// Services / Refrigeration Oil Analysis

Condition analysis for refrigeration compressor oil

Condition and wear analysis for chiller and refrigeration compressor oil — mineral, alkylbenzene, POE and PVE. Chiller and refrigeration compressors, and nothing else. That is narrower than most oil laboratories advertise, and the narrowness is the point.

Scope of certification — read this first

Our AHRI Certified^® Refrigerant Testing Laboratory certification covers refrigerant testing. It does not cover oil analysis. There is no oil scope in that programme for anyone to borrow, us included. This service is not performed under the certification, and no result on an oil report of ours is certified by it.

What the oil service stands on instead is the rest of this page: which parameters are measured, how each one is read against the others, and what has to be known about the machine before any of them means anything.

01Scope

Four oil families, one kind of machine

A general industrial oil laboratory is built around engines, hydraulics and gearboxes. That is a reasonable business and it is the wrong frame for a centrifugal chiller.

Refrigeration oil is the only oil on the bench that lives with a refrigerant dissolved in it, that is expected to be additive-lean rather than additive-rich, and whose acid number can climb without a single oxidation event. Read a POE result out of a centrifugal against a hydraulic-oil frame and the errors are predictable: a low viscosity gets called shear or a cross-charge when it is dissolved refrigerant, a silicon figure gets called dirt when it is sealant from last month’s gasket, and a water figure gets held against the wrong saturation curve entirely.

Knowing what normal looks like is the other half of it, and normal is per machine. A flooded centrifugal with a purge unit, a screw compressor running a high oil-to-refrigerant ratio, and an open-drive reciprocating unit do not share a normal — not for iron, not for copper, not for the particle count. Same oil, same refrigerant, different expected result. A laboratory that sees one refrigeration sample a week among a thousand hydraulic ones has no way to hold that distinction, because it has never needed to.

So this service takes mineral, alkylbenzene, polyol ester and polyvinyl ether oil out of chiller and refrigeration compressors. It is not a general oil-testing service and we do not offer one.

Oil families in scope, and what changes how each one reads
MineralMO	Legacy low-pressure centrifugals and older reciprocating plant on CFC and HCFC charges. Holds very little dissolved water. A water figure that is unremarkable in an ester is a large fraction of saturation here, and free water appears sooner.
AlkylbenzeneAB	Retrofits and low-temperature systems, frequently blended with mineral oil rather than replacing it. Oxidatively stable. A rising acid number is more likely contamination or a refrigerant-side reaction than oil oxidation, which changes what you go and look at.
Polyol esterPOE	The default for HFC and HFO chillers and rack compressors. Solest is one trade name among several. Hygroscopic and hydrolysable. Water and acid are coupled — hydrolysis produces the acid that catalyses more hydrolysis — so water is the governing number, not a secondary one.
Polyvinyl etherPVE	HFC and HFO systems on OEM equipment that specifies it, including much variable-flow plant. At least as hygroscopic as an ester, but it does not hydrolyse the same way. A high water figure alongside a flat acid number is not the alarm here that it would be in a POE.

02The panel

What is measured, and what each group answers

Four groups. Wear metals and contaminant elements say what is in the oil; particle counts say how much of it there is; the condition set says whether the oil is still the fluid you charged.

AIs a surface losing material, and which surface?

Wear metals

ppm by weight, per element

Wear metals — parameters and what each one comes from
Iron	Journals, shafts, gears, cast-iron crankcases and cylinder walls, screw rotors. The general wear indicator, and the one with the most sources.
Copper	Bearing cages, bushings and thrust washers — and, separately, tube and cooler surfaces under chemical attack. Two unrelated causes, one number.
Aluminium	Pistons, housings and some bearing alloys. Reads as abrasive wear when it moves with silicon.
Chromium, nickel	Alloy steels, rings and shafts. A chromium rise against flat iron is unusual enough to justify looking at the machine rather than the next sample.
Tin, lead, silver	Babbitt and bearing overlay. Silver also comes from some cages and some brazed joints, so it is read with the bearing metals, not alone.
Molybdenum	Ring and surface coatings — but also an additive in some oils. Which is exactly why the oil charged has to be on the sample record.

BIs something in the oil that did not come off a wear surface?

Contaminant and additive elements

ppm by weight, per element

Contaminant and additive elements — parameters and what each one comes from
Silicon	Airborne dirt, silicone antifoam and RTV squeeze-out all report here. Sequence separates them; a single figure does not.
Sodium, potassium	On a water-cooled machine, treated water through a leaking tube or oil cooler. These two moving together with water is the case worth chasing the same day.
Calcium, zinc, phosphorus, barium, magnesium	Additive chemistry. Refrigeration esters are formulated lean, so an element that was not in the oil you charged means a cross-charge or a flush that did not leave.
Boron	An additive element in some oils and, in some plants, a marker for coolant-side ingress. Read against the oil charged, never in isolation.

