Jump to main content or area navigation.

Contact Us

Technology Transfer Network
Emission Measurement Center Spectral Database

Coal Fired Boiler Emissions Test




EPA Contract No. 68D20163

Work Assignment No. I-34

Prepared by:

Research Division

Entropy, Inc.

Post Office Box 12291

Research Triangle Park, North Carolina 27709

Prepared for:

Lori Lay

U. S. Environmental Protection Agency

Emissions Measurement Branch

Research Triangle Park, North Carolina 27711

June 15, 1994


This document was prepared by Entropy, Inc. under EPA Contract No. 68D20163, Work Assignment No. I-34. This document has been reviewed by the U.S. Environmental Protection Agency (EPA).

The opinions, conclusions, and recommendations expressed herein are those of the authors, and do not necessarily represent those of EPA.

Mention of specific trade names or products within this report does not constitute endorsement by EPA or Entropy, Inc.


    1.1  BACKGROUND 
    4.5  PROCESS DATA 
     NOTE: Appendices A-D are not available



The U. S. Environmental Protection Agency (EPA) Office of Air Quality Planning and Standards (OAQPS), Industrial Studies Branch (ISB), and Emission Measurement Branch (EMB) directed Entropy, Inc. to conduct an emission test at New York State Electric and Gas Company's (NYSEG) Kintigh Station (Site 12), a coal-fired electric generating unit in Somerset, New York. The test was conducted from July 26 to July 29 1993. The purpose of this test was to identify which hazardous air pollutants (HAPs) listed in the Clean Air Act Amendments of 1990 are emitted from this source. The measurement method used Fourier transform infrared (FTIR) technology, which had been developed for detecting and quantifying many organic HAPs in a flue gas stream. Besides developing emission factors (for this source category), the data will be included in an EPA report to Congress.

Before this test program, Entropy conducted screening tests using the FTIR method at facilities representing several source categories, including a coal-fired electric utility. These screening tests were part of the FTIR Method Development project sponsored by EPA to evaluate the performance and suitability of FTIR spectrometry for HAP emission measurements. These tests helped determine sampling and analytical limitations, provided qualitative information on emission stream composition, and allowed estimation of the mass emission rates for a number of HAPs detected at many process locations. The evaluation demonstrated that gas phase analysis using FTIR can detect and quantify many HAPs at concentrations in the low part per million (ppm) range and higher (sub-ppm for some HAPs), and a sample concentration technique was able to detect HAPs at sub-ppm levels.

Following the screening tests, Entropy conducted a field validation study at a coal-fired steam generation facility to assess the effectiveness of the FTIR method for measuring HAPs, on a compound-by- compound basis. The flue gas stream was spiked with HAPs at known concentrations so that calculated concentrations, provided by the FTIR analysis, could be compared with actual concentrations in the spiked gas stream. The analyte spiking procedures of EPA Method 301 were adapted for experiments with 47 HAPs. The analytical procedures of Method 301 were used to evaluate the accuracy and precision of the results. Separate procedures were performed to validate a direct gas phase analysis technique and a sample concentration technique of the FTIR method. A complete report, describing the results of the field validation test, has been submitted to EPA.[1]

This report was prepared by Entropy, Inc. under EPA Contract No. 68D20163, Work Assignment No. I-34. Research Triangle Institute (RTI) provided process information included in Sections 2.1, 2.2 and 3.3.3.


The FTIR-based method uses two different sampling techniques: (1) direct analysis of the extracted gas stream (hereafter referred to as the gas phase technique or gas phase analysis) and (2) sample concentration followed by thermal desorption. Gas phase analysis involves extracting gas from the sample point location and transporting the gas through sample lines to a mobile laboratory where sample conditioning and FTIR analyses are performed. The sample concentration system employs 10 g of Tenax® sorbent, which can remove organic compounds from a flue gas stream. Organic compounds adsorbed by Tenax® are thermally desorbed into the smaller volume of the FTIR absorption cell; this technique allows detection of some compounds down to ppb levels in the flue gas. For this test, approximately 850 to 1100 dry liters of flue gas were sampled during each sample concentration run. Section 4.0 describes the sampling systems.

Entropy operated a mobile laboratory (FTIR truck) containing the instrumentation and sampling equipment. The truck was driven to the site at Kintigh station and parked near to each location. The test runs were performed over three days.

Entropy tested the exhaust gases from the Unit 1 coal-fired boiler. The furnace burns bituminous coal. Gases from the combustion furnace pass through two electrostatic precipitators (ESPs) to control particulate. Gases exiting the ESPs pass through a flue gas desulfurization unit (FGD), to remove SO2, and are then exhausted through a stack. Section 2.0 contains descriptions of the process and the sampling point locations.

Direct gas phase analysis was used to measure carbon monoxide (CO), carbon dioxide (CO2), sulfur dioxide (SO2), nitrogen oxides (NOx), and ppm levels of other species. EPA instrumental test methods were used to provide concentrations of hydrocarbons (HC), CO, CO2, and O2. The sample concentration technique was used to measure HAPs at ppb levels.

Entropy conducted three 4-hour sample concentration runs and one gas phase run at the inlet and outlet of the ESP, a gas phase run at the FGD outlet upstream of the stack, and three sample concentration runs at the stack. Combustion gas volumetric flows were calculated from measurements taken using a 3-dimensional pitot probe and an S-type pitot probe. Section 3.1 gives the test schedule.


This testing program is being funded and administered by the Industrial Studies Branch (ISB) and the Emissions Measurement Branch (EMB) of the Office of Air Quality Planning and Standards (OAQPS) of EPA. An RTI representative collected process data. The following organizations and personnel have been involved in coordinating and performing this project.

  NYSEG Corporate Contact:         Mr. Peter Carney        (607) 729-2551
                                   Mr. Mehdi Rahimi        (607) 762-4212  
  Kintigh Station                  Mr. Donald Freed        (716) 795-9501
  EMB Work Assignment              Ms. Lori Lay            (919) 541-4825
  Managers:                        Mr. Dennis Holzschuh    (919) 541-5239
  ISB Contacts:                    Mr. Kenneth Durkee      (919) 541-5425
                                   Mr. Bill Maxwell        (919) 541-5430
  Entropy Project Manager:         Dr. Thomas Geyer        (919) 781-3851
  Entropy Test Personnel:          Mr. Greg Blanschan 
                                   Mr. Stuart Davis
                                   Mr. Ricky Strausbaugh
                                   Ms. Lisa Grosshandler
                                   Mr. Scott Shanklin
                                   Dr. Grant Plummer
                                   Dr. Ed Potts
                                   Mr. Mike Worthy
  RTI Representative:              Mr. Jeffrey Cole        (919) 990-8606



New York State Electric and Gas Company's (NYSEG) Kintigh Station in Somerset, New York is about 30 miles Northeast of Buffalo on the south shore of Lake Ontario. Unit One is a pulverized coal-fired (bituminous, medium sulfur) base-loaded unit that normally operates 24 hours a day, 7 days a week, except for a 1- to 2-week planned outage for maintenance once a year. Although Unit One is technically a base- loaded unit, a better character-ization would be a load-following unit. The unit operates from 22 to 100 percent capacity (151 to 688 MWe) depending on the need for power.

The fuel source for Unit One is medium-sulfur coal from Pennsylvania and West Virginia. Railroad cars bring the coal to the plant site where the coal is first crushed and then fed by conveyer sequentially into one of six coal storage bunkers. From these bunkers, the crushed coal is gravity fed to six pulverizing mills directly below the storage bunkers. These mills pulverize the crushed coal into talcum powder consistency. While the coal is being transported to the silos by either of two belts, it is sampled every 2.3 minutes by an automatic sampler. A second transfer belt carries the sample to a mill that grinds the sampled coal and places it in a sampling container. Proximate/ultimate analysis of a composite coal sample are obtained every 24 hours. Table 2-1 shows coal analysis results from days the plant was tested and Table 2-2 shows an analysis of the limestone used for SO2 control. During the test, the unit consumed coal at an average rate of 234.2 ton/hr.

Unit One is a pulverized coal-fired, dry bottom boiler with an opposed firing configuration. Low NOx burners are used to control NOx emissions. Normal operating temperature in the combustion zone is approximately 2,000 - 3,000 degF. The boiler produces 4,500,000 lb/hr of steam for a General Electric steam turbine generator rated at 688 MWe.

