Module datasheets are not written to a common standard. The same words — accuracy, response time, calibration-free — carry different meanings on different pages. Six checks turn them back into comparable numbers.

None of these checks requires equipment. They require the supplier to answer.

Check 1 — Accuracy: percent of what?

±1 % FS and ±1 % of reading are not the same specification.

On a 0–100 %vol range, ±1 % FS means ±1.0 %vol at every point. At a reading of 5 %vol that is a 20 % relative error. ±1 % of reading at 5 %vol is ±0.05 %vol.

Real datasheets rarely quote the same percentage on both sides. Compare a ±1 % FS module against a ±3 % of reading module on the same 0–100 %vol range and the two cross at 33 %vol. Below that point the reading-based specification is tighter; above it the full-scale one is. Which specification is better depends on where you operate.

0 1.0 2.0 3.0 Error band, ±%vol 0 25 50 75 100 Reading, %vol ±1 % FS → ±1.0 %vol everywhere ±3 % of reading Cross at 33 %vol At 5 %vol: ±1.0 vs ±0.15 %vol
At 5 %vol these two specifications differ by nearly a factor of seven, in favour of the reading-based one. They cross at 33 %vol, and above it the full-scale one is tighter. A percentage alone does not tell you which is better for your operating point.

Then ask the second question. Does one accuracy figure cover every gas and every range the module offers?

A module that lists four gases across eight ranges under one ±X % FS is claiming something the optics cannot deliver. A low range has a small full scale, so the same percentage becomes a much harder absolute error. Segmented accuracy — one figure below a breakpoint, a different figure above it — is the form that can actually hold.

Ask for

The denominator, the range each figure applies to, and the breakpoint between segments.

Check 2 — Detection limit: measured over how long?

A detection limit without an integration time cannot be compared.

Minimum detectable limit falls as averaging time rises. A module quoting 1 ppm at 60 s and a module quoting 5 ppm at 1 s can be the same hardware. The two numbers belong together. An Allan deviation plot shows where averaging stops helping and drift takes over — that turning point matters more than the headline figure.

Allan deviation Integration time τ Minimum Averaging removes noise (detection limit improves) Drift dominates (averaging makes it worse)
Averaging improves the detection limit until drift takes over. The turning point is the number that matters, and a headline figure on its own does not tell you where it is.
Ask for

The integration time behind each detection limit, and the point where the Allan deviation turns up.

Check 3 — Response time: a sum, not a property

T90 is a sum of three things, and only two of them belong to the module.

For an extractive installation, the time from a change at the process to a stable reading is: transport time down the sample line, plus displacement of the measurement cell volume, plus the algorithm’s averaging window.

Transport time is yours to calculate. A 10 m sample line of 4 mm internal diameter holds about 126 mL. At a sample flow of 1 L/min, gas takes about 7.5 seconds to reach the module. That time is set by your plumbing, not by the supplier.

Sample line transport Cell displacement Algorithm window Yours The module’s Worked example — 10 m line, 4 mm internal diameter, 1 L/min: line volume 126 mL ÷ 1 L/min ≈ 7.5 s before the module sees anything T90 at the analyser
Only the last two segments belong to the module. The first one is set by your sample line, and you can calculate it before you speak to any supplier.

Some in-situ datasheets state “no T90”. That means no transport lag, because there is no sample line. It does not mean the response is instantaneous. The cell still has to be displaced and the algorithm still has a window.

Ask for

T90 broken into its parts, and the sample flow rate it was measured at.

Check 4 — Pressure: which of the two numbers is this?

Two unrelated pressure specifications often appear in the same row.

Mechanical pressure rating is what the housing withstands. Pressure compensation range is the interval over which the module corrects the reading for gas density. Only the second one determines whether your number is right. A module can be mechanically safe and numerically wrong at the same time.

0 100 200 300 Absolute pressure, kPa Mechanical pressure rating — the housing holds Compensation range — the reading is corrected Dashed: housing is fine, reading is not corrected
A module can be mechanically safe across a range where its reading is no longer corrected. The two figures answer different questions and are frequently printed on the same row.

