile-up limit is
reached. From this point on, the duty factor is fixed at some minimum
value and the average power is gradually reduced while the count rate is
maintained at the maximum value. The important benefits of the pulse mode
of operation result from the fact that the tube is operated either at the
maximum average power level with no power wasted during the system dead
times or at the maximum allowable counting rate. The feedback loops thus
ensure that the system operating in its best condition for a given sample
size.
The quality of the spectra obtained with the pulsed tube is determined
more by the secondary fluoresor geometry more than any electronic innova-
tions which might be employed. Of greater importance is the question of the
X-ray yield for a given tube current since the system is limited by X-ray
tube power over much of its operating range. However, the ultimate test of
the analytical system is the ability to measure the concentrations of ele-
mental constituents with a high degree of sensitivity. For this reason we
show in Fig. 20 a spectrum obtained on a 30 mg/cm2 sample of MBS orchard
leaves.
A Mo secondary target excited by Rh K X-rays was used for this measure-
ment. The design of the anode structure was not optimum for secondary
fluorescence efficiency and some increase over the measured yield is anticipated
-------
-32-
in the final design. Nevertheless an output counting rate of 5 kc/s was
achieved with 5 watts average anode power dissipation. The counting rate
in these measurements was limited by the resolving time of the guard-ring
rejection circuitry which was significantly greater than the X-ray tube
switching times. This difficulty will be overcome in the final system.
F. Conclusion
The design and construction of the prototype pulsed X-ray tube has
been completed. Tests show the system is capable of improved counting
rate performance over a wide range of sample masses. With the incorpora-
tion of improved control circuitry and live time correction methods, the
system will allow the analysis of wide range of sample types with automatic
optimization of excitation conditions.
Experimental data obtained on the prototype tube have been sufficient
to define the final design parameters for a complete analytical system.
This second-generation energy-dispersive X-ray analysis system is currently
being designed and will be completed for operation in the St. Louis RAPS
program in 1974.
-------
-33-
V. DIGITAL FILTER MRKING
After examining various techniques for marking digital computer-
readable codes on the filter holders, we have chosen to use stick-on
bar code labels (JVfonarch Marking Company). A standard bar code is
employed in these labels and identifying numbers are printed to pemLt
visual reading of the identification. The 37 x 9.5 mm area available
on the filter holder permits a 12-digit marking--far more than required
in practice. The bar code is read optically and existing hardware can
be used to connect to the computer.
Digital readers will be installed on both the X-ray elemental
analysis system and on the beta-gauging system. Samples will be moved
by the sample changer into an intermediate position for digital identi-
fication before being moved to the measurement head where X-ray or beta-
particle exposure occurs. The identification number of a particular
filter will then be automatically associated with all the data acquired
on that sample. This computerized bookkeeping capability is essential
for error free analysis of the large number of samples anticipated.
-------
-34-
VI. TOTAL PARTICULATE MEASUREMENTS
< Preliminary tests with a prototype beta gauge have demonstrated
the feasibility of measuring the total particulate mass accumulation dur-
ing sampling periods of six hours or less. Assuming reasonable aerosol
particulate mass concentrations, this would be equivalent to a detectible
limit of 10 ug/cm2 on filter substrates of 3 to 5 mg/cm2.
Using a 147Pm source and Si(Li) detector with appropriate counting
electronics, we have performed a series of measurements on standard plas-
tic thin films. Measurements of errors associated with reproducibility
of the sample position and counting intervals indicate an accuracy of
10 yg/on2 is easily achieved in short counting intervals (30 seconds or
less). Use of higher intensity radioactive sources could reduce this
time even further.
A more serious concern in the measurement of particulates collected
on membrane filters is the variation in mass of the substrate as a func-
tion of ambient relative humidity. A series of tests using different mem-
brane filter types was performed to measure the error due to these humidity
variations. Filters were cycled several times between 100% and 0% relative
humidity and the reproducibility of their mass, after conditioning at con-
stant humidity, was measured by direct weighing on a microbalance. The
three types of filters were teflon, cellulose ester (Millipore) and poly-
carbonate (Nuclepore). The reproducibility of weight was best with the
polycarbonate membranes--an average deviation from the mass of less than
0.5 ng/cm2 for two samples cycled between the humidity limits four times.
-------
-35-
The deviations for teflon were 0.8 yg/cm2 and for the MLllipore 6 yg/on2.
