Environmental Protection Technology Series
VERSATILE
¦
GAS FILTER
CORRELATION SPECTROMETER
U.S. Environmental Protection Agency
Office of Research and Development
Washington, D. C. 20460

-------
EPA-600/2-75-024
VERSATILE GAS FILTER
CORRELATION SPECTROMETER
D. E. Burch, F. J. Gates, D. A. Gryvnak
and J. D. Pembrook	3©^
Aeronutronic Ford Corporation
Aeronutronic Division
Newport Beach, California 92663
Contract No. 68-02-1227
Project Officer
F. M. Black
Emissions Measurement and Characterization Division
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
U.S. Environmental Protection Agency
Sam Nunn Atlanta Federal Center
Region 4 Library
61 Forsyth Street S.W.
Atlanta, Georgia 30303

-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research Laboratory,
U.S. Environmental Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the view and policies
of the U.S. Environmental Protection Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
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 cate-
gories were established to facilitate further development and application of
environmental technology. Elimination of traditional groupings 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 instrumenta-
tion, 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.
This document is available to the public through the National Technical Infor-
mation Service, Springfield, Virginia 22161.
ii

-------
CONTENTS
Page
List of Figures	iv
List of Abbreviations and Symbols	v
Acknowledgements	vii
I	Introduction	1
II	Summary	3
III	Conclusions	5
IV	Recommendations	7
V	Layout and Spectroscopic Principles	9
VI	Formaldehyde Gas-Filter Cell	21
VII	Multiple-Pass Sample Cell	27
VIII	Grating Section	33
IX	Electronics and Processing of Detector	Signal 41
X	1^0 Monitor	45
XI	Results of Formaldehyde Tests	49
XII	Measurements of NH^ and Vinyl Chloride	Near 1000 cm ^ 59
XIII	References	67
iii

-------
LIST OF FIGURES
Number	Page
1	Optical Diagram and Layout of the Instrument	10
2	Photograph of the Instrument	11
3	Two drawings of the Formaldehyde Gas-Filter Cell	22
4	Plot of the Pressure of Formaldehyde Vapor vs the GFC Temperature 25
5	Photographic View of One End of the Multiple-Pass Sample Cell from
the Back Side of the Instrument	28
6	Sample Inlet and Optical Diagram of the Bypass Optics Assembly	29
7	Multiple-Pass Optical System in the Sample Cell	31
8	Optical Diagram of the Retroreflector-Grid Assembly	34
9	Optical Diagram Defining Angles that Relate the Grating Position
to the Wavelength of Energy Passed by the Grating Assembly	35
10	Calibration Curves Relating the Grating Drive Counts to Wave-
length and to Wavenumber	37
11	Curves of &v/6g and cos ff/cos (3 vs Wavenumber for the Three
Gratings	39
12	Plot of Trapezoidal Slit Function	40
13	Block Diagram of Electronics	42
14	Two Optical Diagrams of the 1^0 Monitor	46
15	Diagram of the Receiving Optics for the 1^0 Monitor	47
16	Spectral Curves of Transmittance for Formaldehyde and Five
Hydrocarbons	50
17	Recorder Plots of the Noise on the Signal Output for Various
Operating Conditions	5 7
18	Spectral Curves of Transmittance of NH„, Vinyl Chloride, and
Ethylene Between 92 0 and 985 cm"l	60
iv

-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
GFC	--gas-filter cell, also called correlation cell
T	--traasmittance at a particular wavenumber (or wavelength) that
would be measured with a spectrometer having infinite resolving
power. Subscripts indicate the transraittance of different com-
ponents or gas samples. T indicates the average transmittance
over a specified interval.
Tg	--transmittance of the sample gas being analyzed
T_£	--transmittance of the spectral bandpass filter
T	--transmittance of the gas in the GFC
§
T	--transmittance of an attenuator, assumed to be constant over the
spectral bandpass of interest
T	--average transmittance of a sample as measured with the beam of
^'S	energy passing through the GFC. (Equation 6)
T	--average transmittance of a sample as measured with the beam of
a 's	energy passing through the attenuator. (Equation 7)
A	--absorptance, A = 1 - T. Subscripts used with A are the same as
those used with T
v	--wavenumber of radiant energy (cm
X	--wavelength of radiant energy expressed in micrometers (jj,m)
v X	--wavenumber, or wavelength, of the center of a specified interval
c c
f	--carrier frequency (high) (Hz)
c
f	--alternator frequency (low) (Hz)
a
p	--partial pressure of a particular gas species (atm)
P	--total pressure of a gas mixture (atm)
I	--geometrical path length of the radiant energy beam through a
gas sample (cm)
c	--p/Pj concentration of a gas species, usually expressed in parts
per million (ppm) or percent
u	--p£ (atm cm), absorber thickness of a particular gas species.
One atm cm is equivalent to 10^ ppm meters
k	--average absorption coefficient of a sample over a specified
S	interval (atm cm)" , (Equation (21))
v

-------
M	--constant that accounts for aperture size and field-of-view in
radiant energy beam,(Equation (1))
R	--responsivity of the detector at the wavenurnber of interest
--spectral emissivity of the radiant energy source at the wave-
number of interest
E	--the amount of radiant energy chopped at 360 Hz that is incident
®	on the detector during the GFC-half of the alternator cycle,
(Equation (1))
E	--the amount of radiant energy chopped at 360 Hz that is incident
on the detector during the attenuator-half of the alternator
cycle, (Equation (2))
h	--E - E (Equation (10))
g at
S	--E + E (Equation (11))
g at
V	--voltage component of amplified detector signal at carrier fre-
quency fc
V	--voltage component of amplified detector signal at alternating
frequency f
V1	--Va/Vc, normalized voltage. Normalization is made so that V' =
1 when there is 1007° modulation of the beam at frequency fg.
See Equation (12) and related text
AGC	--automatic gain control.
C	--correlation efficiency (Equations (13) and (15))
e
F	--lim V' Pembrook Factor (atm cm) , (Equation (19))
u~*0 u
D.R.	--discrimination ratio, ratio of the concentration of an inter-
fering gas species to the concentration of the species being
measured that produces the same reading. This ratio may be
positive or negative.
f	--focal length of a mirror or lens (cm)
a,|3 ,P ,0	--angles relating position of grating to the incident and dif-
fracted beams (see Equations (23) through (26))
D	--distance between adjacent grooves of a grating
W ,W	--physical width of the entrance slit and opening in the grid,
n ®	respectively
g	--distance measured at the grid in the direction of the disper-
sion (mm)
vi

-------
ACKNOWLEDGEMENT
The authors would like to express their appreciation for the cooperation and
assistance of four co-workers. Mr. Wayne Leicht and Mr. Kim Choy designed many
of the components and assisted in acquiring the purchased parts. Mr. Oliver
Pacific and Mr. George Hart performed most of the machine shop work and were
very helpful in assembling and checking out the instrument.
vii

-------
THIS PAGE LEFT BLANK INTENTIONALLY
viii

-------
SECTION I
INTRODUCTION
During some experimentation a few years ago with gas-filter correlation instru-
ments we found that instrument stability, sensitivity and discrimination could
be improved significantly by carefully selecting the spectral bandpass. Under
two previous contracts with the Environmental Protection Agency,Aeronutronic
designed and built two gas-filter correlation instruments, each of which included
a small grating assembly that made it possible to select the optimum bandpass.
Commercially available interference filters frequently cannot be obtained with
the optimum transmission characteristics. The grating assemblies are based on
a principle developed by us-' for an instrument that employed a prism as a dis-
persing element instead of a grating. The width and the center position of the
spectral bandpass can each be adjusted independently while measuring the sensi-
tivity and interference as the instrument is being assembled and tested. Thus,
the optimum bandpass can be determined with very good precision.
A single grating assembly can be used to provide the desired spectral bandpasses
for two or more different gas species. The grating assembly in the instrument
described in Reference 1 transmits a spectral band near 3.3 gim for the measure-
ment of CH4 and another band near 4.6 am for the measurement of CO. A single
grating assembly in the instrument described in Reference 2 transmits a spectral
band for CO and one for NO. Each band consists of several very narrow intervals,
each of which includes a strong absorption line of the gas species being measured.
All of the optical components in these grating assemblies were bonded permanently
in place so that the bandpasses could not be shifted after the alignment had been
completed.
1.	Burch, D. E., and J. D. Pembrook. "Instrument to Monitor CH4, CO, and CO2
in Auto Exhaust." Prepared by Philco-Ford Corporation for EPA under Con-
tract No. 68-02-0587. EPA Report No. 650/2-73-030, October 1973.
2.	Burchj D. E., and D. A. Gryvnak. "Infrared Gas Filter Correlation Instru-
ment for In-Situ Measurement of Gaseous Pollutants." Prepared by Philco-
Ford Corporation for EPA under Contract No. 68-02-0575. EPA Report No.
EPA-650/2-74-094. Also, Burch, D. E., and D. A. Gryvnak. "Cross-Stack
Measurement of Pollutant Concentrations Using Gas-Cell Correlation Spectro-
scopy." Chapter 10 of Analytical Methods Applied to Air Pollution Measure-
ments, Stevens, R. K. and W. F. Herget, (eds.). Ann Arbor, Ann Arbor
Science Publishers, Inc., 1974.
3.	Burch, D. E. "Adjustable Bandpass Filter Employing a Prism." Appl. Opt. _8,
649 (1969).
1

-------
The original objective of the project reported here was to incorporate a similar
grating assembly in a gas-filter correlation instrument with a multiple-pass
sample cell to measure the concentration of formaldehyde in automotive exhaust.
This method of measurement promises to be much simpler, and possibly more ac-
curate, than other methods being used for formaldehyde. After the initial phases
of the design, it was decided to maintain the same basic instrument with good
performance for formaldehyde while making it adjustable with several interchange-
able parts so that other gas species could also be studied.
2

-------
SECTION II
SUMMARY
A gas-filter correlation instrument has been built with interchangeable sources,
detectors, windows, lenses, gratings and filters for use from approximately
0.3 am to 11 p,m. A precision screw can rotate the grating to the desired orien-
tation to select the center of the spectral band passed by the instrument.
Interchangeable slits make it possible to select the width of a single spectral
band or of a series of narrow spectral intervals. Spectral curves of transmit-
tance can be scanned by rotating the grating continuously.
The main sample cell has a base length of 1 meter and contains a multiple-pass
optical system that can be adjusted to give path lengths up to more than 40
meters. Electrical heating wire coiled around the body of the cell makes it
possible to heat the cell to approximately 55°C to avoid condensation of H2O.
A temperature controller automatically maintains the temperature at the selected
level.
The instrument has been equipped with components optimized for the 3.6 |j,m region
where there is strong absorption by formaldehyde, the gas species of primary
interest. The majority of the tests were also performed with this gas species
in order to obtain good sensitivity and discrimination against other gases,
particularly hydrocarbons, that absorb in the same spectral region. The gas-
filter cell (GFC) that contains formaldehyde vapor and makes the instrument sen-
sitive to this gas species is heated to approximately 60°C. Powdered paraformal-
dehyde is placed in the bottom of the cell, and the equilibrium pressure of the
formaldehyde vapor over the powder is controlled by regulating the temperature.
When all of the instrument parameters are optimized and the sample cell is ad-
justed to 40 meters path length, the minimum detectable concentration of formal-
dehyde is less than 0.1 ppm.
Water vapor occurs in relatively high concentrations in automotive exhaust and
in many other gas samples that might be analyzed with the instrument. Because
of the many H2O absorption lines throughout the infrared, this gas is likely to
interfere with the measurement of many different gas species. In order to ac-
count for possible H2O interference, an H2O monitor is included in the instru-
ment. The output of the monitor can be calibrated to provide an accurate
measurement of the H2O concentration and can be fed into the main electronics
to automatically account for the 1^0 interference.
In addition to the formaldehyde tests, several tests were performed on NH^
(ammonia) and on vinyl chloride in the spectral region between 10 p,m and 11 u,m
where both of these gas species have very strong absorption features. By
3

-------
employing spectral intervals only a few cm"-'- wide, it is possible to obtain
adequate discrimination against other gases that may be present in the samples.
The minimum detectable concentrations are approximately 0.4 ppm for NH3 and
1 ppm for vinyl chloride. By obtaining a detector with a smaller sensitive
element and a higher detectivity, these values could be reduced by a factor of
5 or more. Additional improvements could also be made by using mirrors with
extra high reflectivity and by making other changes to increase the amount of
energy transmitted through the optical system.
In the visible and ultraviolet, the performance is limited by low reflectivities
of the mirrors and low efficiency of the grating. With the grating, the deuter-
ium-arc lamp source and the photomultiplier provided, the lowest useful wave-
length is approximately 0.32 p,m. This limit could be extended to shorter wave-
lengths by employing a grating blazed for shorter wavelengths and mirrors coated
for high reflectivity in this spectral interval.
4

-------
SECTION III
CONCLUSIONS
The versatility and high sensitivity of the instrument have been demonstrated.
It can operate in the normal mode as a gas-filter correlation instrument with
the spectral bandpass being held constant. Gas-filter cells can be interchanged
to make the instrument sensitive to virtually any gas that contains sharp absorp-
tion features in its infrared, visible, or ultraviolet spectrum. The spectral
bandpass can easily be selected to include the absorption features of the gas
species to be detected.
The sensitivity, stability and discrimination of the instrument are adequate
for most purposes when measuring the concentration of formaldehyde vapor in auto-
motive exhaust. The optimum spectral bandpass and pressure of the formaldehyde
vapor in the gas-filter cell have been determined experimentally. Tests performed
with ammonia and vinyl chloride indicate that either of these gas species can be
measured at low concentrations by gas-filter correlation methods.
By rotating the grating at constant speed, it is possible to scan spectral curves
of transmittance. The liquid-nitrogen-cooled detectors provide high enough de-
tectivity that good signal-to-noise ratios can be obtained with the spectral
slitwidths as narrow as approximately 0.5 cm~l throughout the region between
3 um and 11 p,m.
Because of the high versatility desired for the instrument, it is necessarily
more complex than would be required for a gas-filter correlation instrument de-
signed for a single gas species. In most cases, somewhat better performance
could also be attained with an instrument operating on the same basic principles
but designed for only one gas species. One factor leading to better performance
is the higher reflectivity of the mirrors that would be possible if they were
required to operate only over one narrow spectral interval. Several of the
optical components could also be made more stable if they were assembled per-
manently and were not made to be interchangeable.
5

