Flue gas analysis explained

Every commissioning sheet asks for the same flue gas numbers, and every gas engineer has watched a customer's eyes glaze while the analyser settles. This is what those numbers are, which of them have published limits, where 0.0040 comes from, and how to get them onto the record without a second admin session in the van.

What is flue gas?

Flue gas is what leaves the appliance once the gas has burned. Burn natural gas completely and you get carbon dioxide and water vapour, plus the nitrogen that came in with the combustion air and took no part in the fire, plus whatever oxygen the burner did not use. Nitrogen is the bulk of it, because it is roughly four fifths of the air the burner draws in, and it goes up the flue as ballast carrying heat with it. Carbon monoxide should be a trace, and on a properly set appliance it is.

CO appears when the carbon in the fuel finds too little oxygen, or too little heat and time, to finish the reaction. Complete combustion takes each carbon atom all the way to CO₂; incomplete combustion stops halfway at CO. That is why carbon monoxide is the reading that matters most for safety, and why it climbs when a burner drifts out of adjustment, a heat exchanger fouls, or the air supply is restricted.

The sample is read dry. Combustion makes water vapour as well as carbon dioxide, so gas in a working flue is wet, and a percentage measured with that water still in it would drift about with the flue temperature. The analyser conditions the sample on its way to the sensors, condensing the water out into the trap, so the figures on the screen describe the dry products of combustion. Manufacturers publish their commissioning data on the same dry basis, which is what makes the comparison like for like: a dry reading against a dry reference.

What does the analyser actually read?

An electronic combustion gas analyser (ECGA) reads how completely an appliance is burning its fuel. Four figures do the work: CO in ppm, CO₂ as a percentage, the ratio between them, and oxygen, with flue temperature alongside. The sample is taken where the manufacturer's instructions say, and the readings need to settle before they mean anything.

What are acceptable flue gas readings?

Acceptable means what the appliance manufacturer says is acceptable for that model, at the rate their instructions specify. That is the reference, and no generic table replaces it. Where the maker gives no figure, Gas Safe Register's TB 143 sets the fallback for commissioning a condensing boiler with an air/gas ratio valve: CO at or below 350 ppm and a combustion ratio at or below 0.0040, checked at maximum rate and again at minimum. Manufacturers routinely publish tighter numbers than that, and where they do, theirs are the ones to work to.

One appliance shows how specific the real figures get. Worcester Bosch's installation and maintenance instructions for the Greenstar 4000 combi specify, on natural gas, a CO₂ reading between 9.0 and 9.8% at maximum rated output and between 8.2 and 9.0% at minimum, with the minimum at least 0.2 lower than the maximum you actually measured, and CO under 250 ppm, all taken ten minutes after firing. The same book runs the TB 143 commissioning check at both rates, after a flue integrity check through the air inlet sample point with the casing on, looking for oxygen at or above 20.6% or CO₂ below 0.2%, and both values go on the Benchmark checklist. That 0.0040 line applies to a boiler with an air/gas ratio valve, which covers most modern condensing boilers. Another maker's boiler carries different numbers, which is the point.

What is the combustion ratio, and why 0.0040?

The combustion ratio is simply the measured CO divided by the measured CO₂. It's a sensitive indicator of combustion quality: as a burner goes out of adjustment or a heat exchanger fouls, CO climbs relative to CO₂ and the ratio rises.

• TB 143 and BS 7967 both put the combustion-ratio action level at 0.0040 for a condensing boiler with an air/gas ratio valve.
• At or below 0.0040, with the manufacturer's CO₂ figure met, combustion is within the accepted limit for a correctly operating appliance.
• Above 0.0040, follow the manufacturer's instructions and the relevant standard, investigate, service or take the appliance out of use as appropriate.
• Other appliance types have their own levels in BS 7967: 0.0080 for a central heating boiler without an air/gas ratio valve, 0.0200 for a water heater. Where the manufacturer specifies a figure, that figure applies.

Why the ratio, rather than raw CO, is the screening figure: dilution. Air mixed into the sample lowers CO and CO₂ together, so the ratio between them holds steady while either number alone can mislead. A low CO reading in a heavily diluted sample proves nothing; the ratio still tells the truth.

How do you work out the ratio?

Divide CO by CO₂ with both in the same units. With CO in ppm and CO₂ as a percentage, that is CO ÷ (10000 × CO₂). The analyser does the arithmetic for you, but it is worth knowing, because it shows how little CO it takes to move the number.

A boiler showing 90 ppm CO and 8.5% CO₂ gives 90 ÷ 85,000, which is 0.0011. That is comfortably inside the limit and an ordinary set of figures for a healthy appliance. The same boiler at 400 ppm with 9.0% CO₂ gives 400 ÷ 90,000, or 0.0044, over the action level and over the 350 ppm CO ceiling as well. The interesting case sits between the two: 350 ppm with 8.5% CO₂ works out at 0.0041, so CO is inside the TB 143 ceiling while the ratio is above 0.0040 and the appliance still needs investigating. That is the reading a ppm figure on its own would wave through.

How does a flue gas analyser work?