All of these are well below the required accuracy of 10 yg/on2. However,
the fact that the maximum weight changes with changes in relative humidity
were of the order of 50 yg/on2 emphasises the importance of equilibrating
the filters to a constant relative humidity before mass measurements. The
final beta gauge design will include a provision for controlled humidity
atmosphere, and the operating strategy will be designed to permit suitably
long conditioning of filters before gauging.
-------
-36-
REEERBNCES AND POOTNOTES
1. Jaklevic, J. M., Colliding, F. S. and Landis, D. A., IEEE Trans.
Nucl. Sci., NS-19, No. 3, 392-395 (1972).
2. Andersen, A. A., Amer. Ind. Hyg. Assoc., 27, 160 (1966).
3. Lundgren, D. A., J. Air Poll. Cont. Assoc., 17_, 225 (1967).
4. Whitby, K. T., Husar, R. B. and Liu, B. Y. H., "The Aerosol Size
Distribution of Los Angeles Smog", Publication No. 159, Particle
Technology Laboratory, University of Minnesota.
5. Berner, A., Private Communications.
6. Pierce, J. R., Theory and Design of Electron Beams, Van Nostrand,
New York, 145 ff (1954).
7. Spangenberg, K. R., Vacuum Tubes, McGraw-Hill, New York (1948).
* Earlier Progress Report
Goulding, F. S. and Jaklevic, J. M., "X-Ray Fluorescence Spectrometer
for Airborne Participate Monitoring", Environmental Protection
Technology Series EPA-R2-73-182, April 1973.
-------
-37-
FIGURE CAPTIONS
Fig. 1. Components of the ERC virtual impactor.
Fig. 2. Simplified assembly drawing of the ERC virtual impactor.
Fig. 3. U. V. fluorescence analyzer.
Fig. 4. Self-absorption effect in uranine solution.
Fig. 5. Size separation characteristics and the total loss spectrum.
Fig. 6. Regional loss spectrum, Region 1.
Fig. 7. Regional loss spectrum, Region 2.
Fig. 8. Regional loss spectrum, Region 3.
Fig. 9. Regional loss spectrum, Region 4.
Fig. 10. Regional loss spectrum, Region 5.
Fig. 11. Regional loss spectrum, Region 6.
Fig. 12. Schematic of pulsed tube electronics.
Fig. 13. Input/output counting rate characteristics of pulsed tube
compared to normal operation.
Fig. 14. Cross sectional view of electron optics used in pulsed tube.
Fig. 15. Schematic of control electronics for pulsed tube.
Fig. 16. Calculated plate current vs. voltage curves for prototype X-ray tube.
Fig. 17. Experimental plate current vs. voltage curves for prototype
X-ray tube.
Fig. 18. Plot of input rate vs. non pile-up output rate for two different
values of the maximum "on" time of the tube.
Fig. 19. Plot of X-ray tube power as a function of sample mass as auto-
matically determined by the circuit of Fig. 15. The change in
output counting rate is also shown.
Fig. 20. X-ray fluorescence spectrum of a sample (NBS orchard leaves) using
the pulsed X-ray tube. The counting time was 400 seconds.
-------
-38-
K
XBB 7311-6673A
Fig. 1
-------
-39-
INLET
-»- TO FLOW MONITOR
TO FILTER B
(49 LITERS/MIN.)
INCHES
0123
TO FILTER A
(1 LITER/WIN.)
XB.L 7311-6801
Fig. 2
-------
-40-
MERCURY LIGHT SOURCE
UV TRANSMITTING FILTER
FIBER GLASS OPTICS
SLIDING SHUTTER
GREEN FILTER
RCA 7265
PHOTO
MULTIPLIER
JARRELL-ASH
(26-780)
ELECTROMETER
AND H.V. SUPPLY
XBL 7311-1495
Fig. 3
-------
-41-
0)
o.
CO
§10
-1
10
-2
4 8 16
-7 2
x(x9.942xlO gm/cm
32 64
XBL 7311-1492
Fig. 4
-------
-42-
1 I
j I
o o
o
00
O
(O
o
If)
o
CO
pq
X
oo
N
V)
LU
_J
u
CO
CM
o o
CM rH
NOIllSOd3Q %
-------
-43-
o
14
13
12
11
10
9
2 7
UJ
a
* 6
5
4
3
2
1
REGION 1
567
PARTICLE SIZE (\tm)
10
XBL 7311-1499
Fig. 6
-------
-44-
28
26
24
22
20
18
O 16
12
s?
