Instrumental analyzer methods

EPA Methods 3A, 6C, 7E, 10 and 25A, 40 CFR Part 60, Appendix A

Instead of collecting a sample to weigh later, these methods pipe stack gas continuously to analyzers and read concentrations in real time. The numbers are only as good as the calibration checks that bracket every run.

Answers: What are the O₂, CO₂, NOₓ, SO₂, CO and hydrocarbon concentrations, minute by minute?
Equipment: Heated probe and line, gas conditioner, pump, analyzers, data system, EPA Protocol calibration gases
Produces: Bias-corrected average concentrations for each run, in ppmvd or percent
Feeds into: Emission rates with Method 2 flow, O₂-corrected limits, F-factor rates, CEMS audits

One procedure, many analyzers

Method 7E was written for nitrogen oxides, but its sampling and quality-assurance procedure became the template for the other gas methods. Methods 3A, 6C and 10 each name their own analyzer types and then point back to Method 7E for how to set up, calibrate, and check the system. Learn 7E once and you know most of the family.

Instrumental reference methods and common analyzer principles
Method	Measures	Typical analyzer principle
3A	O₂ and CO₂	Paramagnetic or electrochemical (O₂); nondispersive infrared, NDIR (CO₂)
6C	SO₂	Ultraviolet absorption, NDIR, or UV fluorescence
7E	NOₓ (NO + NO₂)	Chemiluminescence, with a converter that turns NO₂ into NO
10	CO	NDIR or gas filter correlation
25A	Total gaseous organics	Flame ionization detector (FID), calibrated with propane
320	Many gases at once (HCl, NH₃, formaldehyde and more)	Extractive Fourier transform infrared (FTIR), from 40 CFR 63 Appendix A

A chemiluminescence analyzer only sees NO: it reacts NO with ozone and counts the light given off. To report total NOₓ, the sample first passes through a heated converter that reduces NO₂ to NO. Method 7E requires showing the converter is at least 90% efficient.

The sampling system

An extractive system pulls gas out of the stack and delivers it clean and dry to the analyzers. Every part of the path can change the sample: a cold spot condenses water that dissolves SO₂ and NO₂, a leak dilutes the gas with air, and a dirty filter can react with what passes through it. That is why the method checks the whole system, not just the analyzers.

Probe and heated filter pull gas from the measurement point and remove particulate.
Heated sample line carries the gas to the trailer above its dew point so nothing condenses on the way.
Gas conditioner, a chiller or permeation dryer, removes moisture so the analyzers read on a dry basis. Hot, wet systems skip this step and report wet concentrations.
Pump and manifold keep a steady flow and split it between analyzers, venting the excess.
Data system logs every analyzer at short intervals and averages each run.

Where the gas goes

Switch between sampling and the two calibration paths.

Flow path

Calibration gases and span

Analyzers are checked against EPA Protocol gases: cylinders whose concentration is certified traceable to national standards. Before testing you choose a calibration span, the top of the range you will check, and three gases to go with it:

Low-level gas, under 20% of the span. Usually a zero gas such as nitrogen.
Mid-level gas, 40 to 60% of the span.
High-level gas, equal to the span. Its concentration defines the span.

Pick the span to suit the source: high enough that readings stay under it, low enough that the readings are not a sliver at the bottom of the range. A span ten times higher than the emissions makes every percent-of-span criterion ten times easier to pass, and ten times less meaningful.

Calibration error, system bias, and drift

Three checks bracket the data. Each is expressed as a percent of the calibration span.

Analyzer calibration error Once at the start, inject zero, mid and high gases straight into the analyzer. The response must match each cylinder’s certified value.
System bias Before and after every run, inject zero gas and an upscale gas (mid or high, whichever is closer to the stack concentration) at the probe. Compare the response through the whole system with the direct response for the same gas.
Drift Compare each run’s post-run bias response with its pre-run response. A big change means the analyzer moved during the run.