CHow much solid material is in the oil, and is it growing?

Particles and cleanliness

Counts per millilitre, reported as an ISO 4406 code

Particles and cleanliness — parameters and what each one comes from
Particle counts	The whole population, whatever it is made of — wear debris, ingested dirt, desiccant fines, filter media, gasket material.
ISO 4406 cleanliness code	Three range numbers for counts at 4, 6 and 14 µm(c) and above. One code number is a doubling of the count, so three steps is roughly eightfold.
Ferrous / non-ferrous split	Separates magnetic wear debris from everything else, which is the difference between a bearing shedding and a filter bypassing.

DIs the oil still the fluid you charged?

Condition

Water in ppm · acid number in mg KOH/g · viscosity at 40 °C

Condition — parameters and what each one comes from
Water	The governing number in an ester, and the one most easily destroyed by the sample's journey rather than the machine's condition.
Acid number (TAN)	Acid formation. In refrigeration it comes from hydrolysis, from thermal breakdown at hot discharge surfaces, and from a motor burn — rarely from ordinary oxidation.
Viscosity	Moves in both directions for opposite reasons: down for dissolved refrigerant or a lighter makeup oil, up for degradation or the wrong oil.
Appearance and colour	Free water, carbon, varnish and gross particulate are visible before any instrument reports them, and they change what the rest of the panel means.

Reading an ISO 4406 code

Three numbers, in order, for particles at and above 4 µm(c), 6 µm(c) and 14 µm(c) per millilitre. Each number is a range code, not a count, and each step up the scale is a doubling of the particles counted. Three code numbers is therefore roughly an eightfold change, which is why the size of the step between two samples carries more than the absolute code does.

Two things the code will not tell you

Composition. A count treats desiccant fines, gasket material and bearing debris identically. The elemental groups are what tell them apart, which is why both are on the same panel.
Whether it is even solid. Undissolved refrigerant coming out of solution in the bottle counts as particles. A sample that was not degassed before sealing can report a cleanliness code that describes the bottle, not the machine.

03Interpretation

Nearly every parameter is read in company

Single numbers are ambiguous almost everywhere on this panel. What resolves them is which other parameters moved, in which direction, and in what order.

Iron and copper, read as a pair

Iron is the broad wear signal. Copper is the one that needs a second question, because bearing bronze and chemically attacked tube surfaces produce the same element. The pairing separates them: copper rising with tin and lead is an overlay being lost, and copper rising on its own alongside acid number is chemistry — the oil is attacking metal, and no bearing inspection will explain it. After a motor burn, copper also plates onto steel surfaces, which suppresses the very number you are watching.

Silicon has three sources and one number

Dirt ingress through a leaking seal, silicone antifoam, and RTV squeeze-out from a joint made up last week all report as silicon. Sequence tells them apart. A step change immediately after a teardown is almost always sealant, and it decays over the next two samples. A slow rise that tracks aluminium and iron is abrasive third-body wear — a machine problem rather than a housekeeping one, and it does not decay on its own.

Water is the enemy of POE specifically

In an ester, water is not merely a contaminant — it is a reagent. Hydrolysis splits the ester into acid and alcohol, and the acid catalyses further hydrolysis, so the curve runs away instead of levelling off. Downstream come acid attack on metal and winding insulation, and free water at the expansion device. The figure is only meaningful with the oil family attached: the same ppm reads as routine in one fluid and as most of the way up the saturation curve in another.

Acid number matters most in company

A rising acid number on its own says the oil is changing. Rising with copper says it is taking metal with it. Rising with water in an ester says the reaction is feeding itself and the interval to the next sample should shorten. Read alone, the same number supports all three readings and settles none of them — which is why acid number is the parameter most often over-interpreted on a single-sample report.

A viscosity shift is a diagnosis, not a result

Down means dilution: refrigerant still in solution, or a makeup oil of a lighter grade. Up means degradation or a cross-charge. Both are worth knowing and they lead opposite ways. The most common cause of a low viscosity result, though, is not the machine at all — it is a sample sealed before the dissolved refrigerant came out of it, which is why degassing is a procedure step rather than a suggestion.

Two measurements, each blind where the other sees

Elemental analysis reads dissolved and fine material well and grows steadily blinder as particles get larger. A particle count reads the whole population and says nothing about what any of it is made of. Neither settles a case alone, and the disagreement between them is the informative part: flat iron with a rising count points at ingress or a filter bypassing, while rising iron with a flat count points at fine corrosive wear rather than anything spalling.

04Trending

One sample is a position. Two are a direction.

