Refrigerant leak detection and recovery field guide for HVAC

Leak detection and recovery, and why they are one job

Refrigerant leak detection is finding where charge is escaping a sealed system. Recovery is pulling the refrigerant that is left into a recovery cylinder before you ever break a joint open. They read like two separate tasks, but on a real call they are one sequence, because you cannot legally or safely repair a leak without recovering first, and you cannot prove the repair without the leak check that closes it out.

The order almost never changes. Confirm the system is low and find the leak. Recover the remaining charge. Make the repair. Pressure test and evacuate to prove the system is tight and dry. Weigh in the new charge and verify it. Skip a step and you pay for it later, usually as a comeback on the same unit.

What makes this trade work different from a textbook is the temptation to shortcut it. The fast move is to top off the charge and drive away. That is not a repair. It is venting on a delay, because the refrigerant you added is on its way out the same hole, and now you have put more of it into the air. The whole guide below is about doing the slow version that actually holds.

What a leak actually costs

A leak is not just lost refrigerant. A system low on charge loses capacity and runs longer to do the same work, so the power bill climbs while the building stays warm. That part the owner notices. The part they do not notice is the compressor.

Run a system low on charge and the compressor loses the cool suction vapor that carries its motor heat away and the oil return that keeps it lubricated. It runs hot, the oil thins or stops coming back, and the compressor that was supposed to last fifteen years fails in three. Most of the dead compressors pulled on service calls did not die of old age. They died slowly, starved by a charge that was leaking out faster than anyone tracked.

Then there is the law and the air. Refrigerant venting is prohibited under the Clean Air Act, and the prohibition covers the HFCs, not just the old ozone-depleting refrigerants. The newer A2L refrigerants add a flammability angle on top: they are mildly flammable, so a leak in the wrong space is both a charge problem and an ignition concern. A leak is a money problem, an equipment problem, a legal problem, and on A2L systems a safety problem, all at once. Treat it like one.

What are the signs of a refrigerant leak?

A refrigerant leak shows up as low-charge symptoms first and physical evidence second. The system runs long, cools poorly, and the low side reads below where the chart says it should at the conditions you measured. Superheat climbs and subcooling drops on a system that is short of charge, which is the gauge pattern that points you toward a leak instead of an airflow problem.

The physical tells are the ones a pro looks for before reaching for a detector. Oil traces are the big one, because the refrigerant carries oil with it, and where the refrigerant leaks out the oil leaves a wet, dirt-collecting smudge at the joint, the valve, or the coil hairpin. Ice on the evaporator or at the metering device says the charge is low enough that the coil is starving and running below freezing. A hiss you can hear means a leak big enough that you do not need a detector to find the area.

The history is evidence too. A unit that has been topped off two or three times in as many seasons does not have a mystery. It has a leak nobody chased to ground. When the service record shows a recharge every spring, the job is not another top-off. The job is finding the hole.

Leak detection methods and when each one fits

There is no single best leak detector. There is the right one for the leak you have, the system state, and the refrigerant. The methods sort into electronic sniffing, bubbles, dye, and pressure testing, and most real calls use two of them: one to find the area and one to confirm the exact spot.

Match the method to the situation. If the system still holds charge, an electronic detector or bubbles works because there is refrigerant to find. If the system is empty or you just brazed it, you pressurize with dry nitrogen and run a standing pressure test, because there is nothing to sniff. If the leak is slow or intermittent and you cannot catch it live, UV dye buys you time by marking the spot over days of run time. The table sorts the trade-offs.

Electronic leak detectors: heated diode vs infrared

The two electronic detectors a tech actually carries are the heated diode and the infrared. They find the same leaks differently, and knowing which you are holding changes how you use it. A heated diode heats the sampled air enough to break the refrigerant molecules apart and reads the result, triggering on the concentration so the alarm speeds up as you close in on the leak. An infrared detector shines an infrared beam through the sampled air and reads the refrigerant by how it absorbs that light.

Infrared is the more stable sensor. Because the absorption is specific to the gas, an infrared detector throws fewer false alarms from oils and solvents, holds its accuracy across more refrigerants, and the sensor lasts longer. The heated diode is the cheaper tool and works well, but its sensor wears and commonly needs replacement every year or two, and it leans on the chlorine and fluorine content that the newer low-GWP refrigerants carry less of. For the A2L and low-GWP blends showing up now, infrared has the edge.

However you sweep, sweep slow and sweep low. Refrigerant is heavier than air and falls, so probe under the suspect joint, not over it. Move at roughly an inch per second, kill the breeze from fans and wind that disperses the trace, and when the detector lights up, confirm the exact spot with bubbles before you cut anything. An electronic detector finds the area. It does not always find the hole.