Analyzer calibration error, system bias, and drift (% of calibration span)

ACE = \frac{C_{Dir} - C_V}{CS}\times 100 \qquad SB = \frac{C_S - C_{Dir}}{CS}\times 100 \qquad D = \lvert SB_{final} - SB_{i}\rvert

C_V: certified concentration of the calibration gas
C_{Dir}: analyzer response when the gas goes directly to the analyzer
C_S: analyzer response when the gas goes through the whole sampling system
CS: calibration span

Calibration checks for one run

Push the analyzer, system and drift errors and watch which checks fail.

Calibration span, ppm NOₓ

Analyzer error at mid gas

System bias (sample loss or leak)

Drift during the run

Raw run average

Raw average

—

Bias-corrected

—

Each check also passes if the absolute difference is 0.5 ppmv or less, which matters on low spans: try the 10 ppm span, where 0.5 ppm is already 5% of span.

Bias-corrected run concentration (Method 7E, Equation 7E-5)

C_{Gas} = \left(\bar{C} - C_O\right)\frac{C_{MA}}{C_M - C_O}

\bar{C}: average analyzer reading over the run
C_O: average of the pre- and post-run system responses to the zero gas
C_M: average of the pre- and post-run system responses to the upscale gas
C_{MA}: certified concentration of that upscale gas

Method 7E calibration limits

Analyzer calibration error: within 2.0% of the calibration span, or 0.5 ppmv absolute.
System bias: within 5.0% of the span, or 0.5 ppmv, before and after each run.
Drift: within 3.0% of the span, or 0.5 ppmv, between pre- and post-run checks.
NO₂-to-NO converter efficiency: at least 90%.

A failed post-run system bias check invalidates the run. A failed drift check means recalibrating before the next run. Confirm the corrective actions in the current method text and your protocol.

Running the test

With the system calibrated, the probe goes into the stack. The system response time, measured during the first bias check, is how long the analyzer takes to settle after a step change in concentration. It sets the pace of the run: at each sampling point, wait at least two response times before counting data, so the reading reflects that point and not the last one.

Each analyzer also needs an interference check, showing that other gases in the stream, such as CO₂ or water vapor, do not shift its reading. It is typically done once per analyzer rather than every test.

Stratification: how many points?

Gas analyzers measure concentration, not velocity, so they do not need a full velocity-weighted traverse if the concentration is uniform across the stack. The stratification test decides. Sample a full Method 1 traverse (commonly 12 points) and compare each point with the average:

Every point within ±5% of the mean (or ±0.5 ppm; ±0.3% O₂ or CO₂): sample at a single point.
Every point within ±10% (or ±1.0 ppm; ±0.5% O₂ or CO₂): sample at 3 points on one line, at 16.7, 50.0 and 83.3% of the diameter. For stacks wider than 7.8 ft (2.4 m), use points 0.4, 1.2 and 2.0 m from the wall.
Otherwise the gas is stratified: sample at every point of the Method 1 traverse.

Stratification test

Increase the stratification and watch the decision change.

Gas

Pattern

Stratification

Mean

—

Largest deviation

—

Notice how O₂ passes more easily: the absolute allowance of 0.3% O₂ is larger than 5% of a 5% reading.

Total hydrocarbons: Method 25A

Method 25A uses a flame ionization detector. Organic compounds burned in a hydrogen flame produce ions, and the current is roughly proportional to the number of carbon atoms passing through. The FID cannot tell compounds apart, so the method reports a single total.

Propane is the calibration gas. Results are reported “as propane,” or multiplied by three and reported “as carbon.” Always check which basis your permit limit uses.
The sample stays hot and wet. Chilling would condense heavier organics, so the line and analyzer are heated and results are on a wet basis. Convert with Method 4 moisture if the limit is dry.
It counts methane too. If the limit is for non-methane organics, pair it with a methane measurement (for example Method 18 or an analyzer with a methane cutter).

Method 25A has its own calibration levels and limits that differ from Method 7E. Work from the current method text when setting up.

From ppm to permit units

Analyzers report concentration by volume. Permits usually want mass rates, concentrations at a reference O₂ level, or mass per unit of heat input. Three conversions cover most cases.