By pulling a sample of flue gas past sensors that produce a small current in proportion to the gas reaching them. The cells for carbon monoxide and oxygen are electrochemical: gas diffuses into the cell, reacts at an electrode, and the current that reaction produces is what the instrument reads. Carbon dioxide, on most analysers, is not measured at all. Gas Safe Register's TB 019 says so plainly, that the cells in most analysers extrapolate CO₂ from the CO and oxygen they have measured, which is why you tell the instrument which fuel you are burning. Newer instruments can measure CO₂ directly using infrared detection, and TB 019 notes those are now appearing in ECGAs.

Around the sensors sits the plumbing. A pump draws the sample through the probe and down the line, the probe carries a thermocouple for flue temperature, and a water trap takes the condensate out before the sample reaches the cells. That trap is not an accessory. Kane's manual for the KANE958 warns that the electrochemical sensors are affected by condensation or water drawn into the analyser, that the oxygen or carbon dioxide reading then shows as a dash, and that sensors may be permanently damaged; the instrument watches for it with a low flow detector that switches the pump off if water comes through from an over-filled trap.

Which is why the routine around a reading matters as much as the number it produces. The analyser is switched on in fresh air and left to zero itself, a minute on a KANE958, before it ever sees combustion products, and TB 143 asks for exactly that before a commissioning check. Cells also age and drift. Kane's guidance is to have the analyser recertified annually, and TB 143 expects an instrument of the correct type for BS 7967, maintained and calibrated as its own manufacturer specifies. An out-of-calibration analyser can pass a failing appliance, so keep the certificate with your records.

What is chimney and open-flue flue gas analysis?

Chimney is BS 5440's word for the structure that carries the products away, so on a UK domestic call it covers two quite different situations: sampling a room-sealed appliance through the test points on its flue, and sampling an open-flued appliance served by a chimney. The second is much the harder measurement.

The room-sealed case is the routine one, because the appliance is built for it. The Greenstar 4000 has two sampling points on the flue turret or adaptor: an air inlet point for the flue integrity check, and a flue gas point for the combustion readings, with the probe pushed to the centre of the flue gas exhaust and the cone sealing the point. The boiler goes into its own test mode for the readings, which Worcester calls chimney sweep mode, and the checks are taken at maximum output and again at minimum. The plugs go back in when you are done.

Open-flued appliances give you a sample worth arguing about. There may be no purpose-designed test point, the draught diverter is deliberately open to the room, and air drawn in at the diverter dilutes anything sampled beyond it. BS 7967 covers this ground for dwellings, and TB 019 records that it prescribes specialist multi-hole sampling probes for certain appliance types, available from the analyser manufacturer. Sample where the current procedure and the appliance manufacturer direct. Dilution is also why the ratio earns its keep here: air entering the sample pulls CO and CO₂ down together, so a comfortable ppm figure proves little while the ratio between them holds steady.

None of it replaces the flue checks. An analyser tells you how completely the burner is burning; it does not tell you the products are leaving the building, and on an open-flued appliance that question belongs to the flue flow and spillage tests. Flow is proved with a smoke pellet in the chimney, spillage with a smoke match at the edge of the draught diverter or fire canopy once the appliance has run at maximum heat input, five minutes unless the manufacturer says otherwise. Spillage testing follows the manufacturer's instructions or, where there are none, BS 5440-1, and TB 095 adds the fans problem: repeat the test with internal doors open, windows, external doors and adjustable vents closed and every extract fan running, because a kitchen fan fitted years later can pull the flue over. An appliance that spills fails, whatever the analyser said.

What does a high ratio actually tell you?

That combustion is incomplete, and something is making it so. The usual suspects: a burner out of adjustment, a fouled heat exchanger, a restricted air supply, a flue problem, or an appliance overdue a service. TB 019 sets out the loop. Examine the appliance, replace damaged parts with ones the manufacturer approves, service it to their instructions, then retest and see whether the readings improved. If they have not, the manufacturer's technical line is the next call, and where nothing further can be done the appliance is classified under the Gas Industry Unsafe Situations Procedure.

When do you take flue gas readings?

When the manufacturer's instructions and the current procedures say so, which for most modern appliances means at commissioning and at service, and whenever you're investigating suspected spillage or poor combustion. TB 143's own scope is narrower than engineers often assume: it covers commissioning a newly installed condensing boiler with air/gas ratio valve technology, and hands service and maintenance to the manufacturer's instructions and BS 7967. Measuring and recording CO and the combustion ratio on commissioning has been a requirement since 1 April 2014, the Benchmark checklist carries fields for both, and Gas Safe Register will defect an installation where they are missing. Manufacturers treat a completed Benchmark as a warranty condition too.

The analyser also sits alongside, not instead of, confirming heat input. On many modern appliances the combustion readings and the gas rate or burner pressure check are companions on the same visit, each catching what the other can miss.

How should the readings be recorded?

Flue gas readings belong on the appliance record, captured at the appliance while the analyser is still connected. Copying figures into the van later, or off a curling thermal printout at home, is where transposition errors creep in, and a transposed digit in a combustion ratio is not a small mistake. This is why Manifold's FlueScan reads CO, CO₂ and the ratio straight off the analyser screen with the phone camera for you to confirm, so the numbers on the record are the numbers the analyser showed.