-------
THIS PAGE LEFT BLANK INTENTIONALLY
6

-------
SECTION IV
RECOMMENDATIONS
The instrument contains two important features that are not found in any other
available gas analyzers. One is the versatility that makes it adaptable for
the measurement of concentrations of many different gases of interest. The other
important feature is its capability for the real-time measurement of the concen-
tration of formaldehyde in automotive exhaust. Maximum value of the instrument
can probably be realized by making full use of these two features as opposed to
using it for routine measurements that could be performed by other instruments.
Gas-filter correlation techniques have been considered for several gas species
for which adequate instruments are not presently available. The instrument re-
ported here can be very valuable in determining the potential performance of any
gas-filter correlation instrument that could be designed and built for most gases
of interest. Thus, this type of instrument can be evaluated without the cost of
building a prototype. The interchangeable and adjustable components make it con-
venient to determine the optimum spectral bandpass and the optimum amount of gas
for the gas-filter cell, and to measure the sensitivity and the interference by
gas species other than the one being measured. Different sources of noise and
instability can be investigated and related to what might be expected for another
instrument designed for a single gas.
Recommended use of the instrument to measure the concentration of formaldehyde
is expected to provide valuable information on the overall performance and to
indicate possible improvements. Among the items that should be considered are:
methods of flushing the sample gas through the sample cell, methods of improving
the stability of the "zero-reading", different types of attenuators, methods of
controlling the temperature of the formaldehyde gas-filter cell, and alternate
methods of mounting the optical components.
7

-------
THIS PAGE LEFT BLANK INTENTIONALLY
8

-------
SECTION V
LAYOUT AND SPECTROSCOPIC PRINCIPLES
OPTICAL COMPONENTS AND LAYOUT
The optical layout of the instrument is illustrated in Figure 1, and Figure 2
is a photograph taken from above and in front of the instrument. Except for a
power supply and a few minor accessories, all of the components of the instru-
ment are mounted to a single baseplate of extruded aluminum approximately 38 cm
wide by 180 cm long. The optical components consist of five main sections.
The mirrors in each section are indicated by an identifying letter as follows:
N, entrance section, C, sample cell; X, exit section; G, grating section; and
A, alternator. In order to simplify the diagram, the central ray is not in-
cluded in the sample cell and grating sections. The positions of the extreme
rays on the mirror surfaces are indicated, and the images of the source are
formed at the places where the extreme rays cross.
The entrance optics contain the radiation source, a 360 Hz chopper and the
mirrors that direct the radiant energy into the sample cell. The primary source
is a Nernst glower, and alternate sources can be used by placing removable
mirror N5 in the position indicated. The long sample cell used to measure
formaldehyde concentrations, and other gases of low concentration, has a base-
length of one meter and a multiple-pass optical system that can be adjusted to
produce pathlengths of more than 40 meters. This sample cell (described in
detail in Section VII) has a stainless steel body with inside diameter of
10.2 cm. The sample gas can be left in the cell without flushing, or it can
flow continuously in one end of the cell and out the other. The cell can also
be evacuated.
The exit section of the instrument consists of the mirrors and other components
that direct the beam coming from the sample cell to the entrance slit of the
grating section. The grating section provides a means of obtaining essentially
any desired spectral interval from approximately 3000 angstroms in the uv to
12 microns in the infrared. The center of the spectral interval passed by the
grating section is determined by rotating the grating with a precision screw
mechanism. The width of the spectral bandpass is adjusted by varying the open-
ing in a grid, not shown, between two flat mirrors G4 and G5. These two mirrors
are mounted perpendicular to each other with the plane of each 45° from the
horizontal. Energy in the spectral interval passed by the grid is reflected by
mirror G5 back to mirror G3. The beam is then returned to the grating, which
undisperses it and directs it back to G2 and out the exit slit S2. The emerging
beam passes just below mirror Gl before it reaches slit S2. The retro-reflector-
grid assembly that contains mirrors G4 and G5 is discussed in greater detail in
Section VIII.
9

-------
Figure 1. Optical diagram and layout of the instrument.

-------

Figure 2. Photograph of the instrument.

-------
The alternator section directs the energy of the selected spectral interval from
the exit slit of the grating section to the detector. A rotating mirror chopper,
A3, causes the beam to alternately travel over two different paths to the detector.
When mirror A3 is in the open position, the energy passes by it to flat mirror A1
and on through the gas-filter cell (GFC) to spherical mirror A2. Flat mirror A6
is tilted 45° from the vertical so that the beam reflected by it is directed up-
ward through a bandpass filter to the sensitive element of the detector. The
relative positions of mirror A6, the filter, and the detector have been modified
in the diagram so that they can be shown more easily. When mirror A3 is in the
closed position, it reflects the beam to mirror A4, through an attenuator, and
on to the filter and detector via mirrors A5 and A6. The attenuator is adjusted
to provide the same average transmittance over the spectral interval of interest
in the attenuator leg as that in the GFC leg. This adjustment is made when there
is no sample, or other absorbing gas, in the sample cell. Under this condition,
the amount of energy from the source that reaches the detector is the same during
each of the halves of the alternator cycle that correspond to the open and closed
position of the rotating mirror chopper A3. As discussed in more detail below,
the addition of sample to the sample cell modifies the spectral distribution of
the transmitted energy in such a way that more energy reaches the detector during
the GFC half of the cycle (when the energy is passing through the GFC) than when
the energy is directed through the attenuator. This difference results in a
modulation of the 360 Hz signal; the frequency of the modulation, 30 Hz,is that
of the rotating mirror chopper A3.
The narrow bandpass filter located just below the detector serves two important
purposes. First, it blocks out radiant energy of overlapping orders that are
transmitted by the grating section. For example, if the grating section is ad-
justed to pass through 3.6 p.m energy, it also pass.es 3.6/2, 3.6/3, 3.6/4 vim....
etc. The narrow bandpass filter passes the 3.6 um energy and blocks the over-
lapping orders of shorter wavelengths. It can also be used to pass second or
third order while blocking out all others. Another important function of the
bandpass filter is to reduce stray energy that may reach the detector. This
scattered light may originate from several different places. One possibility
is energy from the source that is chopped at 360 Hz by the high frequency chopper
and is scattered from various components to the detector via paths other than
the desired one. This energy is undispersed and therefore covers a wide spectral
interval. Much of it is eliminated by shields not shown in the figure, but it
is further reduced by the narrow bandpass filter.
The rotating mirror chopper, A3, emits some energy and reflects other energy
emitted by the GFC or other components to the detector. This energy is modu-
lated at the alternator frequency, 30 Hz, because of the rotation of mirror A3.
This energy results from temperature gradients between different components in
the instrument. This modulated energy is not dispersed by the grating section
and covers a wide spectral interval. Therefore, it is greatly reduced by the
narrow bandpass filter in front of the detector. Energy that is not modulated
does not degrade the performance of the instrument. To a large extent, the
30 Hz energy emitted or reflected by mirror A3 is not chopped at 360 Hz and is
accounted for by the electronics used to process the detector signal. However,
if this 30 Hz signal is too large, it can saturate some of the electronic com-
ponents, or the harmonics of the signal can be transmitted to produce errors
in the measurements.
12

-------
Two additional measurable quantities are defined as follows:
[ M N R T, T T dv	E
5 , i	»—t s 5 -	_s .	(6)
S's	f M'N R T, T d^	E°
v f g	g
[ M N R Tr T Tdv	E
- = j—v—f_aL_s	_	(7)
'S ^ M N R T T dv	E°
v f at	at
The quantity Tg s in Equation (6) is the average transmittance of a sample as
it would be measured in the beam that also passes through the GFC. Similarly,
Equation (7) gives the average transmittance of a sample as it would be measured
in the beam that passes through the attenuator side of the alternator. In most
situations, the attenuator transmittance Tat is essentially constant over the
spectral interval being considered; therefore, this factor can be removed from
under the integral si^ns of Equation (7), making this equation identical to
Equation (3). Thus, Tat,s = Tg. The average absorptances that correspond to
the average transmittances defined by Equations (6) and (7) are:
A	= 1 - T , and
g,s	g,s'
A . = A = 1 - T	= 1 - T .
at, s	s	at,s	s
(8)
(9)
When the sample contains gas of species x, the same species as in the GFC, there
is a positive correlation between the spectrum of the sample (Ts) and the spec-
trum of the GFC (Tg). Because of this positive correlation in the spectral
structures, Tg)S is grear.ar than T , and Eg is greater than Eat. A sample may
contain gas other than species x that absorbs in the spectral interval passed
by the instrument (where Tf 4 0). If the spectral structure of such a gas has
no correlation with the spectral structure of species x, Tg,s " Tat, s > an^
E„ = Eat. In the c£tse that the spectral structure of the sample gas is such
tnat Eg < Eat, and TgjS < Tat s, there is said to be negative correlation be-
tween the sample gas and species x. It follows that a gas species in the
sample that is negatively correlated with species x will cause the instrument
to indicate too low a concentration of species x, the species being measured.
We define two additional useful quantities:
A = E	- E ,	(10)
g at'	v J
2 = E	+ E .	(11)
g at	v '
15

-------
The quantity A represents the difference in the chopped energy reaching the de-
tector during the two halves of the alternator cycle and is therefore propor-
tional to Va, the 30 Hz component of the detector signal. The constant factor
relating Va to A depends onthe gain of the system between the detector and the
point where Va is measured. Similarly, £ is the sum of the same two energy
values and is proportional to the average value of Vc, the 360 Hz component of
the amplified detector signal. The voltage Va is zero for no sample and is
directly related to the concentration of the gas species in the sample cell.
From Equations (1), (2), and (10), we see that A and Va are also proportional
to the spectral radiance Nv of the source and to the responsivity R of the de-
tector. Thus, if the electronic gain remains constant, calibration data relat-
ing Va to concentrations are valid only as long as Nv and R also remain con-
stant. Changes in Nv and R also influence £ by approximately the same factor
as they do A. Therefore, the ratio A/£ is independent of changes in either Nv
or R as long as either changes by a constant factor over the spectral interval
of interest. Neither A nor £ is easy to measure directly; however, the ratio
A/£, which is proportional to Va/Vc, is measured by the instrument described
in this report. In adjusting the instrument gain settings,_the radiant energy
beam through the attenuator is temporarily blocked so that Tat and, in turn,
Eat equal zero. Under this condition, A/2 = 1, and the amplifier gain settings
are adjusted so that the voltages Va and Vc are equal.
We define
V' = V /V	(12)
3 C
with the gain settings adjusted as just described. The gain settings that af-
fect this ratio are kept fixed so that V1 = Va/Vc = A/Z when the attenuator is
unblocked and measurements are made.
An automatic gain control (AGC) circuit maintains Vc constant to simplify the
measurements. When this is used, V1 is proportional to Va, a quantity that is
measured directly by the instrument. Thus, only one voltage, Va, needs to be
read in order to determine A/2, the quantity that is related directly to the
absorber thickness of the gas species being measured. Calibration data are
obtained by measuring V' for a series of standard samples of known concentra-
tion.
In practice, small changes in the average Nv due to variations in source tem-
perature are also accompanied by second order changes in the spectral distri-
bution. These slight changes can cause small differences in Eg and Eat if
there is any correlation between Nv and Tg. This results in a slight shift
to the "zero" that is measured with no sample. After a few hours warm-up
time for the instrument, zero-shifts due to source or detector variations over
periods of several minutes usually correspond to changes in V' of less than
10"^, which corresponds approximately to the lowest concentrations being mea-
sured. These small shifts can be determined and accounted for easily by flush-
ing the sample cell with a non-absorbing gas. Ordinary changes in Nv and R
do not significantly affect the relationship between absorber thickness and V1,
if the small zero shift is accounted for.
16

-------
The accumulation of dirt on mirrors or windows can reduce the energy reaching
the detector. If the dirty components occur outside of the alternator section,
the attenuation affects both Eg and Eat by exactly the same factor, provided
the attenuation is the same over the entire spectral interval of interest. Thus,
A/E and V' are unaffected by the dirt as long as enough energy reaches the de-
tector to make the AGC operate properly. Thus, the system involving the two
frequencies, fc and fa, along with the AGC circuits provides a reliable and con-
venient method of determining concentration from a single voltage measurement.
In considering the effect of dirt on windows, or of anything else that may ob-
struct part of the radiant energy beam, it becomes apparent that none of the
components in either leg of the alternator should provide an aperture stop or
limit the field-of-view. This corresponds to a different value of M in Equation
(1) than in Equation (2). If they are different, an obstruction to part of the
beam anywhere outside of the alternator might have a different effect on the
beam through one leg of the alternator than on the beam through the other leg.
For example, let us consider the case in which the windows on the GFC are too
small to accept all of the beam incident upon them, but the attenuator leg
transmits the entire baam. An obstruction in the beam that blocks rays that
would be blocked by the undersize GFC windows would not reduce Eg, but it would
reduce Eat. This would result in a zero-shift of the instrument. The serious-
ness of this type of situation depends on how frequently and how easily a "zero-
reading" can be made with no absorbing gas in the sample cell.
An aperture stop or a limit to the field-of-view in either leg of the alternator
can also reduce stability, which may be a more important consideration than a
slow change in the zero riding such as the one discussed in the previous para-
graph. Consider a "geometrical" attenuator that completely blocks a portion of
the edgy of the beam in the attenuator:-leg of the alternator. A slight dis-
placement of the beam will cause the attenuator to block a different fraction
of the beam with a resulting change in Eat. Since the instrument measures a
small difference, A, between two relatively large signals, Eg and Eat, it is
naturally quite sensitive to small changes in Eat, that are not accompanied by
a corresponding change in E„. Beam displacements that produce this kind of in-
stability can result from slight shifts in many of the different optical com-
ponents, particularly from wobble in the rotating mirror chopper. When good
instrument stability is required so that small values of V1 can be measured,
it is best to employ neutral density attenuators that have the same transmit-
tance over all of the area covered by the radiant energy beam.
In some cases, it is difficult to obtain neutral density attenuators with the
proper Tat, and a geometrical-type attenuator provides the simplest method of
obtaining a balance between E° and E°t. The attenuator illustrated schemati-
cally in Figure 1 contains two parts, an interchangeable neutral density part
and a geometrical type that can be used when needed. In order to reduce the
instability and noise produced by the geometrical attenuator, it is made so
that it attenuates near the center of the beam, rather than at an edge of the
beam. The flat, blade-like attenuator extends horizontally across the entire
center portion of the beam where the illumination is very nearly uniform. Thus,
small displacements of the beam produce only very small changes in Tat. The
amount of attenuation is adjusted by rotating the thin blade about its axis
along the length of the blade. Noise on the output, V1, produced by this geo-
metrical attenuator can be kept to no more than about 10~5, which is much less
17

