22 June 2026 · 7 min read

Isokinetic sampling, explained: why the sample velocity must match the flue

The reference methods behind your CEMS — Part 2 of 5. A series for industry and regulators, drawn from the DOE CEMS Guidelines (Version 8, 2025) we helped develop and the standards-committee work behind MS 1596.

In Part 1 we saw that your CEMS is anchored to a manual stack test — the Standard Reference Method. Now to the single idea that decides whether that stack test is any good: isokinetic sampling.

It sounds like jargon, but the principle is simple, and once you’ve seen it you’ll understand why a stack-test report always quotes an “isokinetic %” — and why a number in the wrong range means the whole test gets thrown out.

The one-sentence version

Isokinetic just means same speed. To pull a representative dust sample from a flue, the gas entering the sampling nozzle has to be travelling at the same velocity as the gas flowing past it in the duct. Match the speed, and the sample you collect genuinely represents what’s in the stack. Get the speed wrong, and your dust result is biased — before the sample ever reaches a balance.

Why speed changes the answer

The reason is inertia. Gas molecules are light and follow whatever path the flow takes. Dust particles are heavier, so they resist changing direction. That difference is the whole problem.

If the nozzle samples too slowly (sub-isokinetic), it can’t swallow all the gas heading toward it, so some gas spills around the nozzle mouth. The light gas curves away — but the heavier particles, carried by their own momentum, keep going straight into the nozzle. You collect more dust than is really there, and the result over-reads.

If the nozzle samples too fast (super-isokinetic), it pulls in extra gas from around the opening. That surrounding gas comes in, but many of its heavier particles have too much momentum to make the turn into the nozzle and sail straight past. You collect less dust than is really there, and the result under-reads.

Only when the nozzle velocity matches the flue velocity do the streamlines run straight into the mouth, particles and all — and the sample is true.

Sub-isokinetic, isokinetic and super-isokinetic sampling Three panels showing gas streamlines and dust particles entering a sampling nozzle. When the sample is too slow, streamlines diverge but heavy particles still enter, over-reading dust. When matched, streamlines run straight in for a true reading. When too fast, streamlines converge but heavy particles bypass the nozzle, under-reading dust. Too slow sub-isokinetic Over-reads dust extra particles pulled in Matched isokinetic True reading ✓ sample represents the flue Too fast super-isokinetic Under-reads dust heavy particles sail past
Light gas follows the streamlines; heavier dust resists turning. Only at matched (isokinetic) velocity does the sample collect dust in true proportion.

The number that decides validity

Because this bias is predictable, the reference-method standards turn it into a simple pass/fail check: the isokinetic ratio, reported as a percentage of the flue velocity. Sample exactly at flue speed and you’re at 100%.

Tests are only accepted inside a tight window around 100% — within ±10% of flue speed, a 90–110% band, under ISO 9096, with the exact figure set by the governing standard. (Those standards — MS 1596, ISO 9096 and EN 13284 — and how they fit together are the subject of Part 3.) Fall outside the window and the run isn’t “a bit off” — it’s invalid, and it has to be repeated.

How a tester actually hits it

Isokinetic sampling isn’t guesswork. Before and during the run, the tester:

  • measures the flue gas velocity with a pitot tube (and the temperature, moisture and pressure that affect it),
  • selects a nozzle size to suit, and
  • continuously adjusts the sampling pump so the velocity into the nozzle tracks the velocity in the duct — across every point of the traverse, as conditions shift.

Done well, the final isokinetic % lands comfortably inside the window. Done carelessly, an otherwise good test is wasted — and as we noted in the QAL2 article, sloppy reference-method work can even make a perfectly good CEMS fail its calibration.

Not one point, but a traverse

There’s a second reason a single reading won’t do. The flue isn’t uniform: gas moves faster down the centre of a duct and slower near the walls, and dust can be spread unevenly across the section. A sample drawn at one spot would represent that spot, not the stack. So the reference method samples a traverse — a set of points spread across the cross-section, each standing for an equal slice of it, with the isokinetic speed re-matched at every one.

Where those points sit isn’t left to the tester’s eye. ISO 9096 divides the measurement plane into equal areas and puts a sampling point at the centre of each. A circular duct is split into equal-area rings, so the points fall along two perpendicular diameters and bunch closer together near the wall, where each ring is thinner. A rectangular or square duct is split into an even grid, with one point per cell.

Traverse points in a circular versus a rectangular duct A circular duct is divided into equal-area rings with points along two diameters that cluster toward the wall, and the probe traverses them; a rectangular or square duct is divided into an even grid with a point at the centre of each cell. Circular duct probe equal-area rings — points cluster toward the wall (5 to 17 by size) Rectangular / square duct equal grid — one point per cell (4, 9 or 16 by area)
The SRM samples a traverse, not a single point. ISO 9096:2017 sets the points at the centres of equal areas — rings for circular ducts (Table 1), a grid for rectangular and square ducts (Table 2 / Annex B) — and the tester re-matches isokinetic speed at each.

How many points depends on the duct: a circular plane takes from 5 points on a small duct up to 17 on one over 2 m across (ISO 9096 Table 1), and a rectangular plane takes 4, 9 or 16 as its area grows (Table 2). A bigger plane means more points — and a longer test, because each point is its own isokinetic sample. The dust caught across all of them, over the full sampling time, is what becomes your reported concentration. It’s also why the stack needs access ports placed and sized for the probe at every traverse line (ISO 9096 recommends at least 125 mm diameter, or 100 × 250 mm).

Why it matters to you

If you’re paying for a stack test — for compliance, or to calibrate your CEMS at QAL2 — the isokinetic % is the first thing worth checking on the report. A figure outside the accepted range means a repeat visit: more cost, more downtime, and a delay to getting your CEMS signed off. It’s also a fair proxy for how carefully the whole test was run.

In Part 5 we walk through reading a stack-test report end to end — concentration, mass flow and that isokinetic % — so you can sanity-check the work you’ve paid for.

Booking a compliance stack test or a QAL2? Talk to us — we run isokinetic reference-method testing for facilities across Malaysia, to the same DOE CEMS Guidelines we helped write.

Related insights

This article is general guidance, not legal advice. For obligations specific to your facility, refer to the current Environmental Quality (Clean Air) Regulations 2014, the EQA 1974, the DOE CEMS Guidelines, and the current editions of MS 1596, ISO 9096 and EN 13284, or speak with us directly.

← All insights