Figure 2-1 is a diagram of the process. Combustion air is supplied from two sources, the primary and secondary air systems. Ambient air is drawn into the primary air inlet by a fan and passes through the primary air preheater (PAH). The preheated air then mixes with the pulverized coal and the mixture is blown into the combustion chamber through the burners. The secondary air system starts at the secondary air inlet where ambient air is drawn into the secondary air system by a forced draft fan and passes through the secondary air preheater (SAH). The secondary preheated air is ducted to windboxes where it is introduced above the burners as secondary air used to complete combustion.

Combustion gases and particulate matter exiting the boiler pass through the secondary and primary superheaters, economizer, and the air preheaters to the electrostatic precipitator (ESP). There are four air preheaters, two primary and two secondary. The primary and secondary air preheater ducts meet just before the ESP entrance. The secondary system is larger and accounts for approximately 85 percent of the gas flow. Air tempering dampers are used on the primary preheaters to balance the air flow. From the ESP, the combustion gases flow through three induced draft (ID) fans and into the flue gas desulfurization unit (FGD). The flue gas then enters a 625.5-ft stack and is exhausted to the atmosphere through the 26-ft 8-in exit diameter.


2.2.1 Nitrogen Oxides (NOx) Control

This unit uses low NOx burners to control NOx. Low NOx burners are specially designed burners that carefully control the amount of fuel and combustion air to reduce combustion temperatures and lower NOx emissions.

2.2.2 Sulfur Dioxide (SO2) Control

A FGD is used to control SO2 emissions. These wet scrubbers (6 modules, 4 in use at full load) use limestone as a reagent. The modules are arranged in parallel so that units can be activated and deactivated without effecting other modules. The liquid-to-gas ratio of the FGD is approximately 94 gallons per 1,000 acf and the reagent ratio is approximately 1.15 moles of reagent, CACO3 in the limestone, per mole of SO2. Only fresh limestone is used; therefore, there is no reagent recycling. The FGD system is approximately 70 to 90 percent efficient.

2.2.3 Particulate Control

A cold-side ESP is used to collect flyash that exits the air preheaters. The total collection area and specific collection area (SCA) of the ESP are 1,931,776 ft2 and 840 ft2/1000 acfm, respectively. The ESP has 5 fields in 8 parallel sections resulting in 40 cells. The flue gas flow is divided into four streams before the ESP and enters separate East and West sections. These sections are physically separated and consist of 5 fields in 4 parallel sections resulting in 20 cells. Only the inlet and outlet of the West section was tested (Figure 2-2), but flow measurements were taken at the East and West sections of the ESP. A rapping system is used to dislodge dust from plates inside the ESP. The dust is collected in hoppers. This system raps plates at different times to prevent large amounts of particulate from re- entering the gas stream. The design efficiency of the ESP is 99.86 percent.


Figure 2-3 is a general schematic showing the four test locations.

2.3.1 Boiler Outlet (ESP Inlet)

The main duct carrying gases from the boiler branches into four ducts before passing through the ESPs. Two ducts carry flue gas through each ESP. Entropy tested flue gas using ports on the western most duct of the west-side ESP (Figure 2-4). A grating platform provided access to the six, 4-inch diameter, sampling ports evenly spaced across the top of the 13.25 ft deep, 12.5 ft wide horizontal duct. The sampling location was 23 ft upstream of the ESP building and 66 ft above ground level. The location was reached by stairs and a catwalk. The FTIR truck was parked next to the ESP building and 200 ft of sample line was used to connect the sample probe to the heated pump.

2.3.2 ESP Outlet (FGD Inlet)

The gas stream exits the two ESPs through four ducts, two ducts emanate from each ESP. Each duct has six four-inch diameter sample ports located about 23 ft downstream of the ESP buildings. Entropy extracted flue gas from the western most duct of the west-side ESP outlet (Figure 2-5). The ports were equally spaced along the top of the 16.5 ft deep, 12.5 ft wide horizontal duct. Access to this location was provided by stairs and a catwalk. This location was 66 ft above ground level and was reached using 200 ft of sample line.

2.3.3 FGD Outlet

The four ducts exiting the ESPs recombine before the gas stream passes through the induced draft (ID) fan which blows the gas through the flue gas desulfurization unit (FGD) and out the stack. There are two sampling locations available at the outlet of the FGD. One location is through ports in the 17 ft 4-inch wide breaching connecting the scrubber and the stack (Figure 2-6). The other location is at the 350-foot level of the stack.

Gas phase sampling was performed through ports in the breaching. Four six-inch diameter sample ports were arranged vertically along the side of the breaching. Sampling was conducted through one of the two center ports about 44 ft above ground level. Access was provided by ladder and a platform. The FTIR truck was parked directly beneath and the location was reached using 100 ft of sample line. No flow measurements were taken because of the difficulty of performing velocity traverses using a 3-D probe. It was assumed that volumetric flow at the stack and at the FGD outlet were equivalent because there are no obstructions between these locations. EPA agreed that flow data from the stack could be used for the FGD outlet for the purpose of calculating mass emission rates.

2.3.4 Stack

The breaching carrying flue gas from the FGD connects to the base of the stack. A platform is located at the 350 ft level of the 30.7 ft diameter stack (Figure 2-7). Access to the platform was provided by an elevator. There were four, 4-inch ID, sample ports evenly spaced around the stack circumference. One of these was occupied by plant CEM equipment. One of the remaining ports was used for the sample concentration runs. velocity traverses were performed through all three ports before and after each run using a standard S-type pitot.



The purpose of the test program was to obtain information that will enable EPA to develop emission factors (for as many HAPs as possible) which will apply to electric utilities employing coal-fired boilers using bituminous coal. EPA will use these results to prepare a report for Congress.

    The specific objectives were:  
          Measure HAP emissions (employing methods based on FTIR
            spectrometry) in two concentration ranges, about 1 ppm and
            higher using gas phase analysis, and at sub-ppm levels using
            sample concentration/thermal desorption.  
          Determine maximum possible concentrations for undetected HAPs
            based on detection limits of instrumental configuration and
            limitations imposed by composition of flue gas matrix.
          Measure O2, CO2, CO, and hydrocarbons using gas analyzers.  
          Perform simultaneous sample concentration runs at the inlet
            and outlet of the west ESP and stack.  Perform separate gas
            phase runs at the inlet and outlet of both control devices. 
          Analyze data to determine effect (if any) of the control
            devices on HAP (and other pollutant) emissions.
          Obtain process information from Kintigh.  This information
            includes the rate of power production during the test runs and
            operating parameters of the control devices.

Table 3-1 presents the testing schedule that was followed at Kintigh.


Entropy had planned to collect gas phase samples sequentially at the inlet and outlet of the ESP during Run 1. The distance between the two locations was too long to connect heated line to both at the same time. Instead, they were tested separately; the inlet during Run 1 and the outlet during Run 2. The FGD outlet was sampled during Run 3.


3.3.1 FTIR Results

Gas phase and sample concentration data were analyzed for the presence of HAPs and other species. All spectra were visually inspected and absorbance bands were identified. The spectra were analyzed to determine concentrations of detected species using procedures developed by Entropy. The results are presented in Tables 3-2 and 3-3. Maximum possible (minimum detectible) concentrations were determined for undetected HAPs. These results are presented in Tables 3-4, 3-5, 3-6, and 3-7. Gas Phase Results -- Each gas phase FTIR spectrum was analyzed for HAPs and other species. The spectra revealed that the gas phase samples were composed of:

      water vapor
      CO2 was detected, but measured using a gas analyzer ( Table 3-8).  
      NO was not quantified in undiluted hot/wet samples because of water
        interference, but was measured at about 520 ppm in condenser and
        diluted samples.
      CO was detected, but measured using a gas analyzer (Table 3-8).  
      NO2, and N2O, were detected and will be quantified in the final
        report after reference spectra become available.  
      SO2 was measured at the ESP inlet and outlet at an average
        concentration of about 1100 ppm, and at the FGD outlet at an
        average concentration of about 220 ppm.