-------
than we have been able to attain with combs or with attenuators that block the
edge of the beam.
The correlation efficiency C£ is defined as
A - A	T - T
c = a^s	= g>s _ at>s .	(13)
A	1 - T „
at,s	at,s
Consider the ideally correlated system in which the GFC is completely opaque at
all wavenumbers where there is any absorption by the sample. In terms of the
above quantities, this can be stated as T = 0 if Ts M and Tf 4 0. By apply-
ing this condition to Equations (1) and (6), we see that Eg is unaffected by the
addition of the sample and that Tg)S = 1. (Ag s = 0). In this idealized system,
we see that Ce = 1, the maximum value possible for a system employing an atten-
uator with Tat constant for all wavenumbers. It can be seen that Ce would be
greater than 1 if the attenuator had spectral features that were negatively cor-
related with the sample gas. This specialized case is not applicable to the
instrument described here and is not discussed further.
The correlation efficiency is an important parameter that is quite useful in
describing the performance of a GFC instrument. It is related to the spectral
structure of the gas species being measured but not to the intensity of the ab-
sorption. If Ce is between approximately 0.7 and 1, the correlation is quite
good, and it is likely that the instrument can be made to operate with good sen-
sitivity and good discrimination against other gases that might absorb in the
same spectral region. Good performance for many applications can also be at-
tained with instruments with lower correlation efficiency. Carbon monoxide is
an example of a gas with spectral structure that is well suited to GFC tech-
niques; the spectral lines are well separated, narrow and of about uniform in-
tensity over intervals several cm"-'- wide. We have designed and built CO instru-
ments with Ce > 0.9. On the other hand, much lower values of Ce are expected
for gases whose spectral structure is not "sharp" but contains portions with
intermediate absorption that is strong enough to be significant but weaker than
the strongest absorption in the interval. Reference 2 discusses the SO2 absorp-
tion near 4 p.m, which is a good example of spectral structure with a low corre-
lation efficiency.
With the instrument described in this report, either of the quantities TgjS or
Tat,s can be measured directly by blocking the attenuator leg or the GFC leg
of the alternator, respectively. The ratio of the carrier voltages Vc measured
with and without the sample gives the values of either Tg(s or Tat s depending
on which leg of the alternator is being used. From these two quantities, the
value of Ce can be determined by the use of Equation (13). The value of Ce
will, in most cases, depend slightly on the amount of the sample gas. Of most
jLnterest is the value for samples of low concentration for which Aat s and
AgjS are « 1.
Values of Ce can also be determined from measurements of V' and Aat s. This
method of measuring Ce has certain advantages when the sample absorptance is
proportional to V' and is measured directly. For AatjS 1, we can show that
18

-------
2 V'
at, s
(15)
Values of Ce can also be determined from measurements of V' along with Ag s, or
from V' along with measurements of the uncontrolled carrier voltage Vc as a
sample is added and removed. In the latter case, Vc is proportional to Eg 4- Ea(-,
and the apparent absorptance, Ag + at s of the sample measured with the combina-
tion of the attenuator and GFC is
+ A
+ at, s
^ .	(16)
If A « 1, (T „ — 1 — T ) , it follows that
at,s	' at, s	g,s
V' = A - A ,	•	(17)
at, s g + at,s
The instrument described in this report is particularly well suited for the
simultaneous measurements of V' and Ag _j_ at g. Thus Ce can be determined easily
for different GFC parameters and different Sandpasses by the use of Equations
(15), (16), and (17). Many of the data discussed in later sections of this
report were obtained in this manner.
For samples at the same total pressure, the absorption is a function of the
sample absorber thickness u given by
u (atm cm) = p (atm) t (cm) = 10 c (ppm) P (atm) *-> (cm) ,	(18)
where p is the partial pressure of the absorbing gas and c is its concentration
in parts per million volume. P is the total pressure and t is the geometrical
length of the optical path through the sample. By this definition, 1 atm cm is
equivalent to 10^ ppm meters. In some investigations dealing with samples
covering a wide temperature range, the absorber thickness is defined differently
in order to account for the change in density at different temperatures. This
difference in definition must be taken into account when comparing results of
different workers.
Another very useful parameter in describing instrument performance that depends
on both Ce and the intensity of the absorption is
= lim
u-*0
V'~i
—J (atm cm)
-1
(19)
19

-------
This quantity has been named the Pembrook factor for John Pembrook, one of our
colleagues, who has done extensive work in evaluating GFC instruments.
It follows from Equations (14) and (18) that for u sufficiently small that
As <¦'' 1 for all wavenumbers of interest,
C A
F = e2 * ?S .	(20)
We define another measurable quantity
k = 	a-'s (atm cm) .	(21)
s	u
This average absorption coefficient ks can be measured without a GFC instrument
and is related directly to the intensity of absorption by the gas; it is nearly
independent of u as long as - JZsasI t « 1. Under this condition, - llX, Tat s
= A , and
at,s
C k
F = -Vs (for A « 1) .	(22)
i.	31 ? S
20

-------
SECTION VI
FORMALDEHYDE GAS-FILTER CELL
The formaldehyde gas-filter cell acts as a selective filter for the detection
of formaldehyde and is located in the position indicated by GFC in Figure 1.
The GFC for formaldehyde is necessarily more complicated than it is for many
other gases because of the low vapor pressure of formaldehyde near room temper-
ature. Paraformaldehyde powder is placed in the GFC, and the pressure of the
formaldehyde vapor above the powder is controlled by controlling the cell tem-
perature .
The construction of the formaldehyde GFC is illustrated in Figure 3. The body
of the cell is made of stainless steel, and the optical path through the cell
is approximately 5 cm. Electrical resistance wire heats the cell,which is covered
with asbestos insulation in order to reduce the amount of heat dissipated to the
remainder of the instrument. The right-hand portion of Figure 3 shows a sectional
view of the main body of the cell; the left-hand portion of the figure is divided
into three parts. The right-hand part is a sectional view, and the center-section
is a drawing of a part of the cell body without the insulation in place. Insula-
tion is shown in place in the left-hand portion of the figure. Heating wire that
is coiled around the body of the cell is not included in the figure. The bottom
plate of the cell is sealed to the main body with a Teflon O-ring and can be re-
moved in order to add more paraformaldehyde powder. Four fiber blocks, each ap-
proximately 1.2 cm long and 1.2 cm in diameter, support the cell and insulate
it from the main baseplate of the instrument.
Cell pressures can be measured with a pressure transducer mounted directly to
the main cell body and insulated to keep its temperature close to the remainder
of the cell. A small valve, which is also insulated, makes it possible to evac-
uate the cell or to add nitrogen or any other gas that may be desired. Two
sapphire windows are used on each end; one provides a vacuum tight seal, and
the outer one provides thermal insulation for the inner window, which is in
contact with the formaldehyde vapor. The windows are approximately 42 mm in
diameter and 1.5 mm thick.
The heating coils and insulation are designed to maintain a temperature gradient
such that the inner window on each end is a few degrees centigrade above the
remainder of the cell in order to avoid condensation of formaldehyde on the
window. A small amount of paraformaldehyde powder is placed in the inside of
the bottom plate, the coolest part in contact with the formaldehyde vapor. Thus,
the pressure of the vapor is expected to be approximately equal to the vapor
pressure of the powder at the temperature of the bottom plate. Most of the heat
21

-------
Figure 3. Two drawings of the formaldehyde gas-filter cell.

-------
is provided by heating wire coiled around the outside edge of the window holders.
Some additional heat is provided by heating wire wrapped around the valve and
the pressure transducer to maintain these parts at approximately the same temper-
ature as the main body of the cell. The asbestos insulation, which is approxi-
mately 1 cm thick, around the body of the cell reduces the amount of heat loss
through the walls of the cell. A large portion of the heat flow is from the
heating coils through the body of the cell and the bottom plate to the main base-
plate of the instrument. By experimenting, we found that the four fiber support-
ing blocks provide about the optimum amount of insulation between the cell and
the main baseplate. With more insulation, there is not enough heat flow through
the cell to maintain the temperature of the baseplate sufficiently far below
the window temperature to avoid condensation on the windows. Small aluminum
blocks, which were tried in the place of the fiber blocks, did not provide enough
thermal insulation and required much more heat in order to maintain the lower
plate of the cell at the desired temperature.
The window holders are circular with insulated resistance wire placed in threads
around the outside surface. The inner window of each end is pressed against a
Teflon O-ring to provide a good vacuum seal. The metal ring that holds the win-
dow makes metal-to-metal contact with the body of the cell, and the Teflon O-ring
keeps the sapphire window from contacting the metal cell body. This type of con-
struction allows the parts to expand differently as the cell is heated or cooled
while maintaining the vacuum seal. The metal surface of the cell body does not
need to be machined as flat as if it were in direct contact with the sapphire
window. Slight irregularities in the metal would cause the window to break as
pressure was applied to it if the metal and the window were in contact. The dead
air space between the outer and inner windows provides insulation so that the
window in contact with the formaldehyde vapor is not cooled by the air. The outer
window is held in place by a silicone rubber cement.
An adjustable temperature controller maintains the temperature of the GFC at any
desired level between approximately 40°C and 75°C. The sensor for the controller
is attached to the bottom plate of the GFC near the paraformaldehyde powder as
indicated in Figure 3. The controlling unit is mounted on the underneath side
of the main baseplate and is accessible for adjustment from the end of the instru-
ment near the source. A transformer used in conjunction with the controller
limits the maximum voltage applied to the heating coils to approximately 30 Volts.
The power is turned on and off with a period of approximately two seconds. The
fraction of each period that the power is applied is proportional to the error
signal, the difference between the sensor temperature and the control temperature
to which the controller is adjusted. When the cell is being heated initially,
or when the control temperature is changed, the cell temperature will stabilize
with an insignificant amount of oscillation of the temperature. When the GFC
temperature is changed, the pressure of the formaldehyde vapor also changes, but
as much as two days may be required for the vapor pressure to stabilize after
the temperatures have come to equilibrium.
The temperature of the GFC can be monitored at five key points by thermocouple
junctions that have been bonded to the cell in the positions indicated in Figure 3.
One of these thermocouple junctions monitors the temperature on the bottom plate
adjacent to the sensor for the controlling unit and provides a check on the con-
sistency of the controller. Thermocouple junctions are also placed on the pres-
sure transducer and on the valve in order to be sure that sufficient heat is
23

-------
being applied to these parts to maintain them at approximately the same temper-
ature as the cell body. The fourth thermocouple junction is embedded in the
main body of the cell and measures the average temperature of the vapor. Leads
from the thermocouple junctions go to a five-position switch mounted on the front
panel of the instrument. The switch connects any one of the five thermocouples
to a BNC connector also mounted on the front panel. A vacuum voltmeter, or other
d.c. voltmeter with high input impedance, can be connected to the BNC fitting to
monitor the temperature at any one of the five thermocouple locations. The volt-
meter reading is proportional to the difference between the temperature of the
thermocouple junction and the temperature of the switch, which is normally not
more thaii 1 or 2 degrees centigrade above room temperature.
The relationship between the pressure of the formaldehyde vapor and the GFC tem-
perature is shown graphically in Figure 4. The data were obtained by monitoring
the pressure with the pressure transducer for different cell temperatures indicated
by thermocouple No. 1, which measures the temperature of the coolest point in con-
tact with the formaldehyde vapor. The pressures were measured after the temper-
ature had been stabilized for several hours; however, in some cases the pressure
may have not reached equilibrium. Pressures obtained from Figure 4 may be in
error by as much as 10 or 15%, but they are sufficiently accurate to determine
if the cell is operating properly. If more accurate values were reauired, it
would probably be necessary to employ a more sensitive pressure transducer and
to check its calibration at the different temperatures employed. More time
would also be required in order to ensure that the vapor pressure had reached
ecuilibrium for each temperature.
24

-------
iLH
•'f:
! -u4

a
TTTT
hit
$1
.lt
441,
li!!
i : !
-i 1 1 -
! Hi
:n
tw
mi

I i :
It'!
i*i
! : l
it

I. .
t
II"
1M
i; • i
r;!"

-T
M ;
ItH
i-H
• ~ r'
I
:r
r
\1 ¦
+1—
Wfi
lm.
i;: 1
4i-

: 111
'ti1
¦#!
!!!!
ill:
tiTf
40
50	^0	70
Cell Temperature ( C)
80
Figure 4. Plot of the pressure of formaldehyde vapor vs the GFC
temperature. The temperature corresponds to the coldest
spot in contact with the formaldehyde vapor.
25

-------
THIS PAGE LEFT BLANK INTENTIONALLY
26

-------
SECTION VII
MULTIPLE-PASS SAMPLE CELL
OPTICAL AND MECHANICAL
A view of one end of the multiple-pass sample cell from the back side of the
instrument is shown in Figure 5. The cell is supported by two V-shaped fiber-
glass blocks that insulate it from the main baseplate. A portion of one of the
fiberglass blocks can be seen in Figure 5. Adjustable metal straps hold the
cell down against the blocks. A stainless steel tab welded to the underneath
side of the cell body and bolted to the block nearest the window end of the
cell holds the cell in the proper position. The cell body can slip endwise
relative to the other fiberglass block to account for expansion when the cell
is heated. The transmitting optics of the 1^0 monitor seen on the top of the
cell are discussed in Section X.
Figure 6 illustrates the manner by which the gas enters the sample cell. Eight
small holes drilled through the wall of the sample cell are covered by a hollow
ring that is welded in place around the outside of the cell body. The sample
gas enters a pipe fitting that is welded into the hollow ring and flows around
the inside of the ring to enter the cell through the eight small holes. By
introducing the gas through several holes spaced around the cell, turbulence
produced by the incoming gas is greatly reduced, and the beam of infrared energy
is more stable than it would be if the gas entered through a single large hole.
The opposite end of the cell has a similar arrangement for the gas to exit from
the cell. Both ends are essentially identical so that the direction of flow
can be reversed. The inlet and exit ports are placed close to the ends in order
to reduce the amount of stagnant air in the ends of the cell as the sample flows
through. Our tests indicate that with the sample cell adjusted to 40 passes,
no serious problem is created by turbulence when the sample is flowing fast
enough to flush out essentially all of the sample in one minute. A pipe fitting
similar to those used on the inlet and outlet ports is welded directly to the
wall of the cell near its center. A vacuum pump connected to this fitting will
evacuate the cell more quickly than if the gas were pumped through the small
holes in the wall near the inlet or outlet port. If the fitting is not connected
to a vacuum pump, it can serve as a fitting for a gauge as shown in Figure 2.
Figure 6 also shows an optical diagram of the bypass optics assembly that can
be installed so that the monitoring beam does not enter the multiple-pass sample
cell. Operation with the bypass optics in place might be desirable for studies
in which a sample cell is no longer than a few centimeters. In this case, the
sample cell could be placed in the beam between mirror N4 and the bypass optics
assembly. Mirrors Bl, B2 and B3 are permanently mounted to a block that fits
into position and bolts directly to the main baseplate. Mirror XI, whose normal
27

-------
N>
00
Transmitting Optics
of H^O Monitor
Fiberglass
Support
Thermocouple
Photographic view of one end of the multiple-pass sample cell from the back side of the
instrument.

-------
. ; (SMPLE IN)
A
Figure 6.
Sample inlet and optical diagram of the bypass optics assembly. Bypass mirrors Bl,
B2, and B3 are mounted on a block that fits into place when mirror XI is removed.