Soap bubbles and UV dye

Bubbles are the oldest method and still the one that confirms the spot. A bubble solution brushed on a joint that the detector flagged will foam up right at the leak, and unlike the detector it points to the exact pinhole, not the general area. Use a real leak-detection solution rather than dish soap, which can be corrosive on some metals over time, and remember it fails outdoors in wind and freezes solid in cold weather. On an accessible flare or valve, a quick bubble test is faster than fishing around with electronics.

UV dye solves the leak you cannot catch in the act. You inject a small amount of fluorescent dye compatible with the system oil, let the system run for hours or days, and the dye seeps out and dries at the leak where a UV lamp and glasses make it glow. It is the method for the slow seep that loses a pound a season, because you do not have to be standing there when it leaks. The cautions are real: the dye has to be compatible with the refrigerant and oil so you do not contaminate the system, it takes run time to show, and it makes a mess that has to be cleaned so the next read is not confused by old dye. Check the equipment manufacturer's stance, because some warranties and some sealed or specialty systems do not allow dye.

How do you do a nitrogen standing pressure test?

A nitrogen standing pressure test pressurizes the system with dry nitrogen, holds it, and watches the gauge over time to prove the system is tight. You use nitrogen and not refrigerant because nitrogen is dry, inert, cheap, and legal to release, and because there is no point sniffing for a leak when you can just pressurize and watch a needle. This is the method for an empty system and the proof you run after a repair.

Pressurize with dry nitrogen to the test pressure, commonly somewhere in the 200 to 600 psi range depending on the system and the refrigerant, and confirm the figure against the equipment manufacturer's rating so you do not overpressure a low-side component. Isolate the nitrogen source, note the pressure and the time, and leave it. A tight system holds. A leak shows as the pressure walking down over the hold. For the location, brush bubble solution on the joints under pressure and watch for foam.

Temperature is the trap that fools people. Nitrogen follows the gas laws, so when the space cools overnight the pressure falls on its own with no leak at all, and when the sun hits the unit the pressure rises. A pressure drop on a cold morning is often just the temperature, not a leak. Correct for it: either do the math from the temperature change or use a digital gauge with a temperature-compensated test mode that does it for you. Pressure that drops after you account for temperature is a real leak. Pressure that tracks the temperature and recovers when it warms back up is not.

Where do refrigerant leaks usually happen?

Refrigerant leaks concentrate at the joints, the valves, and the vibration points, not in the middle of a straight run of tube. Search the high-probability spots first and you find most leaks fast instead of sweeping the whole system inch by inch.

The usual suspects, roughly in order: flare connections, which back off and leak when they were over- or under-torqued or never had the seat inspected; brazed joints, where a cold joint or a pinhole from poor flow lets go; Schrader cores and valve caps, which are a five-minute fix and an embarrassing number of the leaks found; the coils, where formicary corrosion and rubbing eat through thin aluminum and copper hairpins; and the vibration points, where a line rubs a panel or a bracket until it wears through. Service valves and the access fittings round out the list.

On a big system, isolate before you sweep. If the unit has service valves, you can shut sections and pressure test them one at a time so you are not hunting a single leak across a hundred feet of pipe. Pin the problem to a section, then work that section hard. The slow, dumb way is to sniff everything. The fast way is to know where leaks live and start there.

Recovering the charge before you open the system

Before any joint comes apart, the refrigerant has to come out into a recovery cylinder. This is not optional and it is not a judgment call. EPA Section 608 requires recovery, and releasing the charge to the air to save twenty minutes is venting, which is illegal and which fines have followed. The tool is a recovery machine and a DOT-rated recovery cylinder, and the machine pulls the refrigerant out of the system and condenses it into the cylinder.

There are three ways to recover, and the charge size picks the method. Vapor recovery pulls refrigerant as a gas through the machine, condenses it, and drops it in the cylinder. It is the slowest method and it works on almost anything, so it is the default for small charges. Liquid recovery pulls liquid off the high side first, which is much faster, and you finish with a vapor pull to clear what is left. Push-pull is the method for large charges, commonly cited around 20 lb and up: you use the machine to pull vapor off the top of the cylinder and push it into the system, and that pressure drives the liquid charge straight from the system into the cylinder, bypassing the machine's compressor for the bulk of the liquid. Then you switch back to vapor recovery to finish.