A first sample can only tell you whether something is grossly wrong today. Everything else this panel is good at is a rate of change.

Wear metal concentration is mass over volume. Two machines wearing at exactly the same rate report different numbers if their oil charges differ, and the same machine reports a lower number after makeup oil is added even though nothing about the wear has improved. Concentration alone therefore cannot be compared between machines, and can only be compared within one machine when the interval and the makeup volume are known.

What survives all of that is the slope. Iron per thousand hours is a real quantity; iron in ppm is a reading. The first sample on a machine exists to establish where that slope starts — ideally taken at a known point, such as immediately after an oil change or at commissioning, so the zero is not a guess.

Consistency matters more than frequency. Same sampling point, same operating condition, same procedure. Vary those and the trend you have built is a trend in how the sample was taken.

Six facts that change how the same number reads

01Oil family and grade actually charged — not what the nameplate says, what went in.
02Refrigerant in the machine, and whether it has ever been changed.
03Compressor type and configuration: centrifugal, screw, reciprocating, scroll; open-drive, semi-hermetic or hermetic.
04Running hours since the last oil change, and hours on the machine.
05Makeup oil added since the last sample, and roughly how much.
06Anything unusual since the last sample: a teardown, a burnout, a purge unit running hard, a change of duty.

None of this is administrative box-ticking. Molybdenum is a wear metal in one oil and an additive in another; a silicon step is sealant or dirt depending on whether a joint was opened; a viscosity drop is a cross-charge or a dissolved refrigerant depending on what was added. Without the six facts, half this panel is unreadable and the other half is guesswork.

05Limits

This page publishes no alarm thresholds

A deliberate absence, not an unfinished section. There is no honest single table to publish for refrigeration compressor oil, and a table that looks authoritative and is not would be worse than none.

The published literature is thin here

Engine and hydraulic oil have decades of published condemning limits behind them. Refrigeration compressor oil does not. What exists is scattered through OEM service literature, is specific to a machine family, and frequently does not cover the parameter you are actually looking at.

A limit is a property of the machine

The same iron figure means different things in a screw compressor with a large oil charge and a small reciprocating unit with a fraction of it. Sump volume, duty cycle, filtration and age all sit inside the number. A limit copied from one machine to another silently changes what it is measuring.

And it is a property of the oil

Water at a given ppm is routine in one fluid and well up the saturation curve in another. A rising acid number is a self-catalysing reaction in an ester and something else entirely in an alkylbenzene. One threshold across four oil families would be wrong for at least three of them.

What a result is evaluated against instead

01
The manufacturer’s limit where the machine has one. It outranks anything we would otherwise apply, and it is the first thing to ask your OEM for.
02
The new-oil reference for properties that belong to the fluid rather than the machine — viscosity grade, and the additive elements the oil was formulated with.
03
The machine’s own history — the only reference that already accounts for its charge volume, its duty and its age, and the reason section 04 exists.

Ask what your result was compared against and you will get a straight answer naming which of those three applied. What you will not get is a number invented to fill a column.

06Sampling

The constraint that actually bites

Most sampling advice is about representativeness. On refrigeration oil there is a second constraint that ruins results more often, and it is a clock.

A POE sample that waits measures the wait

Esters pull moisture out of the air they are exposed to, and they keep doing it through a bottle that has been opened once. A sample left in a service truck arrives with a water result that describes its journey as much as the machine. Nothing at the bench can subtract that back out, and the failure is invisible on the report unless someone knows to ask about the dates. In an ester it also drags the acid number with it, so a slow sample can manufacture a second false signal on the way.

POE and Solest oils absorb moisture quickly, even in a sealed bottle. Send samples promptly.

What a usable sample needs

The right bottle: A borosilicate graduated glass sample bottle. Glass because plastic contributes to the very contamination the particle count is measuring; graduated because volume is part of the record.
Degassed before sealing: Two results depend on it. Refrigerant still in solution reads as low viscosity, and gas coming out of solution in the bottle reads as particles. Both errors point the diagnosis somewhere the machine is not.
Sent immediately: The clock on an ester starts the moment the bottle is opened, and nothing at the bench winds it back. Register the sample before it travels so its identity is fixed by you rather than reconstructed by us when it arrives.
Labelled with the machine: The six facts in section 04, on the sample rather than in somebody’s memory. A sample with no asset identity cannot be trended against anything.

The full oil sampling procedure →

// Ready when you are

Send us the machine, not just the oil

Register the sample first and the asset details travel with it, so the first report already has something to read the numbers against.

Disclosure: we share ownership with Chiller Services Refrigerants, a refrigerant reclaimer in the adjacent unit of this building, and the purity we report on a refrigerant batch is an input to what that company pays for it. You are entitled to weigh that rather than discover it. How the laboratory is kept separate.