-------
position is indicated in Figure 6 by a broken line, is removed before the by-
pass optics assembly is put in place. Small pins in the baseplate guide the
bypass optics assembly so that it can be removed and replaced easily with little
alignment necessary.
As indicated in Figure 1, an image of the source is formed inside the multiple-
pass sample cell adjacent to mirror C2. As the beam leaves the multiple-pass
sample cell, a similar image on the opposite side of C2 acts as a source that
is imaged by mirror X2 onto the entrance slit of the grating section. When the
bypass optics are in place, the beam that would normally enter the multiple-
pass cell is intercepted by mirror Bl, and the image of the source is formed
near mirror B3. This image then acts as a source that is imaged on the slit
of the grating section. Spherical mirror B3 acts as a field mirror and greatly
reduces the vignetting that would occur if it were flat.
Two sets of windows are supplied for the multiple-pass sample cell. Sapphire
windows 38 mm in diameter and 2 mm thick are used for wavelengths shorter than
approximately 5.5 vun. Windows of NaCl are used for longer wavelengths out to
approximately 15 u.m; they are not used unless they are required because they
are hygroscopic and must not be exposed to humid air over extended periods of
time. O-rings not shown in the figures provide a vacuum seal between the win-
dows and the end flange. Metal clamps hold the windows and the field lenses
Ll and L2. Field lenses made of CaF2 can be used from the ultraviolet to ap-
proximately 8 um; a set of NaCl lenses are provided for the longer wavelengths.
A vacuum seal between the end of the multiple-pass sample cell and the remov-
able end flange is provided by O-ring gaskets made of buna-N rubber.
The optical principles of the multiple-pass sample cell, first described by
White,^ can be explained with the aid of Figure 7. The upper portion of the
figure shows mirror C2 as viewed from the opposite end of the multiple-pass
cell. Entrance window Wl and exit window W2 are indicated. The large circle
corresponds to the 10.2 cm inside diameter of the cell body. The image indi-
cated by the 0 is the one formed near mirror C2 by mirror N4 of the entrance
optics (see Figure 1). The beam continuing from this image strikes mirror Cl,
the center of curvature of which is near the front surface of mirror C2 in the
position indicated by an X in the upper portion of Figure 7. Image 0 is at a
distance from mirror Cl equal to the radius of curvature of Cl; therefore,
image No. 2 is formed in the same plane as image 0 with the two images sym-
metrically placed around X, the center of curvature of mirror Cl. Mirror C2
is also spherical, and its center of curvature is between mirrors Cl and C3 so
that C2 forms an image of Cl on C3, and vice versa.
Energy from image 2 is reflected to mirror C3, which in turn, forms another
image on mirror C2 at the position indicated by 4. Images 2 and 4 are symmet-
rical about the center of curvature of mirror C3 indicated by the circle ad-
jacent to the X in Figure 7. The beam continues back and forth in this manner
4. White, John U. Long Optical Paths of Large Aperture. Journal of the
Optical Society of America. 32:285-288, May 1942.
30

-------
Figure 7. Multiple-pass pptical system in the sample cell. The upper por-
tion shows positions of the images on mirror C,2 when the cell is
adjusted to 16 passes. The positions of the screws used to ad-
just mirrors CI and C3 are indicated in the lower portion.
31

-------
forming images on mirror C2 as shown. The number adjacent to each image indi-
cates the number of passes the beam has made at the time the image is formed.
In the arrangement shown, the beam misses mirror C2 after 16 passes and exits
through window W2.
After the sample cell is initially adjusted, the position of mirror Cl is left
fixed so that it forms image 2 close to the position shown. The number of
passes is changed by adjusting the azimuth angle of mirror C3 to move its
center of curvature horizontally. When the centers of curvature of mirrors Cl
and C3 are close together, the images on mirror C2 are also close together and
the beam makes several passes before it emerges through the exit window. After
the optical system has been adjusted to give the desired number of passes, a
small tilt adjustment of mirror C3 may be required in order to form the exit
image at the proper height. After a tilt adjustment has been made, a final
azimuthal adjustment may be required because the azimuth adjustment is the most
critica1.
Mirrors Cl and C3 are adjusted from outside the cell by rotating the screws that
pass through the end plate. Jam nuts on the screws lock the screws solidly in
place. If the cell is to be used at less than atmospheric pressure and leak-
tight seals are required, the screws are potted in silicone rubber after the ad-
justments are completed.
TEMPERATURE CONTROL
Electrical heating wire wrapped around the outside of the cell body makes it
possible to heat the cell to any desired temperature up to approximately 55°C.
Heating the cell avoids condensation of H2O vapor when studying raw automotive
exhaust, or any other samples with moderately high concentrations of H2O. One
piece of heating wire is wrapped around the cylindrical body of the cell with
adjacent coils closer together near the ends than in the middle to help account
for heat loss from the ends. A power resistor bonded with good thermal contact
to each endplate supplies additional heat to the ends so that there is less than
2°C variation in the temperature throughout the length of the cell. Fiberglass
insulation with metallic foil coating is wrapped around the cylindrical body and
covers the heating wire.
A temperature controller senses the temperature of the cell wall near the center
and maintains it at a pre-set level. The dial of the controller can be seen near
the right-hand end of the instrument in Figure 2. Five thermocouples sense the
temperature of the cell. One is connected to each endplate, and the other three
are connected to the cell wall, near the center and approximately 25 cm from
each end. A five-position switch connects each thermocouple to a BNC fitting
on the front panel where it can be coupled easily to a voltmeter to measure the
temperature at any of the five positions.
32

-------
SECTION VIII
GRATING SECTION
The grating section acts as a conveniently adjustable bandpass filter that can
provide virtually any desired spectral bandpass from approximately 0.3 p-m to
more than 11 vim. The beam of radiant energy enters the grating section through
the entrance slit Si on which an image of the source is formed. A wide selec-
tion of slits are available, ranging in width from approximately 0.1 mm to
2.4 mm. The slits are cut by a laser in small pieces of stainless steel shim-
stock, approximately 0.08 mm thick by 0.95 cm wide and 4.75 cm long. The slits
can be interchanged easily and fit into a guide that holds the center of all of
the slits at the same place. Each slit is approximately 12.5 mm in height and
its center is approximately 9.5 cm above the baseplate.
After the beam of radiant energy enters slit SI, it is reflected from a small
diagonal mirror Gl to a spherical collimating mirror G2. The optical path from
slit Si to mirror G2 is equal to approximately 60 cm, the focal length of the
mirror. Therefore, the beam of energy directed from mirror G2 to the grating
GR, is nearly collimated. Depending on the angular position of the grating,
certain wavelengths of energy are diffracted in a near parallel beam to mirror
G3, which has the same focal length as mirror G2. The radiant energy incident
on mirror G3 is dispersed, with different wavelengths incident on the mirror at
different angles. Energy of a single wavelength is focused by mirror G3 onto
the grid assembly, forming an image of the entrance slit.
Figure 8 shows the central ray of a beam of monochromatic energy passing through
the grid of the grid assembly. Before passing through the grid, the energy beam
first strikes mirror G4, which is tilted 45° from the vertical. Mirrors G4 and
G5 are normal to each other so that the central ray returns to mirror G3 parallel
to, but displaced upward from, the ray before it strikes mirror G4. The grid
may contain more than one opening although the one illustrated contains a single
opening and passes one narrow spectral i. iter/al.
The central ray from the center of the slit to the center of the grating travels
in a plane parallel to the baseplate of the instrument until it strikes mirror
G4. The ray from mirror G5 to G3 is also parallel to the baseplate, but after
striking mirror G3 it is directed slightly downward to the center of the grating.
From there it continues downward until it strikes mirror G2 at a point approxi-
mately 1.8 cm below the point where it was incident on mirror G2 in the incom-
ing beam. Except for the slight change in the vertical angle, the returning
beam strikes the grating in exactly the reverse of the path on which it entered.
Thus the grating undisperses the beam of energy passed by the retroreflector-
grid assembly so that the image formed at the exit slit S2 is undispersed.
33

-------
Figure 8, Optical diagram of the retroreflector-grid assembly.
The center of slit S2 is 7.7 cm above the baseplate, approximately 1.8 cm below
the center of the entrance slit Si. Because an image of slit SI is formed at
slit S2, only one of these slits is required to provide the desired spectral
resolution. Slit S2 is intentionally made somewhat wider than slit SI; the ef-
fective part of S2 is only that portion conjugate to the narrower SI. Using
the slightly oversize slit S2 instead of a much larger opening reduces scattered
energy reaching the detector; having it oversized does not degrade the resolution
and makes the positioning of the slits much less critical than if both slits
were the same width.
The method by which the radiation is dispersed and the selected portion is re-
covered and imaged is described in Reference 3. This paper describes the method
that uses a prism as a dispersing element; however, the obvious application to
a grating and to systems with a "tailored" bandpass are pointed out. Decker^
has also described an application of the method to Hadamard spectroscopy.
5. Decker, J. A. Jr. Experimental Realization of the Multiplex Advantage
with a Hadamard-Transform Spectrometer. Appl. Opt. _10, No. 3, 510-514, 1971.
34

-------
SELECTION OF SPECTRAL BANDPASS
The optical diagram in Figure 9 is useful in deriving and understanding the
relationship between the angular position of the grating and the wavelength
passed by the grating section. If the entrance slit SI is very narrow, the
horizontal projection of the beam going from mirror G2 to the grating is essen-
tially parallel. The ray AO in Figure 9 can be considered the central ray of
this parallel beam; the central ray OB corresponds to the central ray of the
monochromatic beam that strikes mirror G3 at such an angle that it is focused
on the center of the grid in the retroreflector.
A
( From mirror G2 )
B
( To mirror C3 )
Figure 9. Optical diagram defining angles that relate the grating
position to the wavelength of energy passed by the grating
assembly. Angles a, 3, and 9 are measured from line ON
and are positive in the counter-clockwise direction.
Angle 0 is always positive and is independent of the rota-
tion of the grating.
Reflecting surface
of grating
35

-------
Consider the situation with the grating rotated about a vertical axis through
point 0 from the typical operating position to the position where its normal
ON coincides with the line 0N°. In this case, both the incoming and reflected
rays form the angle 0 with the normal to the grating. This position of the
grating is called the "zero-order" position, and some of the energy of all wave-
lengths is reflected parallel to OB as if the grating were a plane mirror. In
typical operation, the grating is rotated from the zero-order position by an
angle 0 so that its normal is in the position indicated by the line ON in
Figure 9. Four angles are defined by the figure and are inter-related by the
following equations:
a =
= fl + 0.
(23)
3
= 0-0.
(24)
a + 3 =
= 29.
(25)
Ot - 3 =
= 20.
(26)
With the grating in the position indicated in Figure 9, the only wavelengths
of energy diffracted parallel to the line OB are those for which there is con-
structive interference. These wavelengths pass through the center of the grid
and are given by:
mX = 2D cos 0 sin 9, or	(27)
m>. = D(sin a + sin 3).	(28)
D is the distance between the adjacent rulings on the grating, and m is an
integer. The first order wavelength is the longest one passed and corresponds
to m = 1; higher orders correspond to m = 2, 3,	 etc. Note that the zero-
order position of the grating corresponds to 9 = 0 and m = 0; in this position,
3 is negative and sin & = -sin P. In the following discussion m = 1 unless
otherwise indicated.
The wavelength \c of the radiant energy passing through the center of the grid
is adjusted with the screw mechanism shown in Figure 1. As the screw rotates,
the nut moves along the axis of the screw and drives the arm on the grating
assembly, changing angle R while 0 remains fixed. A linear scale on the side
of the screw mechanism makes it easy to measure the displacement of the nut
from the end of the screw opposite the screw handle. One complete revolution
of the screw produces a displacement of 1 mm, or one grating count. A scale
on the outside surface of the drum to which the handle is attached makes it
possible to read the grating counts directly to 0.01.
From the calibration curves in Figure 10 it is possible to determine the wave-
length passed by the grating from the grating counts without determining the
values of any of the angles. The values of wavelength and wavenumber, written
in parentheses in the left-hand panel, correspond to the grating with 150 lines/mm.
36

-------
180
160
140
120
100
80
60
40
20
0
(
WAVE NUMBER ( cm )
(2000) (1500) (1250) (1000)
4,000	3,000 2500	2000
WAVENUMBER (cm )
20,000	10,
I I I I I I I	1	1	1	1
(7) (8) (9) (10) (11)
WAVELENGTH (micrometers)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 O.o 1.
WaVEL£NGTH (micrometers)
re 10. Calibration curves relating the grating drive counts to wavelength and to
wavenumber. The numbers in parentheses correspond to the 150 line/mm grati

-------
The other values, not in parentheses, represent the 300 line/mm grating. The
right-hand panel of Figure 10 is the corresponding plot for the 1800 line/mm
grating used in the uv, visible, and near infrared. Each grating is blazed at
the following wavelength: 150 line/mm at 6 p,m, 300 line/mm at 3 urn and 1800
line/mm at 0.50 ytm.
It is apparent from Figure 9 and Equation (28) that energy of wavelength Xc + o,\
will not be diffracted with constructive interference through exactly the same
angle (3 as will energy of wavelength \c, which passes from the center of the
entrance slit to the center of the grid. The difference 6|3 between the angles
at which the two nearby wavelengths leave the grating can be determined from
Equation (28). Because the central rays for the beams of the two different
wavelengths are incident on mirror G3 at different angles, the two wavelengths
are focused at different positions in the plane of the grid. The distance 6g
between the two images of the two different wavelengths in the plane of the
grid is given by
6g = f (6P),	(29)
where f is the focal length of mirrors G2 and G3. The dispersion, oX/bg^ in
the plane of the grid can be calculated from Equations (28) and (29) by holding
sin a constant and differentiating Equation (28) with respect to 3. If we as-
sume a very narrow entrance slit, the result of this dispersion calculation
provides the distance between two different wavelengths in the plane of the
grid. In many cases it is preferable to express dispersion in wavenumbers per
millimeter, rather than in micrometers per millimeter. When v is expressed in
cm"! and \ in micrometers,
Sv = - 5X xl0\	(30)
xz
The solid curves in Figure 11 relate the dispersion in cm per mm in the plane
of the grid to wavenumber and wavelength for first-order radiation. If the phy-
sical width Wn of the entrance slit is much narrower than the width of an
opening in the grid, the spectral interval 6v passed by the grid is given by
Sv (cm ¦*") = W (mm) (cm ^ per mm). (for W « W )	(31)
§	§	n g
Consider the image of the entrance slit that is formed at the plane of the grid
for monochromatic energy with a perfect optical system;, i.e., one for which
there are no aberrations and the diffraction-limited image size is very small
compared to the physical slitwidth. The spread 6ctf in the beam incident on the
grating is
6a = W /f.	(32)
38

-------
WAVELENGTH ((im)
.7 .6 .5 .4	.3
LO




-¦HLT
W
sdnr
jf: 1800
¦ - 1 i n P q / m

-• L=.