Check which pressure the range refers to. Absolute or gauge changes the envelope completely.

The gap is not small in practice. A Russian type-approved gas drainage parameter station is registered for 53.3 to 114.7 kPa absolute. A Chinese drainage parameter unit covers 10 to 200 kPa absolute. Both are correct for their design intent. They differ by an order of magnitude in vacuum depth, and a drainage line under suction sits in the part where they differ.

Ask for

The compensation range in absolute pressure, stated separately from the mechanical rating.

Check 5 — “Calibration-free”: free of what?

The term describes one specific property.

In wavelength modulation spectroscopy, 2f/1f normalisation makes the signal insensitive to losses that are not absorption: beam attenuation, vibration-induced misalignment, window fouling, particle scattering. In a dusty or vibrating installation that property is worth a great deal.

It does not mean the module needs no calibration. Calibration-free operation still depends on accurate spectroscopic line parameters and on accurate pressure and temperature measurement inside the module. Factory calibration is still performed with traceable reference gas.

Ask for

What the term covers on this specific module, and what the factory calibration procedure and interval are.

Check 6 — Are the test conditions there at all?

A number without its test conditions cannot be reproduced and cannot be compared.

Every performance figure needs five things attached: sample flow, temperature, pressure, background gas, and whether the module had finished warming up. Background gas in particular is not a detail. Move from nitrogen to a mixture carrying 40 % CO₂ and the cross-interference picture changes.

If a datasheet gives you a table of figures and no conditions block, you are not holding a specification.

Ask for

The test conditions for every figure you intend to design around.

The same six checks, applied to our own datasheets

We wrote this page, so it is fair to run the checks on us.

Denominator and segments

Our figures are segmented, not single-point. For the Pipeline Insertion TDLAS Methane Measurement Module: below 1.00 %vol, ±0.05 %vol; above 1.00 %vol, ±3.5 % of reading. The two segments meet at 1.00 %vol, where the low-segment figure governs — we state that rather than leave the overlap to interpretation.

One figure per gas and range

The Box-level Extractive TDLAS Gas Measurement Module carries four standard gases: CH₄, CO₂, CO and NH₃. Each gas and each range has its own two-segment figure. We do not publish one ±X % FS across all of them, because that specification cannot hold.

Reference conditions

Every figure on both datasheets carries the conditions it applies at.

For the extractive module: a sample flow of 1 L/min at 20 °C and 101.3 kPa, module at 25 °C, ambient pressure, nitrogen balance, after warm-up. T90 is 30 seconds at that same 1 L/min. Change the flow and the response time changes with it.

For the pipeline insertion module: 25 °C, 101.3 kPa absolute, methane in nitrogen balance, after warm-up. There is no sample flow, because the cell sits in the pipe.

Why the in-pipe module has no flow condition

Absorption depends on the number density of molecules along the optical path, and number density is set by pressure and temperature. Gas velocity does not enter that relationship. Once pressure and temperature are compensated, the concentration reading does not depend on how fast the gas moves past the window.

Velocity does decide how quickly the cell contents are replaced, which is a response time question. Our measurement window is 40 mm long, so at a pipe velocity of 0.1 m/s the gas in the window is replaced in 0.4 seconds. Our T90 of 5 seconds applies at 0.1 m/s and above, where the limiting term is the algorithm window rather than the flow. Below that velocity the gas in the window can lag the gas in the pipe, which is a question of how representative the reading is, not of how accurate it is.

Pressure

For the pipeline module, the compensation range is 20 to 200 kPa absolute, and we state it separately from the mechanical envelope rather than merging them into one row.

Where we do not clear the bar

Our own figure, against our own check

At a 22 %vol interlock threshold, ±3.5 % of reading produces an error band of ±0.77 %vol. That band sits on both sides of the threshold. Not crossing it would require ±2.27 % of reading. We publish the figure we can hold, not the one that would read better.