Calculated concentrations of NO, and SO2, for each spectrum are given in Table 3-2. A set of subtracted spectra was generated to analyze for the maximum possible (or minimum detectible) concentrations of undetected HAPs. Reference spectra of water vapor, SO2, NO and CO2 were scaled and subtracted from each sample spectrum. The resulting base lines were analyzed using procedures described in Section 4.6.3. The upper limit concentration of an undetected compound is referred to in Tables 3-4, 3-5, and 3-6 as the maximum possible concentration. This quantity was determined for HAPs in the reference library. Results for hot/wet, condenser and diluted samples are presented in Tables 3-4, 3-5, and 3-6, respectively. The results are averages of the calculated values for all of the spectra in a sample run.

Hot/wet gas phase spectra are the most difficult to analyze due to spectral interference from water vapor. Even so, in results from the hot/wet gas phase data, 90 compounds gave minimum detectible concentrations below 10 ppm; of these, 75 are below 5 ppm, and 25 are 1 ppm or lower.

Previously, Entropy generated program files to analyze for HAPs in FTIR spectra of samples extracted from a coal-fired boiler stack. Statistical analysis showed that the programs were successful in measuring some HAPs in hot/wet and condenser samples.1 The same programs were run on gas phase data from Kintigh. The results are presented in Appendix C. Sample Concentration Results -- The concentrated samples are integrated samples collected over 4-hours. The following compounds were detected in addition to water vapor, CO, and CO2:

      HCl was detected in samples from all three runs at all locations
        with exception of Run 2 at the stack.  Being volatile, HCl does not
        adsorb to Tenax® well: therefore, HCl concentrations from Tenax
        samples represent lower limit concentrations.  The upper limit HCl
        concentration is about 1 ppm from Table 3-4 for ESP inlet/outlet
        samples.  HCl is probably trapped in the water which condensed in
        the tubes.
      HCN traces were detected in samples from Run 3 at the ESP inlet,
        from Run 1 at the ESP outlet, and from Runs 1 and 3 at the stack. 
        Concentrations were estimate using a spectrum from Infrared
        Analysis.  HCN is volatile so the Tenax® data provide a rough
        estimate of the lower limit concentrations.  An estimate of the
        upper limit concentration is not shown in Table 3-4 because the HAP
        library does not contain HCN spectra.    
      Traces of SO2 were detected in all of the samples.  However, sample
        concentration removes most of the SO2 from the FTIR sample and the
        gas phase data (Table 3-2) provide a better estimate of the SO2
        flue gas concentration.  
      Ammonia (NH3) was detected in the samples from Run 3 at the ESP
        outlet, and Runs 1 and 3 at the stack.  Tenax® results give a lower
        limit estimate of the ammonia concentration.  The upper limit is 1
        to 2 ppm as given in Table 3-4.  
      Formaldehyde (CH2O) was detected in samples from Runs 1 and 2 at
        the ESP inlet, from all 3 Runs at the ESP outlet, and from Run 2 at
        the stack.  Tenax® results give a lower limit estimate of the
        formaldehyde concentration.  The upper limit is 1 to 2 ppm as given
        in Table 3-4.   
      Formic acid (CH2O2) was detected in Run 2 at the ESP inlet and
        outlet.  Formic acid is volatile so the Tenax® data provide a rough
        estimate of the lower limit concentrations.  An estimate of the
        upper limit concentration is not shown in Table 3-4 because the HAP
        library does not contain formic acid spectra.    
      Benzene was detected in the sample from Run 1 at the ESP outlet.  
      Carbonyl sulfide (OCS) was detected in the sample from Run 1 at the
        ESP outlet.  
      Methane was detected in the sample from Run 1 at the outlet.  
      Freon(11) (CCl3F) was detected in samples from Runs 1 and 2 at all
        three locations and Run 3 at the stack.  
      Evidence of hexane was detected in samples from all runs at all
        three locations.  Absorbance features assigned to hexane are likely
        due to a mixture of aliphatic hydrocarbons, including hexane, the
        sum of whose spectra is similar to that of hexane.  
      A cyclic siloxane was detected that Entropy first measured in
        spectra of samples taken at the coal-fired boiler validation test.1 
        At that time it was shown to be a product of a reaction between HCl
        or water vapor in the gas stream and materials in the filter
        housing of the Method 5 box.  Entropy took steps to eliminate this
        problem and the cyclic siloxane, if it is a contaminant, is present
        (in the samples from Kintigh) at very low levels relative to
        validation data. It was detected in samples from Run 1 at the ESP
        inlet, and all Runs at the ESP outlet and the stack.  

Table 3-3 presents calculated concentrations of HCl, hexane, formaldehyde, CCl3F, HCN, formic acid, OCS, methane and NH3 in samples where these species were detected. Concentrations of CCL3F, HCN, and formic acid were estimated using spectra prepared by Infrared Analysis Inc. Flue gas concentrations in Table 3-3 were determined by dividing the in-cell concentration by the concentration factor (Section 4.6.5) and are based on the volume of gas sampled.

Table 3-7 gives minimum detectible concentrations for HAPs undetected using Tenax®.

Other unidentified absorbance bands were observed in the spectra. None of these features were attributed to HAPs listed in Table 3-7. The same absorbance features did not consistently appear in every sample spectrum.

Spectral analysis programs were also developed for validation of sample concentration spectra. The programs were run on the data collected at Kintigh and the results for compounds that have been measured using thermal desorption FTIR with Tenax® 1 are presented in Appendix C.

3.3.2 Instrumental and Manual Test Results

Table 3-8 summarizes the results of the EPA Methods 3A, 10 and 25A tests as described in Section 4.3. All CEM results in the table were determined from the average gas concentration measured during the run and adjusted using the pre- and post-run calibration results (Equation 6C-1 presented in EPA Method 6C, Section 8). Although not required by Method 10, the same data reduction procedures as that in Method 3A were used for the CO determinations to ensure data quality. All measurement system calibration bias and calibration drift checks for each test run met the applicable specifications contained in the test methods.

3.3.3 Process Operation During Test Runs Process Data -- The process data collected during the test runs have been tabulated and are presented in Appendix B. Process data are summarized in Figures 3-1, 3-2, 3-3, 3-4, 3-5, 3-6, 3-7, and 3-8.

Occurrences during operations that may have affected the recorded data are listed below:

    1.    During Run 1, a computer problem deleted the plant process data
            that was being automatically collected.  However, other logs with
            similar data have been used to re-construct a record of the plant
            operation during Run 1.  
    2.    Data included in Appendix B, but not shown in graphical form, for
            the scrubber inlet temperatures for all three runs showed a 100 degF
            lower value for the absorber module 'A' than for the other three
            modules.  The lower value was assumed to be in error, presumably
            because a sensor was defective or improperly placed.

The following problems and/or Variations occurred during the test:

    1.    During Run 1 (10:30 a.m. to 2:30 p.m., 7/27/93), the plant operated
            at a steady state without any notable problems.  Process data were
            actually recorded until 3:45 p.m. to allow completion of flow
            measurement testing. 
    2.    During Run 2 (10:15 a.m. to 2:15 p.m., 7/28/93), the plant operated
            at a steady state without notable problems.          
    3.    During Run 3 (9:45 a.m. to 1:45 p.m., 7/29/93), the plant operated
            at a steady state without notable problems.     Calculations -- The accompanying data sheets in Appendix B summarize the calculated average values of the differential pressure across, and inlet temperature to, the ESP. Also explained below are the calculations for the average corona power input, total average corona power input, total average corona power density, and the average cell corona power density.

Ignoring any pressure contribution by the primary air system, the average differential pressure across the ESP, and associated ductwork, is equal to the absolute value of the difference between the average ID fan inlet pressure (in.wg) and the average SAH pressure (in.wg). The ESP inlet temperature (degF) is equal to the SAH average gas temperature (degF) multiplied by 0.85 plus the average of the PAH 'A' gas temperature (degF) and the PAH 'B' gas temperature (degF) multiplied by 0.15. This second computation is weighted because of the gas flow differences of the SAH and the PAH.

The Secondary currents and Secondary voltages, measured at 15 minute intervals during each emission test, were averaged. These averaged numbers were multiplied together to obtain the average corona power input used by an ESP cell during the emissions test.

The average corona power input for each cell was used in two calculations. In the first, all average corona power inputs were summed to obtain the total average corona power input, which was then divided by the total plate area (ft2) to obtain the total average corona power density. In the second calculation, the average corona power input (for each cell) was divided by each cell's plate area. This resulted in the average cell corona power density and, with the total average corona power density, is displayed in Figures 3-9, 3-10 and 3-11, respectively.