'=
: n



w
•:.F=
hit?.
-•
"i 'l







-sk


/¦


. .Tt "
1
p--=





-7=K
~?rlr=-
Pri'L
—•t r_
-:-jr

—


lL
p



"a

--j.
: 5-i

r-r

•:-V-
",|'j
/

-
&*&¦





: -~z-_



¦ .\V

r T :

~"F
k~-
~.i~r
' i':
=5
HH-—



/i..

.
;=
k-
.a'i.

=hr:*


iTI'-L-
•jn

J1-

: :

•"
p
i:;
• n
~ j'/
• -U

hMb;


'/Is::
_7:I .
-M-j-

" i. j
-


.hTiHf
=?===
E==j


"1 "1~ "
-

- L!:~


":".C
prf--

Si

—\ "




r-
hh

UJ"-




_j L: -»•

Liz :t
r::.
.2
=rb

:=¦£=!
—.:ar

•r k

—-r
f?
?:
f'Iht


lis—
£-r±
:;r::
7=fT
rr.
:~-
~i]V'


~-f=
~Th


7-


ifp
==l5ri

=H~
—riii
¦=~
.d_

"H?: _• •
_™v

B-j
20000
1.0
.9
30000
WAVENUMBER (cm )
WAVELENGTH (,,m)
5.0 4.0 3.5 3.0 2.5
2000
3000
-1>
4000
WAVENUMBER (cm )
WAVELENGTH (//_m)
12 10 8
1000 1200 1400 1600
WAVENUMBER (cm"1)
Figure 11. Curves of 5v/6g and cos a/cos (3 vs wavenumber for the three gratings.
39

-------
By differentiating the quantities in Equation (28) we can show that for con-
stant A
6(3
cos &
cos P'
(33)
Thus, the width of the image at the grid plane = Wn x 63/6a. The negative sign
in the equation refers to the relative directions in the changes of the angles.
When Wn is not negligible in comparison to Wg, the width of an opening in the
grid, the "transmission function" of each opening in the grid is "trapezoidal"
and is represented by Figure 12. In most applications of the instrument des-
cribed in this report, Wn cos c/cos 3 Wg. The maximum effective Wn is limited
to approximately 1.6 mm, the width of the image of the source on slit Si. Values
of cos a/cos 0 can be obtained from the curves of Figure 11.
The aberrations in the optics of the grating assembly produce a blurring of less
than 0.1 mm, and the diffraction-limited image width is also less than 0.1 mm,
except for X greater than about 8 |j,m. Therefore, the simple method of deter-
mining the transmission function just described is approximately valid if either
Wn or Wg is greater than about 0.1 to 0.2 mm. For narrower slits or grid open-
ings, the spectral bandpass is somewhat greater than would be calculated by the
simple method that ignores aberrations and diffraction limitations.

-------
SECTION IX
ELECTRONICS AND PROCESSING OF DETECTOR SIGNAL
The electronics process the signal from the photodetector and produce a dc
voltage that is proportional to the concentration of the gas species being
measured. The relationship between the output voltage and the gas concentra-
tion is determined by calibrating the instrument with samples of known concen-
tration. During the normal mode of operation, the spectral interval passed by
the instrument remains fixed, and the sample gas is placed in the sample cell.
Zero-readings are made by evacuating the sample cell or by filling it with a
non-absorbing gas. In addition to the normal mode of operation, the instrument
can be used to measure the transmittance of radiant energy by gases or other
materials, such as windows. This can be done with the wavelength fixed, or
spectral curves of transmittance can be obtained while the wavelength is being
scanned continuously.
The electronic components are contained in the compartment indicated in Figure 1.
The preamplifiers and amplifiers are mounted on circuit cards and are easily
accessible by removing the cover of the electronics compartment. The switches,
controls, panel meter, and recorder output jacks are mounted on the removable
cover. All of the electronic circuits for the H2O monitor described in Section X
are also included in the same electronics compartment.
A block diagram of the electronics used to process the photodetector signal is
shown in Figure 13. The 360 Hz component of the signal from the photodetector
serves as the carrier for the 30 Hz component that is to be measured. The de-
tector signal resulting from any 30 Hz variation of radiant energy on the de-
tector that is not also chopped at the carrier frequency is rejected by the
electronics. The electronics consist basically of two synchronous demodulators
ia series. The first demodulator operates at the carrier frequency, and the
average dc component is proportional to the amount of radiant energy chopped
at 360 Hz. If the chopped energy incident on the detector is also modulated
at 30 Hz, the 30 Hz component of the signal output of the 360 Hz synchronous
demodulator is proportional to the modulation. This 30 Hz component then passes
through a series of gain controls and amplifiers to the second synchronous
demodulator that operates at 30 Hz. The output of the second synchronous de-
modulator is then proportional to the 30 Hz modulation, and thus to the con-
centration of the gas species being measured.
When H2O in the sample interferes with the gas species being measured, the
30 Hz signal also contains a component due to the H2O. This interference is
accounted for by an H2O monitor described in Section X that produces a dc
signal proportional to the H2O concentration. An adjustable portion of the
41

-------
From
Photodetector
Figure 13. Block diagram of electronics.

-------
output of this H2O monitor is fed into the "H2O Correction" circuits in the
place indicated in the block diagram of Figure 13. The signal output from the
H2O correction circuits has been automatically corrected for any interference
by H2O in the sample and is proportional to the concentration of the gas species
being measured. The electronic components used to adjust the portion of the
H2O monitor output that is fed into the correction circuits are discussed as
part of the H2O monitor in Section X.
A five position switch makes it possible to select an electronic time constant
of 0.3, 1, 3 or 10 sec. When the switch is in the fifth position, marked "off",
no additional damping is applied to the signal and the effective time constant
is approximately 0.1 seconds. The main signal output can be determined from a
panel meter or from either of two output jacks. A signal that corresponds to
a full-scale reading of the panel meter produces a 10 volt dc signal on one of
the output jacks. The corresponding full scale reading for the other output
jack can be adjusted from less than 1 mv to 10 v. Thus, the instrument can be
used with a wide range of recorders or meters, provided the input impedance is
sufficiently high.
As discussed in Section V, it is desirable that the signal output during the
normal mode of operation be proportional to the ratio, Va/Vc = V1. The instru-
ment calibration relates the concentration of the gas being measured to V', and
it is desired that this relationship be independent of source brightness, de-
tector sensitivity, etc. During the normal mode of operation Vc is maintained
at a constant value by the AGC circuits. If, for example, the signal from the
photodetector decreases because of dirt on a window or a decrease in source
brightness, the gain of the gain-controlled amplifier increases to maintain Vc
constant. The increase in gain also increases the amplification of both the
360 Hz signal and the 30 Hz signal by the same factor. Therefore, V' is kept
directly proportional to the output Va and maintains a constant relationship
with the gas concentration. The AGC circuits can account for changes in the
360 Hz component of the signal from the preamplifier by approximately a factor
of 4. When the AGC switch is in the off position, Vc is not maintained con-
stant but can be adjusted with a manual gain control.
Both reference pickups consist of a light-emitting diode (LED) and a small
phototransistor. The 360 Hz pickup is mounted near the high-speed chopper so
that the output is modulated as the spokes of the chopper pass between the LED
and the phototransistor. The 30 Hz pickup senses the light from the LED that
is reflected by the mirror chopper. The amplifier for each reference pickup
is tuned and contains an adjustable phase shifter to produce a "clean" signal
with the proper phase for the synchronous demodulators.
The zero-balance assembly electronically accounts for any misbalance between
the two beams passing through the two legs of the alternator when there is no
absorbing gas in the sample cell. This electronic assembly would not be re-
quired if it were practical to maintain a perfect optical balance between the
two legs. Approximately + 10 percent variation in the transmittance through
either of the legs of the alternator can be accounted for by this electronic
zero-balance assembly. The assembly changes the amplification ahead of the
360 Hz synchronous demodulator during the two halves of the alternator cycle.
A signal from the 30 Hz reference pickup switches the zero-balance assembly
between the two different gains with the proper phase. The difference in the
43

-------
gains between the two halves is adjustable by the "zero" potentiometer on the
top of the electronics compartment.
Five ranges of sensitivity are provided. The gain in each range can be adjusted
independently over a factor of approximately 5. A five-position switch mounted
on the top of the electronics compartment makes it convenient to change from
one range to another. An additional potentiometer is also provided to change
the gain of all 5 ranges simultaneously by the same factor. At the time the
instrument was delivered to EPA, the range adjustments, also called span ad-
justments, were made so that full-scale output corresponded to V' = 0.003,
0.01, 0.03, 0.1 and 0.30.
Three additional output jacks are provided on the cover of the electronics com-
partment. One output jack, labeled "preamp", makes it possible to monitor the
electronic signal from the preamplifier before it passes through the tuned
filters or amplifiers. The output of the 360 Hz demodulator can also be moni-
tored from the output jack labeled "carrier". The transmission of gas samples,
attenuators, windows, or other optical components can be measured easily by
monitoring the output from this recorder jack while the AGC switch is in the
off position. Spectral curves of transmittance of gases can be obtained by
recording the signal from this jack while the grating is being scanned. The
360 Hz reference signal is also available from another output jack. Having
this signal readily accessible is convenient if a separate synchronous demodu-
lator is used to scan spectra or make other transmission measurements.
A four-position switch mounted near the panel meter makes it possible to use
the same meter to monitor other quantities besides the signal output. The 30 Hz
reference signal and the average carrier voltage Vc can be checked with this
meter to see if they are in the right operating range. When the instrument is
operating in the normal mode with the AGC switch turned on, the carrier signal
produces approximately 50 percent of full-scale reading of the panel meter.
While the selector switch is in the "signal-in" position, the meter reads the
level of the signal going into the 360 Hz synchronous demodulator.
44

-------
SECTION X
h2o MONITOR
OPTICAL
The H2O monitor produces a dc voltage that is proportional to the concentration
of H2O in the multiple-pass sample cell. The output voltage is displayed on a
panel meter and can be fed into the main electronic circuit to automatically
account for any interference by H2O in the sample. The H2O concentration is
determined by comparing the transmission of the sample at two wavelengths,
one in the region of strong H2O absorption near 1.9 urn and one in a nearby
region of weak absorption.
Two views of the optical layout of the H2O monitor are shown in Figure 14. A
small tungsten iodide bulb serves as a source of infrared radiant energy. The
beam is interrupted at 360 Hz by a small, 6-bladed chopper that is rotated at
3600 rpm by a synchronous motor. The position of the transmitting optics on
the multiple-pass sample cell is shown in the left-hand portion of Figure 14
and in the photograph in Figure 5.
After the beam enters the multiple-pass sample cell through window W3, it is
reflected by a small flat mirror Hi to spherical mirror C2, which is shown in
Figure 1 but not in Figure 14. Mirror C2 directs the beam back to another
flat mirror H2 and down through the exit window W4, which is mounted on the
underneath side of the sample cell. An image of the source is formed near the
small flat mirror H3, which is tilted 45° from the vertical and directs the
beam to the receiving optics that include the detectors and filters. Window
W3 is a small plano-convex lens that enlarges the image of the source near
mirror H3. Windows W3 and W4 are bonded to the stainless steel body of the
sample cell with silicone rubber cement. Mirrors Hi and H2 are each bonded with
epoxy cement to a small aluminum block that is bolted to the cell wall.
Detectors A and B have photoconductive PbS elements, 4 mm x 4 mm, that are
mounted on a small aluminum block as shown in Figure 15. Lens L4 forms images
of mirror C2 on the detectors. Filter A acts as a dichroic beam-splitter,
transmitting a narrow spectral interval in the region of strong H2O absorption
near 1.9 |j,m. The wavelengths not transmitted by filter A are reflected to
filter B, which transmits a narrow spectral interval of weak H2O absorption.
The gains of the preamplifiers are adjusted so that the 360 Hz signals from
the two detectors are equal when no absorbing gas is in the sample cell. It
is apparent that the addition of H2O to the sample cell would reduce the amount
of energy incident on detector A more than that on detector B. The resulting
difference in the detector signals is proportional to the H2O concentration.
45

-------
Figure 14. Two optical diagrams of the monitor

-------
(From
Mirror H3)
Filter A

,Detector B
/
•Filter B
Detector A
Electrical Terminals
(2 for each detector)
Small variations in the temperatures of the detectors can produce drift in the
zero-reading of the monitor, making it less stable than monitors that employ
only one detector. However, we have found from experience with similar instru-
ments that this 2-detector type of monitor is quite adequate for the present
purposes. Ordinarily the correction for H2O interference in the main instru-
ment channel is small so that the size of the correction signal from the H2O
monitor need not be highly accurate. By employing two fixed filters and two
detectors, it is possible to keep the optics relatively simple without any
moving optical components other than the simple chopper.
SIGNAL PROCESSING
The bias voltage for detectors A and B is provided by a + 15 v dc supply used
with the main electronics. A separate preamplifier amplifies the output of
each detector, and the outputs of both preamplifiers are fed into a difference
amplifier. When the sample cell is free of any H2O, the monitor is zeroed by
adjusting the outputs of the preamplifiers so that they are equal, producing a
zero input to the difference amplifier. The addition of H2O to the sample cell
creates a misbalance between the signals from the two preamplifiers and results
in a signal from the difference amplifier. This signal passes through a tuned
amplifier and is then demodulated to produce a dc voltage that is proportional
to the H2O concentration. A reference signal for the synchronous demodulator
is obtained from the output of the preamplifier for detector B.
A variable-gain amplifier makes it possible to adjust the "span" so that full-
scale readings of the panel meter can be made to correspond to 1^0 concentra-
tions between approximately 1 percent and 10 percent. An output jack connected
in parallel with the panel meter has an adjustable attenuator that can be used
47

-------
to vary the signal at the output jack that corresponds to full-scale reading
of the m~ter from less than 1 mv to 10 v dc.
An adjustable fraction of the output of the variable-gain amplifier is fed into
the main electronics as indicated in Figure 13 to correct the signal output for
interference by H2O in the sample. The "correction factor" is adjusted to con-
trol the correction signal so that it is of the proper size and polarity.
Interference by H2O is positive for some gases and negative for others; thus
it is necessary that the polarity of the correction signal be changeable. In
order to adjust the correction factor, N2 or argon is bubbled through water and
through the sample cell to produce an H2O concentration approximately equal to
that expected in the samples to be studied. The potentiometer used to change
the correction factor is then adjusted to cancel any signal from the main elec-
tronics that results from interference by the H2O. With the H2O correction ad-
justed in this manner, the proper correction signal is applied for all R2O con-
centrations over a wide range that includes the concentration used in making
the adjustments.
If the H2O concentration is less than approximately 2 percent, the relationship
between the concentration and the output of the H2O monitor is nearly linear.
A similar linear relationship also exists between the output of the main elec-
tronics and the concentration of the gas species being measured when the con-
centration is low. When the concentrations of H2O and the gas being measured
are both in the linear region, the correction factor is valid for all concen-
trations when it is adjusted as described above. However, because of the slight
non-linearity in the responses, a single adjustment of the correction factor
made at one H2O concentration may not be valid for concentrations that are quite
different. Therefore, it is best to follow the above adjustment procedure and
use H2O concentrations similar to those expected in the samples. When measuring
formaldehyde concentrations, the interference by	is relatively small so that
it can be corrected adequately for all H2O concentrations up to 4 percent with
a single setting of the correction factor. A different setting may be required
if higher H2O concentrations occur in the samples.
An automatic gain control circuit is employed so that the amplified difference
signal is nearly independent of slight changes in such things as source bright-
ness or dirt on windows that would change the amount of energy incident on both
detectors by about the same factor. The AGC maintains the signal from the pre-
amplifier for detector B at a constant level. For example, if a change in the
optics occurs that would ordinarily decrease the detector signal, the bias volt-
age is increased automatically to compensate for the decrease in radiant energy.
The bias voltages for both detectors are increased by the same factor; therefore,
to a good approximation, the change in bias voltage increases the outputs of both
detectors by the same factor. It follows that a given output of the difference
amplifier represents a certain fractional difference in the outputs of the two
preamplifiers. This difference can then be related to the H2O concentration and
is nearly independent of small changes in source brightness, dirt on windows,
etc.
48