Two hard rules on the cylinder. Never overfill it. The standard is to fill a recovery cylinder to no more than 80 percent of its capacity, leaving 20 percent vapor space, because liquid refrigerant expands hard when it warms and a cylinder filled liquid-full can rupture. That limit traces to DOT under 49 CFR 173.304a and is codified as industry practice in AHRI Guideline K. Weigh as you fill or use a cylinder with a float or shutoff, and never use a cylinder past its hydrostatic retest date.

Required recovery levels by appliance

How far you have to pull the system down during recovery depends on the appliance and the refrigerant, and the 608 rules set the targets. The table below is the common framework for high-pressure equipment, but the required levels and the dates that trigger them have shifted across rule revisions, so confirm the current 608 recovery-level table for the appliance and refrigerant in front of you before you call it done.

The one exception worth carrying in your head: on the very high-pressure refrigerants like R-410A, R-404A, and R-507, the target is 0 psig rather than a deep vacuum, because pulling a vacuum during recovery on those risks freezing moisture into the system and doing more harm than the recovery is worth. Recover to the required level, isolate, and watch that the pressure does not climb back up, which would tell you refrigerant is still boiling out of the oil.

EPA Section 608: venting, recovery, and leak repair

Section 608 of the Clean Air Act is the rule that governs all of this, and the core of it is short. You may not knowingly vent refrigerant, the work has to be done by a 608-certified technician with certified recovery equipment, and the larger appliances carry leak-repair and recordkeeping duties on top.

The venting prohibition is the part with no gray area. Intentionally releasing refrigerant is prohibited, and the prohibition was extended to cover the non-ozone-depleting refrigerants, the HFCs, beyond just the old ODS refrigerants. Small releases that are unavoidable during good-faith service, like the wisp left in a hose, are treated differently from dumping a charge, but the intent of the rule is that the charge gets recovered.

For the larger systems there are leak-repair thresholds. On appliances holding 50 lb or more of refrigerant that leak above a set annual rate, the owner has to repair the leak and verify the repair. The thresholds were tightened in recent rulemaking, commonly cited at 30 percent per year for industrial process refrigeration, 20 percent per year for commercial refrigeration, and 10 percent per year for comfort cooling. Recordkeeping follows: service records with the date, the type of service, and the amount of refrigerant added, kept on site and available to EPA. The exact thresholds, the covered refrigerants, and the dates have moved across rule revisions, so verify the current Section 608 requirements for the appliance you are servicing rather than trusting a number from memory.

Repairing the leak the right way

Once the charge is recovered and the system is open, the repair depends on what leaked. A leaking brazed joint gets cleaned, refluxed, and rebrazed, or cut out and replaced if it is a bad joint. A leaking flare gets the seat inspected, deburred, and remade to the right torque, or replaced if the flare is split. A leaking Schrader core gets a new core, which is the cheapest repair on the list and one you should always check first. A weeping service valve gets rebuilt or replaced.

The coil is the hard call. A pinhole in a coil hairpin can sometimes be brazed, but formicary corrosion rarely comes alone, so a coil that leaked once often leaks again somewhere else within a season. On an aging coil with a corrosion leak, replacing the coil usually beats chasing pinholes that keep reappearing, and the math on a third callback is what makes that decision for you.

Where the leak is at a spot with no access fitting, add one. A brazed-in access valve or a tee with a Schrader gives the next tech a place to read pressure and recover from without breaking a joint, and it earns its keep the first time the system is serviced again. The repair is not finished when the leak is fixed. It is finished when the system is proven tight and dry, which is the next two steps.

Why braze with nitrogen flowing?

You flow dry nitrogen through the tubing while you braze to keep the inside of the copper clean. Heat copper in the presence of the oxygen that is sitting inside the tube and it forms copper oxide, a dark scale that at brazing temperature turns into black cupric oxide flakes. Those flakes break loose and the refrigerant and oil carry them straight into the metering device, the small orifices, and the compressor, where they clog and score and cause the failure you were trying to prevent.

Flowing nitrogen pushes the oxygen out of the tube so there is nothing for the hot copper to react with, and the inside of the joint comes out bright instead of scaled. Set a gentle purge, commonly around 2 to 3 cubic feet per hour at a couple of psi, enough to displace the oxygen without blowing through the molten braze or chilling the joint. Start the flow before the torch touches the work and keep it running until the joint cools below roughly 500 degrees F.

This is the step that separates a clean install from a future service call, and it is the step the rookie skips because the joint looks fine from the outside. The damage is on the inside, where you cannot see it until the orifice plugs months later. Braze dry and braze under nitrogen, every joint, no exceptions.

What micron level for evacuation?