Mass concentration and mass rate (68 °F, 29.92 in. Hg)

C_{lb/dscf} = \frac{C_{ppmvd}\times MW}{385.3\times 10^{6}} \qquad E_{lb/hr} = C_{lb/dscf}\times Q_{sd}

385.3: ft³ occupied by one lb-mole of ideal gas at 68 °F and 1 atm
MW: molecular weight; NOₓ is reported as NO₂ (46.01)
Q_{sd}: dry standard flow rate from Method 2, dscf/hr

Correction to a reference O₂ level, and Method 19 heat-input rate

C_{@X\%O_2} = C\,\frac{20.9 - X}{20.9 - \%O_{2d}} \qquad E_{lb/MMBtu} = C_{lb/dscf}\,F_d\,\frac{20.9}{20.9 - \%O_{2d}}

X: reference O₂: 7% is common for boilers and incinerators, 15% for combustion turbines
F_d: dry F-factor: dscf of combustion gas per million Btu of fuel at 0% excess air

Unit converter

Enter a run result and see it in permit units.

Pollutant Concentration, ppmvd O₂, % dry Flow, dscfm Fuel, F_d in dscf/MMBtu

Reference O₂

Mass rate

—

At 7% O₂

—

Heat-input rate

—

Certifying a CEMS: the RATA

Many sources run a continuous emission monitoring system (CEMS) all year. To prove it reads correctly, a test crew runs the reference method side by side with it in a relative accuracy test audit. For SO₂ and NOₓ monitors, Performance Specification 2 (40 CFR 60 Appendix B) sets the procedure:

Run at least 9 paired runs, typically 21 minutes or longer each. You may run more and discard up to 3, but at least 9 must remain, and every run must be reported.
Compute the difference for each pair, the mean difference, and a confidence coefficient from the spread of the differences.
The relative accuracy usually must be no more than 20% of the mean reference method value. Low-emitting sources may qualify for an alternative limit based on the emission standard.

Relative accuracy (Performance Specification 2)

RA = \frac{\lvert\bar{d}\rvert + \lvert CC\rvert}{\overline{RM}}\times 100 \qquad CC = t_{0.975}\,\frac{S_d}{\sqrt{n}}

\bar{d}: mean of the differences between reference method and CEMS values
S_d: standard deviation of the differences
t_{0.975}: Student’s t at n − 1 degrees of freedom: 2.306 for 9 runs, 2.262 for 10, 2.228 for 11, 2.201 for 12

RATA calculator

Edit the run data or drop up to three runs.

Runs performed

Mean difference

—

Confidence coefficient

—

Relative accuracy

—

Acceptance criteria

Instrumental methods at a glance

EPA Protocol calibration gases: low under 20% of span, mid 40–60%, high equal to the span.
Analyzer calibration error within 2.0% of span (or 0.5 ppmv).
System bias within 5.0% of span (or 0.5 ppmv), before and after every run.
Drift within 3.0% of span (or 0.5 ppmv) across each run.
NO₂-to-NO converter efficiency of at least 90% (Method 7E).
At each point, sample at least two system response times before recording data.
Stratification: ±5% for a single point, ±10% for three points, otherwise a full traverse.
RATA (PS-2): at least 9 runs; relative accuracy generally no more than 20% of the mean reference value.

Field mistakes that cost runs

A cold spot in the heated line

Where the heat trace fails, water condenses and dissolves SO₂ and NO₂. Readings drop, and the post-run system bias check often catches it only after an hour of testing is lost.

Choosing a lazy span

A 500 ppm span on a 20 ppm source makes the checks easy to pass and the data hard to defend. Size the span to the emissions.

Calibrating at the analyzer only

Direct calibration proves the analyzer works. It says nothing about leaks, a saturated filter, or a failing conditioner. The system bias check through the probe is the one that protects the data.

Moving too fast between points

If the response time is 90 seconds, data from the first three minutes at a new point still belong partly to the old one.