Since the total collection area of the ESP was known, the collection area for each field was determined as follows.

  Total Collection Area    - 965,888 ft2 (West side only)
  The collection area for each field would then be:
    965,888 ft2 / 20 cells = 48,294.4 ft2  
    The number of plates in each field was not known.  It was assumed to be
    equal for the purposes of Figures 3-9, 3-10, and 3-11.


The FTIR analysis is done using two experimental techniques. The first, referred to as direct gas phase analysis, involves transporting the gas stream to the sample manifold so it can be sent directly to the infrared cell. This technique provides a sample similar in composition to the flue gas stream at the sample point location. Some compounds may be affected because of contact with the sampling system components or reactions with other species in the gas. A second technique, referred to as sample concentration, involves concentrating the sample by passing a measured volume through an absorbing material (Tenax®) packed into a U-shaped stainless steel collection tube. After sampling, the tube is heated to desorb any collected compounds into the FTIR cell. The desorbed sample is then diluted with nitrogen to one atmosphere total pressure. Concentrations of any species detected in the absorption cell are related to flue gas concentrations by comparing the volume of gas collected to the volume of the FTIR cell. Desorption into the smaller FTIR cell volume provides a volumetric concentration related to the volume sampled. This, in turn, provides a corresponding increase in sensitivity for the detection of species that can be measured using Tenax®.

Infrared absorbance spectra of gas phase and concentrated samples were recorded and analyzed. In conjunction with the FTIR sample analyses, measurements of (HC), (CO), (O2), and (CO2) were obtained using gas analyzers. Components of the emission test systems used by Entropy for this testing program are described below.


An extractive system, depicted in Figure 4-1, was used to transport the gas stream from the sample location to the infrared cell.

4.1.1 Sampling System

Flue gas was extracted through a stainless steel probe. A Balston particulate filter rated at 1 micron was installed at the inlet of the sample probe (in-stack). Teflon® sample line (3/8-inch O.D.) was used to connect the probe outlet to the heated sample pump (KNF Neuberger, Inc. model number N010 ST.111) located inside the mobile laboratory. The temperature of the sampling system components was maintained at about 300 degF. Digital temperature controllers were used to control and monitor the temperature of the transport lines. All connections were wrapped with electric heat tape and insulated to ensure that there were no "cold spots" in the sampling system where sample might condense. All components of the sample system were constructed of Type 316 stainless steel or Teflon®. The heated sample flow manifold, located in the FTIR truck, included a secondary particulate filter and valves that allowed the operator to send sample gas directly to the absorption cell or through a gas conditioning system.

The extractive system can deliver three types of samples to the absorption cell. Sample sent directly to the FTIR cell is considered unconditioned, or "hot/wet." This sample is thought to be most representative of the actual effluent composition. The removal of water vapor from the gas stream before analysis is sometimes desirable; therefore, a second type of sample was provided by directing gas through a condenser system. The condenser employed a standard Peltier dryer to cool the gas stream to approximately 38 degF. The resulting condensate was collected in two traps and removed from the conditioning system with peristaltic pumps. This technique is known to leave the concentrations of inorganic and highly volatile compounds very near to the (dry-basis) stack concentrations. A third type of sample was obtained by dilution. The cell was partially filled with sample gas and the partial pressure of the sample was recorded. The cell was then filled to ambient pressure using dry nitrogen. The procedure could also be reversed with the nitrogen being introduced to the cell first. A dilution factor of 2:1 significantly reduced spectral interference from water vapor without removing any species from the sample while minimizing the possibility of reducing HAP concentrations below detectible levels. Lowering the water vapor concentration, in addition to protecting the absorption cell components, relieved spectral interferences, which could limit the effectiveness of the FTIR analysis for particular compounds.

4.1.2 Analytical System

The FTIR equipment used in this test consists of a medium-resolution interferometer, heated infrared absorption cell, liquid nitrogen cooled mercury cadmium telluride (MCT) broad band infrared detector, and computer (Figure 4-2). The interferometer, detector, and computer were purchased from KVB/Analect, Inc., and comprise their base Model RFX-40 system. The nominal spectral resolution of the system is one wavenumber (1 cm-1). Samples were contained in a model 5-22H infrared absorption cell manufactured by Infrared Analysis, Inc. The inside walls and mirror housing of the cell were Teflon® coated. Cell temperature was maintained at 240 degF using heated jackets and temperature controllers. The absorption path length of the cell was set at 22 meters.

4.1.3 Sample Collection Procedure

One gas phase test run was performed at each location (ESP inlet, ESP outlet and FGD outlet) concurrent with a sample concentration run. Table 3-1 presents the test schedule. Gas phase analysis was not performed at the stack. During a run, flue gas continuously flowed through the heated system to the sample manifold in the FTIR truck. A portion of the gas stream was diverted to a secondary manifold located near the inlet of the FTIR absorption cell (Figure 4-2). The cell was filled with sample to ambient pressure and the FTIR spectrum recorded. After analysis, the cell was evacuated so that a subsequent sample could be introduced. The process of collecting and analyzing a sample, then evacuating the cell to prepare for the next sample required less than 10 minutes. During each run, at least 12 gas phase samples were analyzed.


Sample concentration was performed using the adsorbent material Tenax®, followed by thermal desorption into the FTIR cell. The sample collection system employed equipment similar to that of the Modified Method 5 sample train.

4.2.1 Sampling System

Figure 4-3 shows the apparatus used in this test program. Components of the sampling train included a heated stainless steel probe, heated filter and glass casing, stainless steel air-cooled condenser, stainless steel adsorbent trap in an ice bath, followed by two water-filled impingers, one knockout impinger, an impinger filled with silica gel, a sample pump, and a dry gas meter. All heated components were kept at a temperature above 120 degC to ensure no condensation of water vapor within the system. The stainless steel condenser coil was used to pre-cool the sample gas before it entered the adsorbent trap. The trap was a specially designed stainless steel U-shaped collection tube filled with 10 g of Tenax® and plugged at both ends with glass wool. Stainless steel was used for the construction of the adsorbent tubes because it gives a more uniform and more efficient heat transfer than glass.

A sample run lasted 4-hours at approximately 0.12 to 0.16 dcfm for a total sampled volume of about 30 to 40 dcf. The rate depended on the sampling train used and was close to the maximum that could be achieved. Collection times provided a volumetric concentration that was proportional to the total volume sampled. The resulting increase in sensitivity allowed detection to concentrations below 1 ppm for some HAPs.

4.2.2 Analytical System

Before analysis, condensed water vapor was removed from the collection tubes using a dry nitrogen purge for about 15 minutes. Sample analysis was performed using thermal desorption-FTIR. The sample tubes were wrapped with heat tape and placed in an insulated chamber. One end of the tube was connected to the inlet of the evacuated FTIR absorption cell. The same end of the tube that served as the inlet during the sample run served as the outlet for the thermal desorption. Gas samples were desorbed by heating the Tenax® to 250 degC. A preheated stream of UPC grade nitrogen was passed through the adsorbent and into the FTIR absorption cell. About 7 liters of nitrogen (at 240 degF) carried the desorbed gases to the cell and brought the total pressure of the FTIR sample to ambient pressure. The infrared absorption spectrum was then recorded. The purging process was repeated until no evidence of additional sample desorption was noted in the infrared spectrum.

4.2.3 Sample Collection Procedure

During each run, concentrated samples were collected simultaneously at the ESP inlet, ESP outlet and the stack. Table 3-1 presents the test schedule. A sample concentration apparatus was set up at each location and ambient samples were collected to ensure each train was uncontaminated. The procedure for obtaining the ambient sample is described in Section 5.4.1. Entropy performed leak checks of the system and the start time of each run was synchronized at all three locations. Sample flow, temperature of the heated box, and the tube outlet temperature were monitored continuously and recorded at 10-minute intervals. At the end of each run, sampling was interrupted and the collection tube removed. The open ends were tightly capped and the tube was stored on ice until it was analyzed. The tubes were analyzed within 12 hours after the end of the sample run.