-------
SECTION XI
RESULTS OF FORMALDEHYDE TESTS
SPECTRAL BANDPASS SELECTION
By investigating a high-resolution spectrum of formaldehyde provided for us by
Dr. Philip Hanst of EPA, we concluded that enough strong spectral structure
existed in the 2700 - 3000 cm"^ region for the efficient operation of a gas-
filter correlation spectrometer for this gas. It was possible to select several
narrow spectral intervals within the 2700 - 3000 cm"-'- region that might provide
better performance than the full 300 cm-'- wide interval. We decided to assemble
the components and use the instrument to determine the optimum of the several
promising spectral intervals for use in the normal GFC mode. The two most im-
portant factors in the choice of the spectral interval are sensitivity and dis-
crimination against other gases that occur in automotive exhaust.
Among the possible interfering gases are hydrocarbons and H2O. Although H2O
absorption is relatively weak in this spectral region, a small amount of inter-
ference by this gas is almost certain because of its high concentration (typically
27c-37o) and the long sample path lengths required to produce adequate sensitivity
to formaldehyde. Interference by H2O was not considered in the final selection
of the bandpass because it was assumed that the H2O monitor with its correction
circuitry could account for the small amount of interference.
Figure 16 shows spectral curves of transmittance for formaldehyde and for five
other hydrocarbons that absorb in the same spectral region and may be expected
to interfere. All of the curves were scanned with the GFC spectrometer without
the gas-filter cell (GFC) in place. The grating counts scale at the top of the
figure refers to the position of the grating screw as discussed in Section VIII.
The formaldehyde sample was contained in the heated GFC discussed in Section VI.
All of the hydrocarbon samples represented in the lower two panels were pure and
at 1 atm pressure in a 1 cm sample cell. Acetylene, another hydrocarbon found
in automotive exhaust, is not represented in Figure 16 because it does not absorb
in this spectral region. The 0.8 cm~l spectral slitwidth used in scanning the
spectra is too wide to show all of the structure in the spectra of formaldehyde,
methane, ethylene, and ethane. Little additional structure would be observed
in the spectra of propane and butane if narrower slits were employed because the
lines are so closely spaced that they blend together when the sample pressure
is greater than a few-tenths of an atm.
The hydrocarbon samples represented in Figure 16 contain much more absorbing
gas than would be expected in a typical sample of automotive exhaust gas, but
the curves point out clearly that excessive interference might be expected if
49

-------
OR A riNCi C.OUN \'S
9#	97	96	95	94	9?	92	91	;0	M i	tth
2800	29O0
WAV ENUMBEK (cm" 1)
Figure 16. Spectral curves of transmittance for formaldehyde and
five hydrocarbons. The smooth, upper curve in each panel
corresponds to 100% transmittance and was scanned with
the sample cell evacuated. The spectral slitwidth is
approximately 0.8 cm"^,
50

-------
the bandpass extended above approximately 2830 cm"^. Interference tests were
made with the GFC instrument operating in its normal mode with several different
spectral bandpasses. As expected, troublesome interference by some of the hydro-
carbons was observed when the bandpass extended to the higher wavenumbers.
Methane is usually present in automotive exhaust at a higher concentration than
any of the other hydrocarbons investigated. Thus it is important that inter-
ference by this gas be kept as low as is practical. It is likely that the con-
centration of methane in the samples will also be measured with a separate in-
strument; therefore, it will be possible to account for a small amount of inter-
ference by this gas from a predetermined value of the discrimination ratio.
By conducting several tests for different spectral intervals by the methods des-
cribed below, we concluded that the optimum spectral bandpass is centered at
approximately 2803 cm~l. A trapezoidal slit function such as the one illustrated
in Figure 12 determines the spectral bandpass. The approximate widths of the
"top" and the "bottom" of the trapezoid are 11.4 cm" 1 and 18.6 cm"'-, respectively.
This relatively narrow interval includes enough radiant energy that detector noise
is not the major source of error, as would be the case if a much narrower inter-
val were employed. Wider spectral intervals produce more interference by the
five hydrocarbons represented in Figure 16. Most of the wider intervals inves-
tigated also produced a lower Pembrook factor, F, than did the selected interval.
As discussed in Section V, the lower Pembrook factor would increase the uncertainty
caused by turbulence, slight optical misalignment, and other types of optical noise.
TEST PROCEDURES AND PERFORMANCE
Formaldehyde samples used in the tests were contained in a small, heated cell
that was constructed similar to the GFC. The by-pass optics were put in place
as illustrated in Figure 6 so that the beam did not pass through the multiple-
pass sample cell. The heated formaldehyde sample cell was placed in the beam
adjacent to the by-pass optics. A 1 cm long sample cell replaced the heated
sample cell while investigating the interference by other hydrocarbons. In order
to use. longer sample paths for the study of interference by H2O and CO2, the by-
pass optics were removed and the multiple-pass sample cell was used at 40 passes.
No pressure transducer was included with the sample cell; therefore it was not
possible to measure directly the pressure of the formaldehyde vapor in the sample.
As in the GFC, a small amount of paraformaldehyde powder was placed in the sample
cell, and the vapor was formed as the heated (approximately 55°C) powder evaporated.
The cell was evacuated to less than 0.001 atm for a minute or less in order to ob-
tain "zero-readings". As soon as the valve on the sample cell was closed, the
pressure started increasing again as the powder evaporated. Typically, one-half
to one hour was required for the pressure to increase to the desired value
for another set of readings. The pressure was not in equilibrium after this re-
latively short time, but it was nearly constant while a set of measurements was
being made._ On several occasions, as many as 5 measurements of V' and either
Ag+at,s or Aat,s were made at intermediate pressures as the pressure was build-
ing up, or as the cell was partially evacuated in steps.
As explained below, many of the tests involved measuring the quantities required
to determine the correlation efficiency C£>, ond it was not necessary to know
the pressure of the formaldehyde vapor except for a few samples. In order to
estimate the formaldehyde pressure for j u-w samples, the values of Aat s were.
51

-------
related to the formaldehyde pressure by using absorption data obtained over
the same spectral interval for gas in the GFC at several different pressures.
The pressure transducer made it possible to measure the pressure of the vapor
in the GFC when its absorptance was measured. After the pressure had been de-
termined in this manner, the absorber thickness was computed and the important
Pembrook Factor F was calculated by use of Equation 20.
Several samples were investigated with N2 added to the formaldehyde vapor to
produce pressures up to 1 atm, the pressure at which most samples of interest
will probably be investigated. The N2 was added to the cell just after it had
been evacuated by first pressurizing the line to the cell then opening the cell
valve to allow the N2 to enter. The pressure was measured before the formalde-
hyde vapor had time to diffuse through the valve and the N2 to the unheated gas
line where it would have condensed. Nitrogen was also added to the vapor in
the heated GFC in the same manner in order to investigate the influence of
pressure-broadening of the absorption lines.
Because of the difficulties in accurately measuring the pressure of formalde-
hyde vapor in mixtures containing N2, it was decided that we would provide only
approximate calibration data to relate V' to formaldehyde absorber thickness.
The Environmental Protection Agency has developed more accurate methods of de-
termining formaldehyde partial pressures; therefore, the final, more accurate,
calibration will be performed by EPA scientists after they receive the instru-
ment. The data obtained by us have been limited primarily to small samples for
which a nearly linear relationship exists between V' and the absorber thickness
During most of the tests the instrument was operated in the normal mode so that
V1 could be measured directly. In addition the signal from the output jack
labeled "preamp" (see Figure 13) served as the input to a separate amplifier
and synchronous demodulator. This demodulator received its reference signal
from the output jack labeled "360 Hz Ref.". The demodulator output was pro-
portional to V , the average value of the carrier voltage. By monitoring the
output of this demodulator with a strip-chart recorder, we measured the absorp-
tance (or transmittance) of samples as they were introduced into the sample
cell or pumped from it.
In accordance wit1! the discussion and definitions in Section V, the absorptance
measured for a given sample depends upon the manner by which it is measured.
The absorptance Ag+at s is given directly from (D0 - Ds)/D0 if the instrument
is operating in its normal mode with the alternator switching the beam between
the formaldehyde GFC and the attenuator (see Equation (16)). The recorder
deflections observed with the sample cell filled and with the sample cell
evacuated are Ds and D0, respectively. In order to measure Aat s, the beam
through the GFC was temporarily blocked so that the only energy reaching the
detector was that passing through the attenuator. The recorder deflections
were observed with and without the sample in the cell, just as in the measure-
ment of Ag+at,s•
Of course, V1 could not be measured directly with the main electronics while
the beam through the GFC was blocked during the measurement of Aat g. There-
fore, the following procedure was followed in order to measure both V' and
Aat,s for a given sample: After the sample had been in place long enough to
be reasonably well stabilized, the value of V* was observed. The main
52

-------
electronics were then turned off, the beam through the GFC was blocked, and the
deflection Ds of the recorder connected to the synchronous demodulator was noted.
The sample cell was then evacuated and D0 was noted. Finally, the main elec-
tronics were turned back on to observe the zero-reading of V'. The difference
between the two values observed for V1, with and without the sample, represents
the true value of V'.
From Equations (15) and (17) we see that the correlation efficiency Ce can be
determined from a measurement of V' and of either Agt g or As+at . It is ap-
parent from the previous paragraph that ^g+at s •*-s much simpler to measure than
Aat s while measuring V'. However, when Ce is high, between approximately 0.5
and the maximum value of 1.0, the quantity Aat s is considerably greater than
Ag+at,s an^ can therefore be measured more accurately. The improved accuracy
is particularly important for samples with absorptances of only a few percent.
Therefore AatjS, rather than_Ag+gt s was measured for most of the samples. In
a few cases, both Aat,s and Ag+at,s were measured along with V1. The results
were consistent with Equations (15) and (17) within the estimated uncertainties.
Measuring the quantities required to determine Ce was much easier than measur-
ing the Pembrook Factor F because of the above-mentioned difficulties in deter-
mining the absorber thickness u. For a given spectral bandpass and a given set
of conditions of the formaldehyde GFC, values of Ce were quite reproducible and
nearly independent of the amount of formaldehyde in the sample provided ^5at,s
was less than approximately 0.2 and the sample pressure remained constant.
Table 1 summarizes the results of the measurements of correlation efficiency
Ce for the recommended spectral bandpass, which is centered at 2803 cm""l and
has a half-power bandwidth of 15 cm"l. The data in the table represent the
averages of several sets of measurements. Each group of data, A, B, and C,
corresponds to a given condition for the formaldehyde gas-filter cell (GFC).
Pure formaldehyde vapor was contained in the GFC for groups A and B. Nitrogen
was added to the GFC in order to obtain the data in group C.
For each set of GFC conditions, Ce is greater at low sample pressures than at
the highest pressure (1 atm) investigated. This dependence on sample pressure
results from the smoothing of the line structure due to collision broadening
of the lines as the pressure increases. It will usually be preferable to
operate with sample pressures near 1 atm so the sample can be flushed through
the sample cell without using a vacuum pump or a complex system of pressure
regulators. The maximum value of Ce tabulated for 1 atm samples is 0.28, which
was observed with enough pure formaldehyde in the GFC to reduce its average
transmittance Tg to approximately 0.37. Other data not included in the table
indicated that somewhat higher values of Ce could be achieved by employing even
lower transmittances of the GFC. However, the advantage of the small increase
in Ce is probably more than offset by the accompanying decrease in chopped
radiant energy that reaches the detector. For optimum performance, it is re-
commended that the GFC be operated with the temperature, and thus the formalde-
hyde vapor pressure, adjusted to produce an average transmittance T„ between
0.3 and 0.4.
The data in group C of Table 1 were obtained in order to see if Ce might be
improved by adding N2, or any non-absorbing gas, to the GFC to broaden the
absorption lines. It is most meaningful to compare data group B with group C
53

-------
TABLE 1


SUMMARY OF MEASUREMENTS OF
CORRELATION EFFICIENCY
FOR FORMALDEHYDE



GFC
Parameters



Group
Formaldehyde
Pressure
(atm)
Total
Pressure
(atm)
Average
Transmittance
Sample
Pressure
(atm)
Correlation
Efficiency
Ce
A
0.070
0.070
0.37
1.00
0.50
0.20
0.10
0.28
0.39
0.50
0.53
B
0.051
0.051
0.48
1.00
0.50
0.20
0.10
0.19
0.26
0.40
0.50
C
0.044
0.50
0.51
1.00
0.50
0.10
0.16
0.24
0.28
because Tg is approximately the same for both groups. It is apparent that Ce
is not improved by broadening the lines of the gas in the GFC so they more
nearly match the widths of the lines of the sample gas. In fact, Ce is de-
creased slightly by the increased GFC pressure, even when the sample pressure
is the same as the GFC pressure, or higher. As might be expected, C£ is lower
for high GFC pressures than for low GFC pressures when the sample pressure is
0.1 atm. At this low sample pressure, full advantage of the sharp structure
in the spectra is not realized if the spectral structure of the lines of the
gas in the GFC is not also sharp.
The average transmittance, TCT, of the vapor in the GFC was measured at several
pressures up to approximately 0.1 atm. The spectral bandpass was the same as
that used for the data represented in Table 1. The results indicate that the
average absorption coefficient kg for this spectral interval is approximately
4.0 (atm cm)"-'-. From Equation (22), we see that the Pembrook Factor F can be
determined from the values of Ce listed in Table 1.
C x 4 (atm cm) ^
F (atm cm) = —		^	 = 2Ce	(34)
It follows from Equations (18) and (19) that V', the quantity measured by the
instrument, can be related to Ce. For samples of low concentration c (ppm)
54