10
8
6
4
2
REGION 2
34567
PARTICLE SIZE(pm)
10
XBL 7311-1496
Fig. 7
-------
-45-
oo
r-
CQ
o>
o -
00
CO
z
o
LU
in
oo
bO
51
CO
c
(J
CM
j
CM r-l
o> oo r»»
-------
-46-
o
C3
UJ
ce
o>
00
00
o-
CP
X
to
ui
o
IT)
0
I I I I I I
J I
CM
CM <-H O O>
fl - -!
oo
in ^- co CM «-
NOIllSOd3Q
-------
-47-
34
32
30
28
26
24
22
z
° 20
I 18
tu
0 16
14
12
10
8
6
4
2
0
REGION 5
PARTICLE SIDE (pm)
10
XBL 7311-6800
Fig. 10
-------
-48-
r i r
(O
o
UJ
ec
bO
is:
in
o-
CM tH
O) CO
If)
NOiiisodaa %
-------
-49-
CQ
X
-------
-50-
100
I I I I
PULSE
SHAPE
O
^
UJ
10
o»
h-
o
17
4 41 (IS »
(DEAD TIME)
i i i i i i i i
Iiiii i i jt
PULSED X-RAY
EXCITATION
s -
41 psec DEAD TIME _
NORMAL
RANDOM
PULSES
10
INPUT RATE Kc/s
till
100
XBL 733-288
Fig. 13
-------
-51-
6.5 cm
CATHODE
LENS
XBL 736-859
Fig. 14
-------
-52-
CTl
LO
VO
LO
b
-------
-53-
a.
E
- 4
-100
Vg(volts!
Fig. 16
-------
-5*-
150
100
50
Vp Vg(CUTOFF)
30 KV -18 V
40 KV -24 V
50 KV -28 V
o / £
try ' AT /
// / O '
/ /
' /
/ t
.
-20
-15 -10
Vg(VOLTS)
-5
XBL 7U2-U18
Fig. 17
-------
-55-
lOus ON
2us ON
10
INPUT RATE
XBL
Fig. 18
-------
-56-
A) 10 MS MAXIMUM 'ON'
B) 20 IS MAXIMUM 'ON' I
PULSE POWER
" (A)
COUNTING RATE
AVERAGE POWER
1 10 100
SAMPLE MASS (mgm/cm )
XBL 7^2-1+17
Fig. 19
-------
-57-
Z
O
a, OQ
Q> «. rf
£< «2
^S UJ *vi ^*
o-^mx
< o " ..r en
I oo 3> <
c/>o E o ^
00 (K OO
z o co * m
o
en
o
m 7>
uidd ££ JS ». '..'v!'!.
uidd zi qd -'<:^.,
"''' ' "^ >
uidd OT J8 :C a
\
uidd trl s\/ 'uidd gfr qj '*''.'.'. "*
uidd 92 NZ *"''
uidd 21 no ^
5
uidd 00£ 9J '==::-mi:r:J
uidd 06 u|/\|/:
J
o
10
o
en
OJ
A
bo
S1NH03
-------
-------
-59-
APPENDIX A
A. Hardware
The T. I. 960A computer system required as a component in the
analysis facility for the 1974 St. Louis RAPS program has been assembled
and fully tested. It has been used in studies of the analysis programs
and for testing associated hardware as discussed below.
The link between the TI 960A computer and the X-ray system has
been changed to eliminate potential and existing noise problems. In place
of the multi-wire parallel-driven cable, we now have a 4-coax serial trans-
mission communicator. The communicator consists of two shift registers at
the computer and two shift registers at the X-ray machine. The data
shifted out of one of the computer shift registers is transmitted via a
120-ohm coaxial cable to the data input lead of one of the X-ray system
shift registers. Conversely, data from one of the shift registers at the
X-ray machine is transmitted via another 125-ohm coaxial cable to the
data input lead of a shift register at the computer. Clock signals, hav-
ing correct time relationships relative to the data signals are trans-
mitted with each data signal. Thus we have the four coaxial cables linking
the two machines.
The reasons for converting to this scheme were two fold.
i) Experience with the existing cable and parallel-transmission
scheme demonstrated severe noise problems requiring rather drastic noise-
suppression measures.
ii) There was considerable uncertainty regarding the length of the
cable in possible future configurations. The production, debugging and
-------
-60-
maintenance required to change the four 125-ohm coaxial cables is very
small. To change the multi-wire cable is a relatively major job. The
characteristic impedence of the cables is 125 ohms and the need for cor-
rect relative timing between clock and data signals requires that the
cables be of almost the same length. We have successfully tested the
system with 30 meter cables separating the computer from the X-ray sys-
tem. The maximum permissible cable length is a function of cable quality.