Entropy's extractive measurement system and the sampling and analytical procedures used for the determinations of SO2, NOx, HC, CO, O2, and CO2 conform with the requirements of EPA Test Methods 6C, 7E, 25A, 10, and 3A, respectively, of 40 CFR 60, Appendix B. A heated extractive sampling system and a set of gas analyzers were used to analyze flue gas samples extracted from each location. The analyzers received gas samples delivered from the same sampling system that supplied the FTIR cell. These gas analyzers require that the flue gas be conditioned to eliminate any possible interference (i.e., particulate matter and/or water vapor) before being transported and analyzed. All components of the sampling system that contact the gas sample were Type 316 stainless steel and Teflon®.

A gas flow distribution manifold downstream of the heated sample pump was used to control the flow of sample gas to each analyzer. A refrigerated condenser removed water vapor from the sample gas analyzed by all the analyzers except for the HC analyzer. (Method 25A requires a wet basis analysis.) The condenser was operated at approximately 38 degF. Condensate was continuously removed from the traps to minimize contact with the gas sample.

The sampling system included a calibration gas injection point immediately upstream of the analyzers for calibration error checks and also at the outlet of the probe for sampling system bias and calibration drift checks. The mid- and high-range calibration gases were certified by the vendor according to EPA Protocol 1 specifications. Methane in air was used to calibrate the HC analyzer.

A computer-based data acquisition system was used to provide an instantaneous display of the analyzer responses, compile the measurement data collected each second, calculate data averages over selected time periods, calculate emission rates, and document the measurement system calibrations.

Table 4-1 presents a list of the analyzers that Entropy used in this test program. Figure 4-1 presents a simplified schematic of Entropy's reference measurement system.

The test run values were determined from the average concentration measurements displayed by the gas analyzers during the run and are adjusted based on the zero and upscale sampling system bias check results using the equation presented in Section 8 of Method 6C. The CEM data are presented in Appendix A.


Flue gas flow rates were measured at the ESP inlet and outlet and the stack. The exhaust duct from the boiler branched into four ducts (labeled A - D in Figure 2-3) upstream of the ESP inlet location. Two ducts carried flue gas through each ESP building. The streams were recombined downstream of the ESP outlet location before passing through the FGD. During Run 1 Entropy collected velocity data using a 3-Dimensional (3-D) pitot probe at the inlet and outlet locations of the A and B ducts on the west-side ESP. This information indicated the potential for flow disturbances at the outlet. There were no indications of flow disturbances at the ESP inlet location; therefore, during Runs 2 and 3, the inlet flow determinations were made in accordance with EPA Methods 1, 2, and 3A using the S-type pitot probe. Entropy adopted the following plan, with the approval of Kintigh and EPA, to determine total flow on both sides of the ESPs. Table 4-2 summarizes the schedule of flow measurements.

      During Run 1 flow was measured using the 3-D probe through all 24
        ports (in A and B ducts) on the inlet and outlet of the west-side
      During Runs 2 and 3 flow was measured using the S-type pitot
        through all 12 ports (in the A and B ducts) on the inlet side of
        the west-side ESP.  
      During Run 2 flow was also measured using the 3-D probe through all
        24 ports (in the C and D ducts) on the inlet and outlet of the
        East-side ESP.  
      During Run 3 the S-type pitot was used to measure flow through all
        12 ports (in the A and B ducts) at the west side ESP inlet and the
        3-D was used to measure flow through all 12 ports (in the A and B
        ducts) at the West side ESP outlet.

Total flow, for each run, was determined by combining the flow through the west-side ESP (measured each test run) with the flow through the east- side ESP (measured only once during Run 2).

Gas flow was not measured at the breaching of the FGD outlet. This location did not meet Method 1 criteria and access was difficult for 3-D measurements. The stack location did meet Method 1 criteria and pre- and post-test flow data were obtained using the S-type pitot. With EPA approval, it was assumed that volumetric flow at the stack and the FGD outlet were the same. This assumption is reasonable because there are no obstructions between the FGD outlet and the stack locations.

The wet-bulb/dry-bulb technique was used for the measurement of the flue gas moisture. Pitot traverse point locations and the measurements made at these points are presented in the data sheets included in Appendix A.

The 3-D probe was used to identify off-axial flow. Measurements were obtained at 7 points through each port (42 data points for each duct) to establish a flow profile in the sampling region.

During the sampling runs, an S-type pitot tube was positioned adjacent to the point where the sample concentration probe was inserted. Single point P values were recorded at 10 minute intervals to verify that flow characteristics, at the sampling point, were not changing significantly during the test run.


RTI collected process data during the test. Results are included in Section 3.3.3.


4.6.1 Description of K-Matrix Analyses

K-type calibration matrices were used to relate absorbance to concentration. Several descriptions of this analytical technique can be found in the literature2. The discussion presented here follows that of Haaland, Easterling, and Vopicka[3].

For a set of m absorbance reference spectra of q different compounds over n data points (corresponding to the discrete infrared wavenumber positions chosen as the analytical region) at a fixed absorption pathlength b, Beer's law can be written in matrix form as

    A =   The n x m matrix representing the absorbance values of the m 
            reference spectra over the n wavenumber positions, containing
            contributions from all or some of the q components;
    K =   The n by q matrix representing the relationship between absorbance
            and concentration for the compounds in the wavenumber region(s) of
            interest, as represented in the reference spectra.  The matrix
            element  Knq = banq, where anq is the absorptivity of the qth
            compound at the nth wavenumber position;
    C =   The q x m matrix containing the concentrations of the q compounds
            in the m reference spectra;
    E =   The n x m matrix representing the random "errors" in Beer's law for
            the analysis; these errors are not actually due to a failure of
            Beer's law, but actually arise from factors such as
            misrepresentation (instrumental distortion) of the absorbance
            values of the reference spectra, or inaccuracies in the reference
            spectrum concentrations.

The quantity which is sought in the design of this analysis is the matrix K, since if an approximation to this matrix, denoted by K, can be found, the concentrations in a sample spectrum can also be estimated. Using the vector A* to represent the n measured absorbance values of a sample spectrum over the wavenumber region(s) of interest, and the vector C to represent the j estimated concentrations of the compounds comprising the sample, C can be calculated from A* and K from the relation

Here the superscript t represents the transpose of the indicated matrix, and the superscript -1 represents the matrix inverse.

The standard method for obtaining the best estimate K is to minimize the square of the error terms represented by the matrix E. The equation represents the estimate K which minimizes the analysis error.

Reference spectra for the K-matrix concentration determinations were de- resolved to 1.0 cm-1 resolution from existing 0.25 cm-1 resolution reference spectra. This was accomplished by truncating and re-apodizing4 the interfer- ograms of single beam reference spectra and their associated background interferograms. The processed single beam spectra were recombined and converted to absorbance (see Section 4.3).

4.6.2 Preparation of Analysis Programs

To provide quantitative results, K-matrix input must include absorbance values from a set of reference spectra which, added together, qualitatively resemble the appearance of the sample spectra. For this reason, all of the Multicomponent analysis files included spectra representing interferant species and criteria pollutants present in the flue gas.

Several factors affect the detection and analysis of an analyte in the stack gas matrix. One is the composition of the stack gas. The major spectral interferants in the coal-fired boiler effluent are water and CO2. At CO2 concentrations of about 10 percent and higher, weak absorbance bands that are normally not visible begin to emerge. Some portions of the FTIR spectrum were not available for analysis because of interference from water and CO2, but most compounds exhibit at least one absorbance band that is suitable for analysis. Significant amounts of SO2, NO, and NO2 were also present in the samples and these species were accounted for in the analysis. A second factor is the number of analytes to be detected because the program becomes more limited in distinguishing overlapping bands as the number of species increases. A third factor depends on how well the sample spectra can be modeled. The best analysis can be made when reference spectra are available to account for all of the species detected in the sample. If reference spectra for a major sample component are unavailable, then it may complicate the analysis of some species.

Before K-matrix analysis was applied to data, all of the spectra were visually inspected. Program files included reference spectra of the detected species were prepared and used to calculate flue gas concentrations. Four baseline subtraction points were specified in each analytical region, identifying an upper and a lower baseline averaging range. The absorbance data in each range were averaged, a straight baseline was calculated through the range midpoint using the average absorbance values, and the baseline was subtracted from the data prior to K-matrix analysis.