Evacuate the system to 500 microns or below before you charge it. Evacuation is pulling a deep vacuum with a vacuum pump to boil off moisture and pull out the noncondensable gases, mainly air, that got in while the system was open. You read it with a micron gauge, not the gauge on the manifold, because a manifold gauge cannot see the difference between a good vacuum and a great one down in the range that matters.

Why so deep. Moisture left in a system reacts with the oil and the refrigerant to form acids that eat the compressor windings, and at 500 microns water boils at about minus 12 degrees F, which is what lets the vacuum actually pull the water out as vapor instead of leaving it pooled. Noncondensables raise head pressure and rob capacity. Pulling to 500 microns is what gets both out, and a system that was open a long time or holds POE oil is worth pulling to 300 for margin.

Hitting 500 microns and holding 500 microns are not the same thing, which is the part techs get wrong. Run the standing vacuum decay test: pull to target, isolate the pump, and watch the micron gauge. A small slow rise that levels off, on the order of 50 to 100 microns over 10 minutes or so, means the system is tight and dry. A fast climb means a leak. A steady climb that stalls higher up means moisture is still boiling off and you are not done. On a system that fought you, a triple evacuation, pulling down, breaking the vacuum with dry nitrogen, and pulling again, sweeps out moisture that a single pull leaves behind. Set the target and the decay criteria, do not just watch the pump run.

Recharging after the repair

With the system tight and evacuated, weigh in the charge. The most reliable way to charge a system you opened is to weigh in the amount the data plate calls for, using a charging scale, because you know exactly how much went in and the number lands on the record. On a system with line-set length that differs from the rated length, adjust the charge for the extra liquid line per the manufacturer's table.

Then verify, do not assume. A weighed-in charge gets you close, but the proof is the superheat and subcooling read at the operating conditions against the equipment targets, the same method you would use charging any system. Charge a fixed-orifice system by superheat and a TXV or EEV system by subcooling, and confirm the two numbers agree with the chart before you sign off. The leak detection guide ends where the charging guide begins, and the two together close the call.

Last, leave the system better than you found it on the detection side. After recharge, run the electronic detector around the joint you repaired and the rest of the suspect area one more time. You fixed one leak. Confirm you did not miss a second one, because the second leak is what brings you back next month.

Working leaks on A2L refrigerants

The A2L refrigerants, R-32 and the R-454B blend among them, are mildly flammable, and that changes how you treat a leak. A2L is the ASHRAE Standard 34 safety class for low-toxicity, lower-flammability refrigerants, where the 2L means a burning velocity below 10 cm per second. They do not ignite easily and they tend to burn slow and self-extinguish, but a leak that builds concentration in a closed space near a high-energy ignition source is still a real hazard, which is the reason the codes around them tightened.

On the leak-detection side, A2L equipment is built with the flammability in mind. Many systems carry a built-in refrigerant leak sensor that triggers ventilation, and the handheld detectors rated for A2L are made to a sensitivity standard and built so they do not spark or run a surface hot enough to ignite the gas. Use detection and tools rated for A2L, keep the space ventilated when you open a system, and keep ignition sources away from a suspected leak.

The charge limits are the other piece. ASHRAE Standard 15 caps how much A2L charge is allowed for a given occupied space, tied to the refrigerant's flammability limit and the room size, which is why the charge-limit math matters on VRF and other high-charge A2L systems. Confirm the equipment listing, ASHRAE 15 and 34 as adopted, and the manufacturer's instructions for the specific refrigerant. The VRF commissioning guide covers the charge-limit calculation in more depth.

Leaks on data center and critical cooling

On a data center, a hospital, or any critical-cooling load, a refrigerant leak is not a comfort complaint waiting for a slow fix. It is a clock running against an outage. The cooling cannot go down while you recover, repair, and evacuate, so the work is built around the redundancy: you isolate and service the failed unit while the redundant units carry the load, which is the reason those systems are designed N+1 or better in the first place.

The leak you can least afford is the one nobody is watching for. Critical systems carry large charges across long pipe runs, and a slow leak on a system that is still holding temperature because the redundant capacity is masking it can run for months. The fix is monitoring: leak sensors in the space and at the units, charge tracking against a baseline, and trend data that flags a slow loss before it becomes a tripped unit on the hottest day of the year. Find the leak on the redundant unit while the others hold the room, not after the room is already losing temperature.

Coordinate the recovery and evacuation around the load, not around your schedule. On these jobs the commissioning record and the service log are part of the asset, because the next tech inherits a system where a missed step shows up as downtime measured in dollars per minute.

