Electronic leak detector: what the alarm does not prove

A leak detector answers one question. How much refrigerant is in the air being drawn through its probe at this moment. A pump pulls a sample past a sensor, and the sensor reports what it finds. Everything said afterwards is an interpretation laid on top of that single reading.

Air concentration and leak position are different quantities. INFICON explains the gap in the manual for its D-TEK Stratus. Refrigerant leaving a system does not spread evenly, and it is denser than air, so it gathers into clouds that sit low. Those clouds are colourless and odourless. A conventional detector alarms on walking into one, which the manual says does not help you find the leak, because the cloud may not be near its source.

The instrument also has to be told what counts as nothing. Testo's manual for the 316-4 describes zeroing as standardising the instrument to the current gas concentration. Whatever refrigerant already hangs in the air at that moment becomes the new zero. A detector zeroed inside a haze of escaped gas has quietly agreed to ignore it.

One tool can be set to answer either of two questions. Testo names the modes separately. Localisation mode optimises sensitivity to the change in the signal. Search mode optimises sensitivity to the leakage quantity. One setting chases a gradient, the other weighs how much is present, and the technician chooses which.

So a beep carries less information than its confidence suggests. It establishes that vapour the sensor responds to reached the tip during the sweep. It does not establish where that vapour started, how much is escaping, or whether the source is the joint the probe happened to be passing.

Sensitivity is a specification with conditions attached, and the conditions do more work than the number. Bacharach's datasheet for the H-10 PRO prints two figures for one instrument. Stationary, it is rated at 0.006 oz/yr. Moving, measured per SAE J2791, it is rated at 0.1 oz/yr.

That is a gap of roughly seventeen times, on the same page, for the same tool. The larger figure is the one that applies while a probe is being swept along a pipe, which is how the instrument gets used. A footnote adds further conditions. The rating covers a named list of refrigerants, at the small leak setting, in manual mode.

Background gas moves the number again. INFICON's specification table gives a minimum sensitivity of 1 g/yr for R134a in clean conditions. In a contaminated environment the same row reads greater than 2 g/yr. Refrigerant already in the air costs the instrument at least half its rating, and a leak hunt happens exactly where refrigerant is already in the air.

Which gas is escaping matters as well. Testo rates the 316-4 at under 3 g/a and qualifies that as specified for reference refrigerants. Its own table lists R410A as detectable, while the reference gas behind the threshold is R134a. The published figure was not established on R410A.

None of this makes the numbers dishonest. It makes a bare number useless. A figure quoted with no probe condition, no refrigerant and no setting behind it describes a laboratory, and the joint in question is on a ledge.

Cross sensitivity is designed for, not incidental. EN 14624 lists cross sensitivity from interference gases among the minimum requirements it specifies. A problem written into the scope of a standard is an old and well-understood one.

INFICON prints the evidence in its own manual. A table of common workshop chemicals marks which ones cause a false trigger on the D-TEK Stratus. Methanol-based windscreen washer solvent triggers it. So do a spot and stain remover, a rust penetrant, a gasket and trim adhesive, a cleaner and degreaser, a brake parts cleaner, a carburettor cleaner and clear silicone rubber.

Several other liquids do not. The same table marks antifreeze, brake fluid, silicone lubricant, automatic transmission fluid, mineral engine oil and a pumice hand cleaner as safe. Volatile solvents, aerosol degreasers and curing adhesives set it off. Heavier oils and water-based fluids largely do not.

That list carries weight because the D-TEK Stratus is an infrared instrument. Infrared is the sensing technology sold as the selective one, and it still responds to cleaning products. Bacharach's H-10 PRO uses a heated diode, which the datasheet describes as ultra-sensitive to halogen-based refrigerants, and a sensor that answers to halogens casts a wider net again. Testo's 316-4 uses a gas-sensitive semiconductor. Three technologies, three different response profiles, one shared weakness.

Nothing about a leak hunt keeps those substances away. A coil that has been chemically cleaned, a joint wiped with solvent, a bracket sealed with silicone, a cabinet sprayed with degreaser: each leaves residue that evaporates for a while afterwards. An alarm over that joint is a true reading about the air and a false reading about the system.

Two further triggers have nothing to do with chemistry at all. INFICON warns that high radio-frequency environments may cause a false alarm, and that blocking the exhaust port can cause false alarms or readings. An instrument can be wrong because of where it is standing and how it is being held.

A detector fails quietly. It does not stop working in a way anyone notices. It grows less sensitive, and a less sensitive instrument produces clean sweeps. Testo's troubleshooting page names the symptom plainly. An instrument that seems to ignore leaks, or misses some refrigerants, has a sensor whose service life is spent.

Service life is short and counted in running hours. Testo puts the 316-3 sensor at 80 to 100 hours, which it equates to about a year of normal use. Bacharach gives the H-10 PRO sensor a typical life of one year, and adds that consistent exposure to high levels of refrigerant deteriorates it faster. Finding leaks wears out the part that finds leaks.

Makers treat the sensing element as a consumable. INFICON's warranty excludes items that deteriorate under normal use, and names batteries, sensors and filters among them. The instrument carries a warranty. The part doing the detecting does not.

Dirt does the same damage faster. Testo warns that tobacco smoke, dirty air, oils, grease and evaporating liquids leave deposits on the sensor. The result it names is reduced sensitivity and a falsified concentration display. A separate caution warns of sensor destruction by substances such as oils. Refrigerant leaving a system carries oil out with it.

Warm-up is the other quiet gap. INFICON's D-TEK Stratus finishes its warm-up in 45 to 90 seconds, and the manual then recommends running it for 15 minutes to reach the specified sensitivities. A detector switched on at the door and used a minute later sits inside its working range and outside its rated one.

Makers answer all of this the same way, by checking against a known leak. INFICON gives a calibration frequency of an annual check with a calibrated leak standard, and Testo recommends yearly servicing at an authorised centre. That same handling code asks for periodic calibration. A beep reveals none of it, which is why the age of the tool is a fair thing to ask about.

The codified answer is to use two methods. The Australian and New Zealand handling code calls it best practice to combine techniques, and gives the example of an electronic detector to test an area, with leak detection spray to identify and verify the exact location. The detector narrows. Something else confirms.

That same code limits what a silent detector settles. Where a leak is suspected but not detected, it requires the refrigerant to be removed and the affected section tested for tightness. A sweep that found nothing is not a finding of no leak, and the code declines to let it close the question.

Open air is where the instrument is weakest. The code notes that plant located outdoors is a case where a hand-held leak detector may not function well, and that indirect detection covers equipment sitting in a well-ventilated environment. It also requires the procedure to account for the distance between the leak and the testing equipment. In Singapore the condenser usually lives on an open ledge, which is the condition being described.

Makers ask for a retest even when the alarm looked convincing. INFICON advises rechecking an apparent leak found at the most sensitive setting, particularly where the probe was held static against a joint, and repeating it with the probe moving. A probe pressed onto one spot is the easiest way to produce a hit that fails to survive a second look.

The useful question is never whether the detector alarmed. It is what happened next. A joint named from a beep alone, with no second method and no repeat, is a suspect promoted to a conclusion with the middle step missing.

That middle step is also what separates a repair from a refill. Refrigerant put into a circuit whose leak was never confirmed buys time and nothing else. Shrinking the search is the instrument's real job, and shrinking it only pays once something else has put a finger on the spot.