4.6.3 Error Analysis of data

The principal constituents of the gas phase samples were water, CO2, SO2, NO, and NO2. A program file was prepared to quantify these compounds. Other than these species and N2O no major absorbance features were observed in the gas phase data. After concentrations of the main constituents were determined, the appropriate standard was scaled and subtracted from the spectrum of the sample mixture. This helped verify the calculated concentrations and generated a base line by successively subtracting scaled standard spectra of water, CO2, SO2, NO, and NO2. The resulting "subtracted" spectrum was analyzed for HAPs and used to calculate maximum possible concentrations for undetected HAPs.

Maximum possible (minimum detectible) concentrations were determined in several steps. The noise level in the appropriate analytical region was quantified by calculating the root mean square deviation (RMSD) of the baseline in the subtracted spectrum. The RMSD was multiplied by the width (in cm-1) of the analytical region to give an equivalent "noise area" in the subtracted spectrum. This value was compared to the integrated area of the same analytical region in a standard spectrum of the pure compound. The noise was calculated from the equation:

    RMSD  =   Root mean square deviation in the absorbance values within a
       n  =   Number of absorbance values in the region.
      Ai  =   Absorbance value of the ith data point in the analytical
     AM   =   Mean of all the absorbance values in the region.

If a species is detected, then the error in the calculated concentration is given by:

    Eppm  =   Noise related error in the calculated concentration, in ppm.
      x2  =   Upper limit, in cm-1, of the analytical region.
      x1  =   Lower limit, in cm-1, of the analytical region.
     AreaR =  Total band area (corrected for path length, temperature, and
                pressure) in analytical region of reference spectrum of
                compound of interest.
    CONR  =   Known concentration of compound in the same reference

This ratio provided a concentration equivalent of the measured area in the subtracted spectrum. For instances when a compound was not detected, the value Eppm was equivalent to the minimum detectible concentration of that (undetected) species in the sample.

Some concentrations given in Tables 3-4 to 3-6 are relatively high (greater than 10 ppm) and there are several possible reasons for this.

      The reference spectrum of the compound may show low absorbance at
        relatively high concentrations so that its real limit of detection
        is high.  An example of this may be acetonitrile.  
      The region of the spectrum used for the analysis may have residual
        bands or negative features resulting from the spectral subtraction. 
        In these cases the absorbance of the reference band may be large at
        low concentrations, but the RMSD is also large (see Equation 7). 
        Condenser spectra give lower values because it is easier to perform
        good spectral subtraction on the spectra of the dryer samples.
      The chosen analytical region may be too large, unnecessarily
        including regions of noise where there is no absorbance from the
        compound of interest.

In the second and third cases the calculated maximum possible concentration may be lowered by choosing a different analytical region, generating better subtracted spectra, or by narrowing the limits of the analytical region. Entropy has taken these steps to minimize the calculated values in Tables 3-4 to 3-7.

4.6.4 Concentration Correction Factors

Calculated concentrations in sample spectra were corrected for differences in absorption pathlength between the reference and sample spectra according to the following relation:


    Ccorr  =  The pathlength corrected concentration.
    Ccalc =   The initial calculated concentration (output of the Multicomp
                program designed for the compound)
       Lr =   The pathlength associated with the reference spectra.
       Ls  =  The pathlength (22m) associated with the sample spectra.
       Ts  =  The absolute temperature of the sample gas (388 K).
       Tr  =  The absolute gas temperature at which reference spectra were
                recorded (300 to 373 K).

Corrections for variation in sample pressure were considered, and found to affect the indicated HAP concentrations by no more that one to two percent. Since this is a small effect in comparison to other sources of analytical error, no sample pressure corrections were made.

4.6.5 Analysis of Sample Concentration Spectra

Sample concentration spectra were analyzed in the same way as the gas phase samples. To derive flue gas concentrations it was necessary to divide the calculated concentrations by the concentration factor (CF). As an illustration, suppose that 10 ft3 (about 283 liters) of gas were sampled and then desorbed into the FTIR cell volume of approximately 8.5 liters to give concentration factor of about 33. If some compound was detected at a concentration of 50 ppm in the cell, then its corresponding flue gas concentration was about 1.5 ppm. The volume of flue gas sampled was determined from the following equation:


    Vflue =   Total volume of flue gas sampled.
    Vcol  =   Volume of gas sampled as measured at the dry gas meter after
                it passed through the collection tube.
    Tflue =   Absolute temperature of the flue gas at the sampling location.
    Tcol  =   Absolute temperature of the sample gas at the dry gas meter.
       W  =   Fraction (by volume) of flue gas stream that was water vapor.

The concentration factor, CF, was then determined using Vflue and the volume of the FTIR cell (Vcell) which was measured at an absolute temperature (Tcell) of about 300 K:

Finally, the in-stack concentration was determined using CF and the calculated concentration of the sample contained in the FTIR cell, Ccell.


Quality control (QC) is defined as the overall system of activities designed to ensure a quality product or service. This may include routine procedures for achieving prescribed standards of performance in the monitoring and measurement process. Quality assurance (QA) is defined as a system of activities that provides a mechanism of assessing the effectiveness of the quality control procedures. It is an integrated program for assuring the reliability of monitoring and measurement data.

The specific internal quality assurance and quality control procedures used during this test program are described in this section. Each procedure was an integral part of the test program activities.


This section details the QC procedures that were followed during the manual testing activities.

5.1.1 Velocity/Volumetric Flow Rate QC Procedures

The QC procedures for velocity/volumetric flow rate determinations followed guidelines set forth by EPA Method 2. Incorporated into this method are sample point determinations by EPA Method 1. Gas moisture content was approximated using the wet bulb-dry bulb technique.

The following QC steps were followed during these tests:

             The S-type pitot tube was visually inspected before sampling.
             Both legs of the pitot tube were leak checked before and after
             Proper orientation of the S-type pitot tube was maintained while
               making measurements.  The roll and pitch axis of the S-type pitot
               tube was maintained at 90ø to the flow.
             The magnehelic set was leveled and zeroed before each run.
             The pitot tube/manometer umbilical lines were inspected before and
               after sampling for leaks and moisture condensate (lines were cleared
               if found).
             Cyclonic or turbulent flow checks were performed prior to testing the
             Reported duct dimensions and cross-sectional duct area were verified
               by on-site measurements.  
             If a negative static pressure was present at sampling ports, checks
               were made for air in-leakage at the sample port which could have
               resulted in possible flow and temperature errors.  Leaks were sealed
               when found.
             The stack gas temperature measuring system was checked by observing
               ambient temperatures prior to placement in the stack.

The QC procedures that were followed in regards to accurate sample gas volume determination are:

             The dry gas meter is fully calibrated every 6 months using an EPA
               approved intermediate standard. 
             Pre-test and post-test leak checks were completed and were less than
               0.02 cfm or 4 percent of the average sample rate.
             The gas meter was read to a thousandth (.001) of a cubic foot for the
               initial and final readings.
             Readings of the dry gas meter, meter orifice pressure ( H), and meter
               temperatures were taken every 10 minutes during sample collection. 
             Accurate barometric pressures were recorded at least once per day.
             Post-test dry gas meter checks were completed to verify the accuracy
               of the meter full calibration constant (Y).

5.1.2 Sample Concentration Sampling QC Procedures

QC procedures that allowed representative collection of organic compounds by the sample concentration sampling system were:

             Only properly cleaned glassware and prepared adsorbent tubes that had
               been sealed with stainless steel caps were used for sampling.
             The filter, Teflon®  transfer line, and adsorbent tube were maintained
               at +-10 degF of the specified temperatures.    
             An ambient sample was analyzed for background contamination.
             Clean sample tubes were analyzed for contamination prior to their use
               in testing.