-------
at 1 atm pressure in the 40-meter cell when the GFC parameters are the same as
those indicated in group A of Table 1,
-1	3
V' = Fu = 2 (atm cm) Cg x c (ppm) x 4 x 10 cm x
(35)
10 ^ (atm x ppm ^) = 2.24 x 10 ^ c.
Discrimination against hydrocarbons is quite good when using the spectral band-
pass centered near 2803 cm"-'- that was recommended above. As might be expected
from the curves in Figure 16, there is no measurable interference by ethylene
because the absorption by this gas is negligible in the spectral region of
interest. Although both butane and propane absorb significantly in this region,
there is little structure in the spectral curves. Consequently, the gas-filter
method discriminates quite effectively so that large discrimination ratios are
observed: +1000 for propane and -2000 for butane. The interference by ethane
is so slight that it could not be measured reliably; the discrimination ratio
is probably > 5,000.
The discrimination ratio for methane is also quite high and is sensitive to
changes in the bandpass. By adjusting the center of the bandpass to several
positions within a 4 cm"l interval, the discrimination ratio was made to vary
from approximately +66 to M (no measurable interference) to -500. Thus, if the
methane concentration is high enough that very good discrimination is required,
a slight "tuning" of the bandpass may be made by placing a methane sample in
the cell .^nd nulling the interference. Changes of only 1 or 2 cm" ^ in the
center of the bandpass has only a very slight influence on the sensitivity to
formaldehyde.
When the H2O correction circuitry was de-activated, a sample of 2 percent H2O
in N2 at 1 atm total pressure produced an output that corresponds to approxi-
mately 0.25 ppm of formaldehyde. This corresponds to a discrimination ratio of
80,000 for H2O without using the H2O monitor and the correction circuits. When
the H2O correction circuits are employed and adjusted properly, the apparent
remaining error due to a 5 percent H2O mixture is less than 0.05 ppm of formal-
dehyde .
When the interference by other gases is negligible, the minimum detectable con-
centration is limited by the noise on the output signal of the instrument when
it is operating in the normal mode. Most of this noise comes from two basic
sources, detector noise and "optical" noise. The relative contribution by each
source depends on the operating condition. The amount of noise produced by the
PbS detector is essentially independent of the amount of chopped energy incident
upon it. Therefore, the relative amount of detector noise can be observed by
blocking the beam of chopped energy from the detector. Of course, the automatic
gain control must be turned off during this measurement so tha^ the amplifier gain
remains constant when the chopped energy is blocked. Noise generated by the pre-
amplifier and the other electronics that process the detector signal is much less
than the noise generated by the detector. For the sake of the present discussion,
this small amount of amplifier noise can be treated as part of the detector noise
55

-------
because its magnitude is essentially independent of the amount of chopped energy
incident on the detector.
In the present discussion, optical noise is defined as the noise in the output
signal that results from variations in the amount of chopped energy incident on
the detector. Thus, if the minimum detectable concentration is limited by opti-
cal noise, the signal-to-noise ratio cannot be improved by merely increasing
the chopped energy. Optical noise may be caused by turbulence and other types
of beam steering due to non-uniformities in the temperature or composition of
the gas in the optical path. Instabilities in mirrors, lenses, choppers, and
other optical components may also contribute to optical noise. This type of
noise may be greatly increased if some of the mirrors are not properly aligned,
particularly those in the alternator section.
Figure 17 illustrates the different types of noise on the signal output when
the instrument is adjusted to measure the formaldehyde concentration. The spec-
tral bandpass has a halfwidth of 15 cm"1 and is centered at 2803 cm"^. Three
ballasts for the Nernst glower source were turned on to produce a source current
of approximately 0.85 amp. The four upper strips were recorded with the multiple-
pass sample cell adjusted to 40 passes and filled to 1 atm with nitrogen. The
sample path-length and spectral bandpass are those recommended for maximum sen-
sitivity and accuracy in measuring formaldehyde concentrations. The AGC was
turned off, but the amplifier gain was manually adjusted to produce the same
value of Vc that the AGC maintains. With the gain adjusted manually in this
manner, it is possible to record the detector noise by blocking the chopped
energy from the detector.
Each short strip in Figure 17 was cut from a longer strip recorded while monitor-
ing the noise. The short strips, each corresponding to a 2-minute interval,
were then pieced together. The output signal V' was adjusted to read zero before
the recordings were traced. The plots were traced from right to left. As ex-
pected, the amplitude and the frequency of the noise decreased when the time
constant was increased from 1 sec to 3 sec. The detector noise and optical
noise combined are observed when the chopped signal is incident on the detector.
With the sample cell adjusted to 40 passes, the noise obviously decreased when
the radiant energy was blocked from the detector, indicating that the optical
noise made a sizeable contribution to the combined noise.
The optical noise may increase to a level greater than that indicated in the
upper right-hand panel if gas is flowing through the sample cell. This increase
in noise is probably due to turbulence and is particularly noticeable if the in-
coming gas has a different composition, or is at a different temperature than
the gas already in the cell.
The plots in the middle row were obtained in a similar manner but with the
multiple-pass cell adjusted to 4 passes. More chopped energy was incident on
the detector after the cell was set to 4 passes; therefore, the amplifier gain
was reduced to maintain the same value of V£. The detector noise observed with
the detector blocked decreased in direct proportion to the decrease in ampli-
fier gain. As with the cell adjusted to 40 passes, both the optical noise and
detector noise contribute significantly to the combined noise when the time con-
stant is 1 sec. Increasing the time constant reduces the detector noise to a
56

-------
DETECTOR BLOCKED
SIGNAL ON
DETECTOR
Figure 17. Recorder plots of the noise on the signal output for various operating conditions.
Each short strip represents a 2 minute scan with the electronic time constant given
below each section. The number of passes of the multiple-pass sample cell is in-
dicated on the left-hand side of the figure. The recorder deflection corresponding
to V' = 0,0001 is the distance between the short parallel lines that have been drawn
for reference in the middle of the figure.

-------
very small value when the cell is at 4 passes. Much of the optical noise is
apparently at a lower frequency and is not decreased as much by the increase in
time constant.
The noise with the signal on the detector is seen to be approximately twice as
much with the cell at 40 passes as at 4 passes. However, the sample path length
is approximately 10 times greater at the higher number of passes. Therefore,
the noise level is equivalent to a sample concentration 5 times as high at 4
passes as at 40 passes. It follows that the minimum detectable concentration is
one-fifth as high at the longer path, provided that noise is the limiting factor
Of course, increasing the sample path-length will not improve the accuracy if it
is limited by interference from other gases in the sample.
The four strips shown in the lower panel of Figure 17 were obtained with the GFC
and the attenuators removed from the alternator section. When these components
were removed, the amount of chopped energy incident on the detector increased by
approximately a factor of 3. Thus the amplifier gains had to be reduced by this
same factor in order to maintain Vc constant. The decrease in gain accounts for
the decrease in noise observed with the detector blocked. With the GFC removed,
the noise is much greater when the signal is on the detector than when the de-
tector is blocked. Thus, the optical noise is much greater than the detector
noise. Turbulence produced by the hot GFC when it is in place apparently does
not contribute a major portion of the optical noise.
From the recorder plots shown in the upper right-hand corner of Figure 17, we
conclude that noise limits the minimum detectable V' to approximately 0.0001
when the multiple-pass cell is adjusted to 40 passes (40 meters). From Equation
(35) we see that this corresponds to a minimum detectable concentration of ap-
proximately 0.05 ppm of formaldehyde.
58

-------
SECTION XII
MEASUREMENTS OF NH^ AND VINYL CHLORIDE NEAR 1000 cm"1
EXPERIMENTAL ARRANGEMENT
Spectral curves of transmittance and data ori the instrument sensitivity were
obtained for NH3 and vinyl chloride in order to determine the feasibility of a
gas-filter correlation instrument to measure the two gases at low concentrations.
Possible problems of interference were investigated, and data were obtained on sensi
tivity and noise levels. The work was limited to relatively small samples, and
somewhat less care was exercised than would have been done if the data were in-
tended for calibration of a finished instrument. The same experimental arrange-
ment was employed for both of these two gases because they have strong infrared
absorption features in the same spectral region. Figure 18 shows a spectral
curve of transmittance for each of these gases along with a curve for ethylene,
a gas that absorbs in the same spectral region and might interfere if it is
present in the sample being studied. The two strong NH3 absorption features
near 931 cm"^ and 966 cm"l were each investigated separately. The tests on
vinyl chloride made use of the strong absorption near 942 cm"-'-. Previous tests
in our laboratory had indicated that this absorption feature was stronger than
the one near 897 cm"'- and would result in better sensitivity if used in a vinyl
chloride sensor.
Most of the measurements were made with the by-pass optics assembly in place so
that the radiant energy beam did not go through the multiple-pass sample cell.
A small, room-temperature cell, 0.43 cm long, was placed in the beam just ahead
of the by-pass optics assembly. The GFC was 1 cm long and was located in the
alternator section in the position indicated in Figure 1. Both the sample cell
and the GFC contained NaCl windows. Lens L3 was not employed at the entrance
slit of the grating section since no lens that was transparent in this spectral
interval was available at the time. A 150 line/mm grating blazed at 1250 cm"-'-
(8 |J.m) dispersed the energy so the desired spectral interval could be selected
by the retroreflector-grid assembly. An interference filter that passes wave-
lengths greater than approximately 7.8 p,m eliminated overlapping orders of energy
of wavelengths shorter than those desired. The detector contained an HgCdTe ele-
ment that was cooled to liquid-nitrogen temperature. Spectral curves of trans-
mission and data on the instrument sensitivity and interference were obtained
in the same manner as the formaldehyde data.
RESULTS FOR NH^
Two dilute mixtures of NH3 + N2 were used for the samples; one was 0.757c NH3 and
59

-------
WAVENUMBER (cm'1)
Figure 18. Spectral curves of transmittance of NH-j, vinyl chloride, and
ethylene between 920 and 985 cm"1. The curves were scanned
with the versatile gas-filter spectrometer with no GFC in
place.
60

-------
the other was 3.37= NH^. Each of these mixtures was studied at total pressures
of approximately 0.2, 0.5, and 1 atm in the 0.43 cm sample cell. Samples were
flushed slowly through the sample cell for several seconds before the valves
were closed and the measurement was made. This allowed adsorption of NH3 on
the walls to take place and approach equilibrium so that the concentration of
the mixture in the sample would be close to that in the container from which
the mixture was drawn. Longer periods would be required for adsorption to come
to equilibrium if high accuracy were required. The estimated combined uncer-
tainty in the absorber thickness of the samples due to adsorption and errors in
pressure measurement is less than + 10%, which was adequate for the preliminary
tests being performed.
The majority of the data were obtained with the grating assembly adjusted to pass
a narrow spectral band that included the strong Q-branch near 966 cm~l. Two dif-
ferent grids were employed in the retroreflector-grid assembly. The wider one
(10 mm) passed a spectral interval approximately 7.2 cm" ^ wide at 966 cm"'- and
6.2 cm"! wide at 931 cm"l. The corresponding bandpasses are 2.9 and 2.5 cm"1
for the other grid, which had a narrower, 4 mm opening. The entrance slit was
1.6 mm wide for all of the gas-filter correlation data. A few data obtained
for the other Q-branch near 931 cm"-*- indicated that this region did not produce
sensitivity quite as high as the other wavenumber region. The dependence of
the correlation efficiency Ce (see Equation (13)) and the Pembrook factor, F,
on the parameters of the gas-filter cell (GFC) was also investigated.
Table 2 summarizes the data obtained for NH3. Each group corresponds to a given
spectral bandpass and a fixed set of GFC parameters with different sample total
pressures. The GFC was filled with NHg + N2 mixtures at different concentrations
to provide the total pressures and values of Tg shown in Table 2. Nearly all of
the samples studied produced values of Aat s less than 0.25 and V' less than
0.03. For samples absorbing this little, (5e and F are essentially independent
of the absorber thickness as long as the total pressure is constant. Each set
of conditions shown in Table 2 represents data for from one to five samples.
Values of Ce have been rounded off to the nearest 0.05 or 0.10 except for values
less than 0.20, for which the rounding was to the nearest 0.00, 0.02, 0.05, or
0.07. Values of F have been rounded off accordingly and are believed to be ac-
curate to between + 107o and + 207o.
If the GFC is at 1 atm, the results of group A indicate that Ce is independent,
within the accuracy of the present experiment, of sample pressure if it is less
than 1 atm. However, if the GFC is at 0.2 atm, Ce is higher for low-pressure
samples than for samples near 1 atm. Correlation efficiencies as high as 0.6
are observed when both the GFC and the sample cell are at 0.2 atm. This result
can be explained on the basis of pressure broadening of the many individual
lines that appear in the spectral bandpass. At the higher pressures, the lines
overlap each other, smoothing out much of the spectral structure on which the
GFC method depends for its sensitivity. When T , the average transmittance of
the GFC, is approximately 0.7, Ce is not as hign as when Tg is lower. The
higher value of Ce results because the GFC corresponding to the lower Tg is
more nearly opaque at the wavenumbers where the sample absorbs. In accordance
with the discussion of Section V, this increases the correlation efficiency.
Larger values of Ce are observed with the 7.2 cm"-*- spectral bandpass than with
the narrow 2.9 cm-l bandpass. The narrower interval contains strong lines
61

-------
TABLE
SUMMARY OF DATA ON NH^ SENSITIVITY
Group
Spectral Bandpass
Vc
(cm "*") (cm
GFC
P
(atm)
T
g
Sample
P
(atm)
C
e
F(a)
(atm cm) ^
A
966
7.2
1.0
0.5
1.0
0.40
2.2
A
966
7.2
1.0
0.5
0.5
0.40
2.2
A
966
7.2
1.0
0.5
0.2
0.40
2.2
B
966
7.2
0.2
0.5
1.0
0.40
2.2
B
966
7 . 2
0.2
0.5
0.5
0.50
2.7
B
966
7.2
0.2
0.5
0.2
0.60
3.3
C
966
7.2
1.0
0.7
1.0
0.20
1.1
C
966
7.2
1.0
0.7
0.5
0.20
1.1
c
966
7.2
1.0
0o 7
0.2
0.20
1.1
D
966
7.2
0.2
0.7
1.0
0.25
1.3
D
966
7.2
0.2
0.7
0.5
0.35
2.0
D
966
7.2
0.2
0.7
0.2
0.60
3.3
E
966
2.9
1.0
0.6
1.0
0.10
1.0
E
966
2.9
1.0
0.6
0.5
0.13
1.3
E
966
2.9
1.0
0.6
0.2
0.17
1.7
F
966
2.9
0.2
0.7
1.0
0.10
1.0
F
966
2.9
0.2
0.7
0.5
0.15
1.5
F
966
2.9
0.2
0.7
0.2
0.30
3.0
G
966
2.9
0.2
0.5
1.0
0.15
1.5
G
966
2.9
0.2
0.5
0.5
0.30
3.0
G
966
2.9
0.2
0.5
0.2
0.60
6.0
H
931
6.2
0.2
0.7
1.0
0.15
0.8
H
931
6.2
0.2
0.7
0.5
0.25
1.4
H
931
6.2
0.2
0.7
0.2
0.45
2.5
I
931
6.2
0.2
0.5
1.0
0.30
1.6
I
931
6.2
0.2
0.5
0.5
0.40
2.2
I
931
6.2
0.2
0.5
0.2
0.50
2.7