B. Software
1. The algorithm used to detect spectrum shifts and correct for
them has been thoroughly analyzed. The previous technique involved
calculating a residual for the test region at each allowed shift point
(up to ± 1 channel in 1/4 channel increments) and selecting the shift
value which minimized this residual. This has been changed, with drama-
tic results, by forcing the centroids of the selected region (one in the
reference spectrum and one in the data spectrum) to be equal. This is
achieved by using the difference in centroids to approximate the required
shift, and then reiterating the process until the centroids converge.
To further ensure that the computed shift locates the data spectrum in
the center of the "dead-zone" of no-difference in centroid values, the
program slides the data spectrum back and forth to determine the bound-
aries of this dead-zone, and then sets the actual shift value at the cen-
ter of this dead-zpne. The shift increments have also been reduced to
1/16 of a channel.
-------
-61-
2. Several convenience features have been added to the program.
First, two markers are now available for manipulation, and an option has
been added to allow typing out the counts in each channel between the two
markers. Second, to allow for expansion of the number of program options
available, the operation of the ALT key on the control panel has been
changed to make the teletype request an option number, which is then
entered from it.
3. Calibration procedures have also been changed. In the CALIBRATE
mode, the calibration samples are counted only into buffer 4 (the double
precision buffer). When the largest count in the buffer exceeds 960,000
or when the timer terminates the count, the data is divided by a suitable
constant to allow it to be stored in a single precision buffer area. The
spectrum is then stored on tape for later use.
The new calibration data-entry routines start by the system
requesting an option number (0-7). Typing a "," makes the routine operate
as it did in the original program. Options 0-7 permit entry into the
program at different points, eliminating many redundant requests for
information. When the program requests new data it displays the previously
stored data.
All spectral calibration data is packed in memory starting at
the bottom of the first buffer area. Data relating to the calibration
standards inserted by the operator is packed from the top of buffer #5
downwards. Sixteen "descriptor" locations are reserved for each element.
If an overlap occurs between incoming spectral data and the "descriptor"
area, an error message occurs and some information is not saved.
-------
-62-
There is no longer any restriction (except length) on the tag
name given to any element. One of the calibration options now allows
assigning of an element to an analysis priority group without erasing it
from any previous group. One of the regular options destroys all calibra-
tion information, producing a virgin program.
4. Error calculations have been improved but further work remains
to allow for inter-.element effects. Results are now printed out in a
fixed sequence going from lighter to heavier elements. Changes have also
been made in the output format to minimize typing time.
-------
-63-
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1 REPORT NO.
EPA-650/4-74-030
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Development of Air Particulate
Monitoring Systems
5. REPORT DATE
May 1974
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
F. S. Goulding and J. M. Jaklevic
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORG "\NIZATION NAME AND ADDRESS
Lawrence Berkeley Laboratory
University of California
Berkeley, California 94720
10. PROGRAM ELEMENT NO.
1A1103
11. CONTRACT/GRANT NO.
IAG-04-0377
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Protection Agency
National Environmental Research Center
Research Triangle Park, N. C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final Jan. 1973 to Jan. 1974
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Progress is described for the first year of a two year effort to fabricate a
high-speed aerosol analysis system. Twelve automated dichotomous sampling stations
which collect particles in the 0 to 2 and 2 to 20-ym diameter ranges will be built.
An evaluation of a prototype virtual impactor for aerodynamically separating
particles into two size ranges is given. Design improvements for minimizing
losses are described. Design parameters have been determined for a high-speed
energy dispersive X-ray fluorescence spectrometer capable of analysis of elements
with atomic numbers above 13. A prototype beta gauge for total mass determination
has been tested, and a detection limit of 10 yg/cm2 has been measured. For the
purpose of sample identification, a system is described by which the X-ray spec-
trometer and the beta gauge will be able to identify a digital code marked on each
filter.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
Airborne Particulate Monitor
Aerosol Size Fractionation
X-ray Fluorescence
X-ray tubes
Trace Element Analysis
Beta Gauge Mass Monitor
8 DISTRIBUTION STATEMENT
Release Unlimited
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21 NO. OF PAGES
65
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
U.S. GOVERNMENT PRINTING OFFICE: 1974 - 640-878/638 - Region 4
-------
-------
"'Mill' I
-------
n
;!
------- |