5.1.3 Manual Sampling Equipment Calibration Procedures Type-S Pitot Tube Calibration -- EPA has specified guidelines concerning the construction and geometry of an acceptable Type-S pitot tube. If the specified design and construction guidelines are met, a pitot tube coefficient of 0.84 is used. Information pertaining to the design and construction of the Type-S pitot tube is presented in detail in Section 3.1.1 of EPA document 600/4-77-027b. Only Type-S pitot tubes meeting the required EPA specifications were used. The pitot tubes were inspected and documented as meeting EPA specifications prior to field sampling. Temperature Measuring Device Calibration -- Accurate temperature measurements are required during source sampling. The bimetallic stem thermometers and thermocouple temperature sensors used during the test program were calibrated using the procedure described in Section 3.4.2 of EPA document 600/4-77-027b. Each temperature sensor is calibrated at a minimum of three points over the anticipated range of use against a NIST-traceable mercury-in-glass thermometer. All sensors were calibrated prior to field sampling. Dry Gas Meter Calibration -- Dry gas meters (DGMs) were used in the sample trains to monitor the sampling rate and to measure the sample volume. All DGMs were fully calibrated to determine the volume correction factor prior to their use in the field. Post-test calibration checks were performed as soon as possible after the equipment was returned as a QA check on the calibration coefficients. Pre- and post-test calibrations should agree within 5 percent. The calibration procedure is documented in Section 3.3.2 of EPA document 600/4-77-237b. 3-Dimensional Probe Calibration -- The following QC procedures were performed when using the 3-dimensional probe.

            The barometric pressure was recorded daily.
            The entire sampling system was leak checked prior to each run.  
            The direction of gas flow was determined before sampling.
            The angle finder was determined to be working properly.
            The manometers were leveled and zeroed every day. 
            The probe was positioned at the measurement point and rotated in the
              gas stream until zero deflection was indicated for the yaw angle;
              this null position occurs when P2 = P3.  Each yaw angle reading from
              the protractor or other angle measuring device was recorded.
            Holding the null reading position, readings were taken and recorded
              for the (P1 - P2) and (P4  P5).  Record the duct pressure or the (P2
              - Pbar).
            The procedure was repeated at each of the measurement points.
            The pitch angle was determined from the F1 calibration curve and F2
              was  determined from the F2 calibration curve.
            Calibration curves were generated from the procedures outlined in
              Draft Method 2E. 


The flue gas was analyzed for carbon monoxide (CO), oxygen (O2), carbon dioxide (CO2) and hydrocarbons (HC). Prior to sampling each day a pre-test leak check of the sampling system from the probe tip to the heated manifold was performed and was less than 4 percent of the average sample rate. Internal QA/QC checks for the CEM systems are presented below.

5.2.1 Daily Calibrations, Drift Checks, and System Bias Checks

Method 3A require that the tester : (1) select appropriate apparatus meeting the applicable equipment specifications of the method, (2) conduct an interference response test prior to the test program, and (3) conduct calibration error (linearity), calibration drift, and sampling system bias determinations during the test program to demonstrate conformance with the measurement system performance specifications. The performance specifica- tions are identified in Table 5-1.

A three-point (i.e., zero, mid-, and high-range) analyzer calibration error check is conducted before sampling by injecting the calibration gases directly into the gas analyzer and recording the responses. Zero and upscale calibration checks are conducted both before and after each test run in order to quantify measurement system calibration drift and sampling system bias. Upscale is either the mid- or high-range gas, whichever most closely approximates the flue gas level. During these checks, the calibration gases are introduced into the sampling system at the probe outlet so that the calibration gases are analyzed in the same manner as flue gas samples. Drift is the difference between the pre- and post-test run calibration check responses. Sampling system bias is the difference between the test run calibration check responses (system calibration) and the initial calibration error responses (direct analyzer calibration) to the zero and upscale calibration gases. If an acceptable post-test bias check result is obtained but the zero or upscale drift result exceeds the drift limit, the test run result is valid; however, the analyzer calibration error and bias check procedures must be repeated before conducting the next test run. A run is considered invalid and must be repeated if the post-test zero or upscale calibration check result exceeds the bias specification. The calibration error and bias checks must be repeated and acceptable results obtained before testing can resume.

Although not required by Methods 10 and 25A, the same calibration and data reduction procedures required by Method 3A were used for the CO and HC determinations to ensure the quality of the reference data.


Data quality audits were conducted using data quality indicators which require the detailed review of: (1) the recording and transfer of raw data; (2) data calculations; (3) the documentation of procedures; and (4) the selection of appropriate data quality indicators.

All data and/or calculations for flow rates, moisture content, and sampling rates were spot checked for accuracy and completeness.

In general, all measurement data were validated based on the following criteria:

            acceptable sample collection procedures
            adherence to prescribed QC procedures.

Any suspect data were identified with respect to the nature of the problem and potential effect on the data quality. Upon completion of testing, the field coordinator was responsible for preparation of a data summary including calculation results and raw data sheets.

5.3.1 Sample Concentration

The sample concentration custody procedures for this test program were based on EPA recommended procedures. Because collected samples were analyzed on-site, the custody procedures emphasized careful documentation of sample collection and field analytical data. Use of chain-of-custody documentation was not necessary. Instead, careful attention was paid to the sample identification coding. These procedures are discussed in more detail below.

Each sample concentration spectrum was assigned a unique alphanumeric identification code. For example, Tinl102A designates a desorption spectrum of a Tenax® sample taken at the ESP inlet during Run 1 using tube number 02. The A means this was the spectrum of the first desorption from this tube. Every collection tube was inscribed with a tube identification number.

The project manager was responsible for ensuring that proper custody and documentation procedures were followed for the field sampling, sample recovery, and for reviewing the sample inventory after each run to ensure complete and up-to-date entries. A sample inventory was maintained to provide an overview of all sample collection activities.

Every sample tube was cleaned and checked for contamination before use. The contamination check consisted of desorbing the clean tube and recording its FTIR spectrum. Each sampling train was checked for contamination before testing and after all testing was completed. The trains were set up according to the procedures of Section 4.2, except that the probe was not inserted into the port. Ambient air was drawn through the entire sample concentration apparatus for one hour (about 10 ft3 of air). This was sufficient to reveal significant contamination in the sampling system components. The charged ambient tube was stored and analyzed in the same manner as those obtained during test runs. If relatively minor contamination was identified from the ambient sample, it was accounted for in the subsequent analysis. Evidence of major contamination was not identified in any instance.

Sample flow at the dry gas meter was recorded at 10 minute intervals. Results from the analyzers and the spectra of the gas phase samples provided a check on the consistency of the effluent composition during the sampling period.

5.3.2 Gas Phase Analysis

During each test run a total of 12 gas phase samples were collected and analyzed. Each spectrum was assigned a unique file name and a separate data sheet identifying sample location and sampling conditions. A comparison of all spectra in this data set provided information on the consistency of effluent composition and a real-time check on the performance of the sampling system. Effluent was directed through all sampling lines for at least 5 minutes and the CEMs provided consistent readings over the same period before sampling was attempted. This requirement was satisfied any time there was a switch to a different conditioning system or a switch between testing locations. The FTIR continuously scanned when the cell was evacuating to provide a spectral profile of the empty cell. A new sample was not introduced until no residual absorbance from the previous sample was observed. The FTIR also continuously scanned during sample collection to provide a real-time check on possible contamination in the system.

5.3.3 FTIR Spectra

For a detailed description of QA/QC procedures relating to data collection and analysis, refer to the "Protocol For Applying FTIR Spectrometry in Emission Testing" in Appendix D. A spectrum of the calibration transfer standard (CTS) was recorded and a leak check of the FTIR cell was performed at the beginning and end of each data collection session. The CTS gas was 100 ppm ethylene in nitrogen. The CTS spectrum provided a check on the operating conditions of the FTIR instrumentation, e.g. spectral resolution and cell path length. Ambient pressure was recorded whenever a CTS spectrum was collected.

Two copies of all interferograms and processed spectra of backgrounds, samples, and the CTS were stored on separate computer disks. Additional copies of sample and CTS absorbance spectra were also stored for use in the data analysis. Sample spectra can be regenerated from the raw interfer- ograms, if necessary. FTIR spectra are available for inspection or re- analysis at any future date.

Pure, dry ("zero") air was periodically introduced through the sampling system in order to check for contamination or if condensation formed. Once, when water condensed in the FTIR manifold, the lines and cell were purged with dry N2, until sampling could continue.

As successive spectra were collected the position and slope of the spectral base line were monitored. If the base line within a data set for a run began to deviate by more than 5 percent from 100 percent transmittance, a new background was collected.


During the test program, it was the responsibility of the field coordinator and the sampling team members to ensure that all data measurement procedures were followed as specified and that data met the prescribed acceptance criteria. Specific procedures for corrective actions are described above.