(a) F =
C k /2 = V/u for small
e s
absorp tance.



k =
11 (atm cm) ^
for 7.2 cm
^ interval at 966 cm ^ and 6.2
cm ^ interval a
931 cm \ and 20 (atm cm) ^
for the 2.9 cm interval at 966 cm
-1
62

-------
throughout most of its width and therefore must depend on the structure between
the closely spaced lines. Thus, Ce is very low when either the sample or the
GFC is at 1 atm or higher. On the other hand, the 7.2 cm"-*- interval is wide
enough to include some of the region of weak absorption next to the strong lines
in the Q-branch. Thus, some spectral structure remains within the spectral in-
terval, even when much of the structure between the lines is filled in. Use of
spectral intervals 15 or 20 cm"-'- wide that include one of the Q-branches would
probably produce slightly larger values of C£ than those obtained with the 7.2 cm
interval. However, the average absorption coefficient, ks (Equation (20)), is
correspondingly less for the wider interval. As a result, the important Pembrook
factor, F, is not expected to increase significantly as the bandpass increases
beyond 7.2 cm"l. As the interval becomes more than 2 or 3 times the width of
the Q-branch, the value of F will decrease because of the corresponding decrease
in ks. In summary, the 7.2 cm~l interval is probably near the optimum for max-
imum F.
After the grid was placed in the retroreflector-grid assembly to fix the width
of the spectral bandpass, the grating screw was adjusted to make a slight change
in vc, the center of the interval, to minimize the transmittance of the GFC.
Adjusting the location of the bandpass in this manner makes the instrument stabil
ity less susceptible to small changes in the spectral bandpass due to temperature
changes or to stresses on the instrument. At the possible expense of some sta-
bility, some increase in Ce and F might be gained by shifting vc a small fraction
of the bandwidth toward higher wavenumbers than those used in the experiment.
As one might expect from the spectral curves in Figure 18, ethylene interferes
with the measurement of ammonia. The discrimination ratio for ethylene varied
from approximately -10 to -30, depending on the pressures of the sample and the
GFC. The greatest interference, corresponding to the lowest discrimination ratio
was observed when both the GFC and sample were at 1 atm. Better discrimination
was observed when both cells were at 0.2 atm. The interference was approximately
the same at 931 cm~l as at 966 cm"-'-. Only the narrow spectral bandpass, approxi-
mately 2.9 cm"l, was employed in the ethylene interference measurements.
Interference by H2O is certain to be slight, and possibly insignificant, because
absorption by this gas in this spectral region is known to be very weak. Further
more, most of the H2O absorption that does occur is due to continuum absorption
against which a GFC instrument has good discrimination. Any possible H2O inter-
ference that might occur can easily be accounted for electronically by the H2O
monitor.
A one-atmosphere sample of 14.57= CO2 in N2 was introduced into the multiple-pass
sample cell when it was adjusted to 40 meters. The spectral bandpass and GFC
corresponded to data Group B in Table 2. The slight response of the instrument
to this large CO2 sample indicated a discrimination ratio against CO2 of about
800,000:1. A 167c mixture of CO2, which is about as rich as ever occurs in auto
exhaust, would produce an error equivalent to 0.2 ppm of NH3. By adjusting
some of the instrument parameters, the small amount of interference could likely
be reduced even further if it produced serious error. The CO2 concentration is
frequently known with fair accuracy so that the residual CO2 interference could
be calculated and accounted for.
We now consider the performance that we might expect from a GFC analyzer built

-------
specifically for NH^• Assume that the samples are kept at 1 atm in a multiple-
pass sample cell similar to the one in the present instrument with a one-meter
baselength and a total path length of 40 meters. The results obtained with for-
maldehyde indicated that a practical minimum lower limit of V1 - 10"^ is imposed
by optical instabilir.ies and turbulence in the multiple-pass sample cell. From
the results summarized in Table 2, we see that we can expect a Pembrook factor
F --2. Thus the minimum detectable absorber thickness that is imposed by a min-
imum V' of 10"^ is 10~^/2 = 0.5 x 10"^ atm cm. In a 4000 cm cell, this corres-
ponds to a minimum detectable concentration of (0.5 x 10"^-)/4000 = 1.25 x 10"8
atm cm = 0.0125 ppm.
This very good sensitivity is, of course, dependent on having an optical system
with high energy throughput, a good source and a detector with high detectivity
so that the instrument sensitivity is not limited by detector noise or amplifier
noise. The present instrument is limited by the noise from the HgCdTe detector.
With the bandpass adjusted as it was when obtaining the data for Group B in
Table 2, a time-constant of 3 seconds, and the multiple-pass cell adjusted to
40 meters, the noise-level corresponded to V1 -- 0.003. By comparing this with
the results from the previous paragraph, we see that this corresponds to a mini-
mum detectable concentration of approximately 0.4 ppm.
Better detectivity than this could be obtained with the same type of instrument
by making a few changes. For example, it is probably possible to gain a factor
of approximately 4 or 5 by decreasing the size of the detector element and pay-
ing a premium price for a detector with higher detectivity. Installing a cold-
filter in the detector dewar just ahead of the sensitive element could further
improve the detectivity by decreasing the amount of background energy incident
on the detector. Gains of 2 to 4 could also be obtained by different methods
of increasing the energy throughput. In summary, it appears that a minimum de-
tectable NH3 concentration as low as 0.02 to 0.04 ppm could be obtained with a
GFC instrument employing the 966 cm~l absorption feature and a multiple-pass
absorption cell. The instrument could easily fit on a table-top and be easy
to use and make nearly real-time measurements with a response time of a few
seconds. Even further reductions in the minimum detectable concentration to
less than 0.01 ppm with such an instrument does not appear to be unrealistic.
RESULTS FOR VINYL CHLORIDE
The data obtained on vinyl chloride are summarized in Table 3 in the same way
that the NH^ data are summarized above. Two different spectral intervals were
employed; both were centered near the strong absorption at approximately 942 cm"''-.
The dependence of Ce and of F on spectral slitwidth, average transmittance Tg .
of the GFC, and sample pressure are similar to those observed with NH3. The
highest values of Ce and F are obtained with low values of Tg. The optimum
value of Tg for a particular instrument depends on other factors, including de-
tector noise. Further decreasing of Tg below values represented in the table
would probably result in additional increases in Ce and F. However, the de-
crease in Tg reduces the amount of energy reaching the detector; thus, the de-
tector noise becomes a bigger factor.
Comparison of data Groups E and G in the table shows that Ce has little, if any,
dependence on the pressure of the gas in the GFC when the spectral interval is
wide enough to include the strong absorption and some weaker absorption on
64

-------
TABLE 3
SUMMARY OF DATA ON VINYL CHLORIDE SENSITIVITY
Sj.u'.ctral Bandpjss
GFC
C.roui
V
c
(cm )
(<
Av
-U
;m )
p(a)
(atm)
T
8
P
(atm)
C
e
F(b)
, N-l
(atm cm)
A
942

2.6
.06
.75
1.0
0.07
.20
A
942

2.6
.06
.75
0.5
0.10
.28
A
942

2.b
» Ob
.75
0.2
0.15
.40
B
942

2.6
.15
.47
1.0
0.17
.50
U
942

2.6
.15
.47
0.5
0.25
.70
r,
94L'

2.6
.15
.47
0.2
0.30
.85
c:
942

2.6
.26
.31
1.0
.30
.85
-
942

2.6
.26
.31
0.5
.30
.85
D
942

6.5
.10
.73
1.0
.15
.28
D
942

6.5
.10
.73
0.5
.17
.32
D
942

6.5
.10
.73
0.2
.20
.37
E
942

0.5
.20
.55
1.0
.25
.46
E
942

6.5
.20
.55
0.5
.25
.46
E
942

6.5
.20
.55
0.2
.25
.46
F
942

6.5
.36
.35
1.0
.30
.56
F
942

6.5
.36
.35
0.5
.30
.56
G
942

6.5
1.0
.55
1.0
.25
.46

(a) Pure
vinyl chloride was used
in the
GFC for groups
A through F; a mixture

of vinyl
chloride
plus nitrogen was
used for G.



(b) F -
c: lc
e s
/2 = V/u
for small a
bsorptance.



k =
s
5.6
(atm cm)
-1 .
1 * , o
for the 2
.6 cm ^
interval and 3
.7 (a tm
cm) * for the
6.5 cm interval.
65

-------
either side. Group A includes data obtained with a narrow, 2.6 cm" , spectral
_interval and a relatively small amount of vinyl chloride in the GFC to produce
Tg = 0.75. In this case the transmittance is low only very near the centers of-
the strongest lines, and the instrument sensitivity is dependent on the structure
between the closely-spaced lines. At low pressures, the lines are narrow and
there is more structure than occurs at higher gas pressures. For Tg as high as
0.75, Ce is very low except when the pressures in both the GFC and the sample are
low.
A few data were obtained on the interference by ethylene, ammonia, CO2 and H2O
with the spectral bandpass and the GFC parameters similar to those in data
Group E in Table 3. The discrimination ratios were observed to be greater than
50,000:1 for both CO2 and H2O. The H2O interference was not determined reliably;
the discrimination ratio may be much greater than 50,000:1. Any interference by
this gas that might occur could easily be accounted for by an H2O monitor with
automatic correction circuitry similar to that in the instrument being reported
here. The largest sample of NH3 investigated for interference consisted of 1 atm
of pure ammonia in a 0.43 cm cell. The absorptance, AatjS, of this sample was
approximately 0.20. However, the correlation of the spectral structure of the
NH3 and vinyl chloride is apparently low because this produced a V1 of 0.003 and
a corresponding discrimination ratio of approximately -100:1. Absorption by
ethylene is also strong in this spectral region, and its discrimination ratio
is approximately 70:1. Although the discrimination ratios for ethylene and
ammonia are relatively low, their concentrations are expected to be low in most
places where vinyl chloride is being monitored; therefore, the interference
would probably not be serious.
In order to estimate the performance of a GFC instrument designed specifically
for vinyl chloride, we can compare the results with those for ammonia. For vinyl
chloride, we can expect F to be approximately 0.7 for 1 atm samples and a 6.5 cm"^
spectral interval. This is 0.35 as high as the values assumed above for NH3;
thus the estimated minimum detectable concentration of vinyl chloride is approxi-
mately 3 times as high as for NH3. With the present instrument adjusted for a
40 meter path-length, the minimum detectable concentration is approximately 1 ppm.
However, by making some of the changes discussed above with regard to NH3, this
quantity can probably be reduced to less than 0.1 ppm.
66

-------
SECTION XIII
REFERENCES
1.	Burch, D. E., and J. D. Pembrook. "Instrument to Monitor CH^, CO, and
CO2 in Auto Exhaust." Prepared by Philco-Ford Corporation for EPA under
Contract No. 68-02-0587. EPA Report No. 650/2-73-030, October 1973.
2.	Burch, D. E., and D. A. Gryvnak. "Infrared Gas Filter Correlation
Instrument for In-Situ Measurement of Gaseous Pollutants." Prepared by
Philco-Ford Corporation for EPA under Contract No. 68-02-0575. EPA
Report No. EPA-650/2-74-094. Also, Burch, D. E., and D. A. Gryvnak.
"Cross-Stack Measurement of Pollutant Concentrations Using Gas-Cell
Correlation Spectroscopy." Chapter 10 of Analytical Methods Applied to
Air Pollution Measurements, Stevens, R. K. and W. F. Herget, (eds.).
Ann Arbor, Ann Arbor Science Publishers Inc., 1974.
3.	Burch, D. E. "Adjustable Bandpass Filter Employing a Prism." Appl.
Optics 8, 649 (1969) .
4.	White, John U. "Long Optical Paths of Large Aperture." Journ. of the
Optical Society of America 32:285-288, May 1942.
5.	Decker, J. A. Jr. "Experimental Realization of the Multiplex Advantage
with a Hadamard-Transform Spectrometer." Appl. Opt. 10, No. 3, 510-514,
1971.
67

-------
TECHNICAL REPORT DATA
(I'! use read Instructions on the reverse before completing)
1 REPORT NO
EPA-600/2-75-024
3. RECIPIENT'S ACCESSIOf^NO
4. TITLE AND SUBTITLE
5. REPORT DATE
VERSATILE GAS FILTER CORRECTION SPECTROMETER
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8 PERFORMING ORGANIZATION REPORT NO
D. E. Burch, F. J. Gates, D. A. Gryvnak, J. D. Pembrook
U-6201
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Aeronutronic Ford Corporation
Aeronutronic Division
Ford Road
Newport Beach, Calif. 92663
10. PROGRAM ELEMENT NO.
P.E. 1A1010 (ROAP 26 ACV)
11. CONTRACT/GRANT NO.
68-02-1227
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final: June '73 - June '75
14. SPONSORING AGENCY CODE
EPA - ORD
15 SUPPLEMENTARY NOTES
16. ABSTRACT
A versatile infrared analyzer employing gas-filter correlation techniques has been
designed and constructed to measure concentrations of pollutant gases from a variety
of sources. The spectral bandpass is determined by an adjustable grating assembly.
By interchanging cell windows, radiant energy sources, gratings, interference
filters, and detectors, nearly any desired spectral bandpass between 0.3 |im and
11 iJbm can be obtained. Spectral curves of transmittance can also be scanned. A
multiple-pass sample cell provides sample paths between approximately 4 m and 40 m.
Shorter sample cells can also be employed. An 1^0 monitor measures the concentra-
tion of H2O in the multiple-pass sample cell and automatically accounts for inter-
ference by H2O in the measurement of other gas concentrations. Tests have been
performed on the measurement of formaldehyde, vinyl chloride and ammonia. The mini-
mum detectable concentration of formaldehyde in automotive exhaust is approximately
0.05 ppm.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. cosati Field/Group
*	Spectrometer
*	Exhaust gases
Air pollution
Formaldehyde
Vinyl chloride
Ammonia
Gas-filter correlation
14B
13B
2 IB
07C
07B
18 DISTRIBUTION STATEMENT
Release to public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
76
20. SECURITY CLASS (This pagej
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
68

-------
ENVIRONMENTAL PROTECTION AGENCY
Technical Publications Branch
Office of Administration
Research Triangle Park, Noith Carolina 2771 1
OFFICIAL BUSINESS
AN EQUAL OPPORTUNITY EMPLOYER
POSTAGE AND FEES PAID .
ENVIRONMENTAL PROTECTION AGENC
EPA -335
Return this sheet if you do NOT wish to receive this material Q .
or if change of address is needed £3 (Indicate change, including
ZIP code.)
PUBLICATION NO. EPA-600/2-75-024

-------