The cheapest leak is the one caught on a planned check instead of a no-cool call, so on the high-charge systems a routine leak inspection belongs in the maintenance. Sweep the joints and the coil with the detector, look for oil traces, check the Schrader cores and caps, and compare running superheat and subcooling to last visit's numbers, because a charge drifting low shows there before it shows as a complaint. Track the charge added: a unit that needed two pounds last spring and three this spring is telling you the leak is getting worse, and on the 50 lb and larger appliances that leak inspection and the post-repair verification are a recordkeeping duty, not just good practice.

What to document

Write up a leak repair or you leave nothing behind it, and on the larger systems you have also skipped a 608 recordkeeping duty. The record answers the next tech's first question, which is whether this is a new leak or the same one coming back, and it is the trail that proves the charge was recovered and not vented.

Capture the leak location and the component, the method that found it, the refrigerant type and the amount recovered in pounds, the repair you made, the evacuation level and the decay-test result, the recharge amount and method, and the certified tech and date. The recovered amount and the recharge amount together tell the story of where the charge went, which is exactly what the EPA recordkeeping is after on the appliances it covers.

Common mistakes

• Topping off the charge and leaving without finding the leak, which is venting on a delay.
• Releasing the charge to the air instead of recovering it, which is illegal under 608.
• Brazing without dry nitrogen flowing, scaling the inside of the tube and clogging the metering device later.
• Calling a nitrogen pressure drop a leak without correcting for the overnight temperature change.
• Stopping evacuation when the gauge hits 500 microns instead of running the standing decay test to confirm it holds.
• Overfilling the recovery cylinder past 80 percent, risking a rupture.
• Fixing one leak and not re-checking for a second one before recharging.
• Skipping the proportional duties on 50 lb and larger appliances: leak repair, verification, and recordkeeping.

Field checklist

Standards and references

EPA Section 608 of the Clean Air Act is the rule that controls leak work: the venting prohibition, the requirement to recover with certified equipment, the recovery levels by appliance, the leak-repair thresholds on the larger appliances, and the recordkeeping. The thresholds, covered refrigerants, and effective dates have moved across rule revisions, so verify the current 608 requirements rather than a remembered number.

Recovery and recovery-equipment performance trace to AHRI standards, with AHRI 740 covering recovery and recycling equipment and AHRI Guideline K codifying the 80 percent cylinder fill practice, while the cylinder transport limit itself sits with DOT under 49 CFR 173.304a. Refrigerant safety classification, including the A2L class, comes from ASHRAE Standard 34, and the charge limits, ventilation, and equipment-placement rules for flammable refrigerants come from ASHRAE Standard 15. The AIM Act drives the transition to the lower-GWP A2L refrigerants now in the field.

Above all of it sits the equipment manufacturer. The test pressure, the charge, the line-set adjustment, the dye compatibility, and the A2L handling for a specific unit come from the manufacturer's instructions and the equipment listing, and where they are stricter than the general rule, they control. Cite the standard that governs the point and confirm the editions and rule versions a jurisdiction has actually adopted.

Units, terms, and conversions

Leak and recovery work crosses pressure, vacuum, and weight units, and the same reading shows up in different forms across a manifold, a micron gauge, and a recovery scale.

System and test pressure read in psig, the gauge pressure above atmospheric, while a recovery target can be given as inches of mercury (in Hg) of vacuum. Deep vacuum for evacuation is read in microns, where 1 micron is one-thousandth of a millimeter of mercury and 500 microns is the common dry-and-tight target. Refrigerant amounts and recovery are tracked in pounds and ounces. Nitrogen purge flow for brazing reads in cubic feet per hour (CFH). Refrigerant safety class (A1, A2L, A3, and the B classes) comes from ASHRAE 34, where the letter is toxicity and the number is flammability.

Recovery

Removing refrigerant from a system into an external cylinder, required by EPA 608 before opening the system

Evacuation

Pulling a deep vacuum to boil off moisture and remove noncondensables, read in microns

Micron

Unit of deep vacuum, one-thousandth of a millimeter of mercury; 500 microns is the common evacuation target

Standing pressure test

Holding nitrogen pressure and watching the gauge over time, with temperature correction, to prove tightness

Decay test

Isolating the vacuum pump and watching the micron gauge for a rise that signals a leak or remaining moisture

Push-pull

A fast recovery method for large liquid charges that drives liquid into the cylinder using machine-generated pressure

A2L

ASHRAE 34 safety class for lower-toxicity, mildly flammable refrigerants like R-32 and R-454B

Noncondensables

Gases such as air that will not condense in the system, raising head pressure until evacuated out

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• Do I charge by superheat or subcooling?
• What is a VRF system?
• Why do data centers need liquid-cooling leak detection?
• What is a leak detection system in a data center?
• Where do most roof leaks come from?