Entropy conducted an emission test at NYSEG's Kintigh Unit 1 bituminous coal-fired electric generating station in Somerset, New York. Entropy performed direct gas phase analysis, and sample concentration testing over three days. Gas analyzers were used to measure CO, O2, CO2, and hydrocarbons in the gas streams. Three 4-hour sample concentrations runs were conducted at the ESP inlet, ESP outlet, and the stack with each run performed simultaneously at the three locations. One gas phase run was conducted at each of three locations, the ESP inlet, ESP outlet, and the FGD outlet. CEM measurements were performed simultaneously with the gas phase analyses. Each gas phase run was performed concurrently with a sample concentration run.

Gas phase analysis revealed the presence of water vapor, CO, SO2, CO2, and NOx. Sample concentration revealed the presence of a number of species. Details are given in Section 3.3.2.

A primary goal of this project was to use FTIR instrumention in a major test program to measure as many HAPs as possible or to place upper limits on their concentrations. Four other electric utilities were tested besides Kintigh. Utilities present a difficult testing challenge for FTIR techniques because: (1) they are combustion sources so the flue gas contains high levels of moisture and CO2 (both are spectral interferants) and (2) the large volumetric flow rates typical at these facilities lead to mass emissions above regulated limits even for HAPs at very low concentrations. This places demand on the measurement method to achieve low detection limits.

This represents the first attempt to use FTIR spectroscopy in such an ambitious test program. The program made significant achievements and demonstrated some important and fundamental advantages of FTIR spectroscopy as an emissions test method:

            Quantitative data were provided for a large number of compounds
              using one method.
            Software was written to provide concentration and detection limit
              results in a timely manner.  The same or similar software can be
              used for subsequent tests with very little investment of time for
              minor modifications or improvements.
            Preliminary data (qualitative and quantitative) is provided on-site
              in real time.
            With little effort at optimization (see below), detection limits in
              the ppb range were calculated for 25 HAPs and below 5 ppm for a
              total of 75 HAPs using direct gas phase measurements.  Sample
              concentration provided even lower detection limits for some HAPs.
            A positive identification of a compound is unambiguous.

It is appropriate to include some discussion about the "maximum possible concentrations" presented in Tables 3-4 to 3-6. These numbers were specifically not labeled as detection limits because use of that term could be misinterpreted.

In FTIR analysis maximum concentrations of non-detects are calculated from the spectra (see Section 4.6.3 and the "FTIR Protocol"). These calculated numbers are not fundamental measurement limits, but they give an indication of the measurement limits for the sampling and instrumental conditions used in this test. A number of factors influence the maximum concentrations:

Some instrumental factors

               Spectral resolution.
               Source intensity.
               Detector response and sensitivity.
               Path length that the infrared beam travels through the sample.
               Scan time.
               Efficiency of infrared transmission (through-put).
               Signal gain.

Some sampling factors

               Physical and chemical properties of compound.
               Flue gas composition.
               Flue gas temperature.
               Flue gas moisture content.
               Length of sample line (distance from location).
               Temperature of sampling components.
               Sample flow.

Instrumental components and settings can be chosen to optimize the measurement capability for the sampling conditions and for particular compounds of interest. For this program instrument settings were chosen to duplicate conditions that were successfully used in previous screening tests and the validation test. These conditions provide speed of analysis, durability of instrumentation, and the best chance to measure a large number of compounds with acceptable sensitivity. Sampling factors present similar challenges to any test method.

An additional consideration is that the numbers presented in Tables 3-4 to 3-6 are all higher than the true measurement limits that can be calculated from the 1 cm-1 data collected at TNP. This results from the method of analysis: the noise calculations were made only after all spectral subtractions were completed. Each spectral subtraction adds noise to the resulting subtracted spectrum. For most compounds it is necessary to perform only some (or none) of the spectral subtractions before the detection limit can be calculated. With more sophisticated software it will be possible to automate the process of performing selective spectral subtractions and optimizing the detection limit calculation for each compound of interest. (Such an undertaking was beyond the scope of the current project.) Furthermore, the maximum concentrations are averages compiled from the results of all the spectra collected at given location. A more realistic detection limit is provided by the single spectrum whose analysis gives the lowest calculated value. It would be more accurate to think of maximum concentrations as placing upper boundaries on the HAP detection limits provided by these data.

Perhaps the most important sampling consideration is the flue gas composition. In Table 3-4 the maximum benzene concentration is reported as 6.83 ppm. This was calculated by measuring the noise in the analytical region between 3020 and 3125 cm-1, where benzene has an absorbance band. Benzene exhibits a much stronger infrared band at 673 cm-1, but this band was not used for the analysis because absorbance from CO2 strongly interfered in this analytical region. Using identical sampling components and FTIR instrumental settings at a lower CO2 emission source would provide a detection limit below 1 ppm for benzene for direct gas analysis (even ignoring the consideration discussed above).

There may be some question as to why certain species were not detected in the direct gas measurements, particularly HCl. From Tenax® results it can be concluded that the HCl concentration was at least 611 ppb. But HCl results from sample concentration measurements are not quantitative because some HCl is lost in the condensed water that passes through the sample tube.

Table 2-1 presents the coal sampling analysis data (supplied to RTI by NYSEG) for coal used during the test. According to the analyses, the coal contained about .12 percent chlorine and about 1.9 to 2.0 percent sulfur (dry basis). The SO2 concentration before the scrubber was as high as 1300 ppm. If chlorine content of the coal translates in a similar way to a flue gas HCl concentration, then the HCl concentration before the scrubber may have been as high as 80 ppm. The scrubber should effectively remove HCl so any HCl concentration measured at the FGD outlet or the stack would have been much lower.

The HCl concentration may actually have been below 970 ppb as given in Table 3-4 for the ESP inlet. If HCl was present above 1 ppm, which is detectible by direct FTIR gas analysis, its measurement could have been affected by the moisture content of the flue gas. The sampling system configuration (including temperature of components) was chosen because the same configuration was used successfully at other tests. HCl was measured at utilities and other emission sources during the FTIR development project, and it was assumed that flue gas conditions at Kintigh would be similar. The moisture content at the ESP locations varied between 8.5 and 11.5 percent and was about 13.5 percent after the FGD and at the stack. This is higher than the 7 percent moisture experienced at the screening and validation tests. But, at least at the ESP locations, it should have been possible to measure HCl at 1 ppm and above. High moisture makes HCl measurements more difficult because of its solubility, but Entropy has detected HCl in gas streams with up to 30 percent moisture after using dilution.

The sample components and the FTIR cell were maintained above 300 degF and at 250 degF, respectively. Nothing would have been gained by using a higher sampling temperature because the flue gas was between 280 and 300 degF at the ESP locations, and only about 120 degF at the stack. The cell temperature was kept at 250 degF, because many of the reference spectra were collected at that temperature. This did not present a problem because no sample condensed in the cell.

Previous studies on HCl sampling and measurement in wet streams indicate that high sample flow rates help to deliver HCl to the measurement system. Entropy has participated in EPRI (Electric Power Research Institute) studies performing FTIR CEM measurements at utilities. In these studies HCl was measured using a sample flow of 10-15 lpm and heated lines at 300 degF. At the Kintigh test it was not possible to achieve a sample flow rate above 6-7 lpm because Entropy was also delivering sample to gas analyzers. In later FTIR field tests Entropy has performed QA spiking with HCl (and other compounds) to verify sampling system integrity, and this would be a good procedure to include in the test method. The important point to emphasize is that moisture presents a sampling system difficulty that any test method must address, not an analytical difficulty.


1) "FTIR Method Validation at a Coal-Fired Boiler," EPA Contract No.68D20163, Work Assignment 2, July, 1993.

2) "Computer-Assisted Quantitative Infrared Spectroscopy," Gregory L. McClure (ed.), ASTM Special Publication 934 (ASTM), 1987.

3) "Multivariate Least-Squares Methods Applied to the Quantitative Spectral Analysis of Multicomponent Mixtures," Applied Spectroscopy, 39(10), 73-84,1985.

4) "Fourier Transform Infrared Spectrometry," Peter R. Griffiths and James de Haseth, Chemical Analysis, 83, 16-25,(1986), P. J. Elving, J. D. Winefordner and I. M. Kolthoff (ed.), John Wiley and Sons,.

Jump to main content.