Water is the quiet enemy of a transformer. You don’t see it, it shows up in single-digit parts per million, and that’s enough to matter. Go from 10 ppm to 30 ppm of water in the oil and the breakdown voltage can drop by half. The water also feeds the slow rot of the paper insulation, which is the part of the transformer you can never replace.
So you need a number. Not “the oil looks fine” — an actual figure in ppm that you can trend, compare to a limit, and act on. Karl Fischer titration is how you get that number. It’s the reference method for water in insulating oil, and it’s accurate down to the trace levels that matter here.
This guide covers what the test does, the two versions of it, how the water gets out of the oil and into the instrument, what the number means once you have it, and the sampling mistakes that quietly wreck the result.
Table of Contents
The reaction it’s built on
Karl Fischer figured out in the 1930s that water reacts cleanly and predictably with iodine in the presence of sulfur dioxide, a base, and an alcohol — usually methanol. The useful part is the ratio: one molecule of water consumes exactly one molecule of iodine. Nothing fuzzy about it. If you know how much iodine got used up, you know how much water was there.
That one-to-one relationship is the whole trick. Every version of the test, old or modern, is just a different way of answering the same question: how much iodine did the water eat?
The reaction needs the alcohol and the base to work properly — they tie up the reaction products so the iodine keeps reacting with water and only water. Old reagents used pyridine as the base, which is nasty stuff. Modern reagents swapped it for safer bases, but the chemistry underneath is the same.
Two versions: volumetric and coulometric
There are two ways to run the titration, and picking the right one matters.
Volumetric. You add iodine from a burette as a liquid titrant. You keep adding until the water is used up and a little excess iodine appears — that’s the endpoint. Volumetric is the right choice when there’s a lot of water to measure, roughly the milligram range and up, or percent-level moisture in solids.
Coulometric. You don’t add iodine at all. Instead the instrument generates it electrically, right inside the cell, by passing current through an iodide solution. Iodine appears at the anode in exact proportion to the charge you pass. Faraday’s law makes the conversion clean: 1 mg of water corresponds to 10.71 coulombs of charge. Count the coulombs, you’ve counted the water.
For transformer oil, you want coulometric. Dry oil sits in the single digits to low tens of ppm. That’s trace water, and only the coulometric method has the sensitivity to measure it accurately. Volumetric simply isn’t fine enough at these levels.
The endpoint, in both cases, is found electrically. A pair of small platinum electrodes watches for the first trace of excess iodine in the solution. The moment it appears, the water is gone and the titration stops. No colour-watching by eye.
Getting the water out of the oil
Here’s the practical wrinkle. Oil isn’t a clean sample. It’s viscous, it carries additives, and if you dump it straight into the cell it can foul the electrolyte. There are two ways to deal with that.
Direct injection. You inject the oil straight into the titration cell, usually with a cosolvent that helps the oil mix into the methanol-based electrolyte. It’s fast and simple. It works well for clean mineral oils. The downside is that oil builds up in the cell over repeated runs, and some additives can interfere.
The oven (vaporizer) method. You seal a measured oil sample in a vial and heat it. The water evaporates, and a stream of dry nitrogen carries that water vapour — and only the water vapour — into the titration cell. The oil itself never enters the cell. This is the cleaner approach for oils with additives, higher viscosity, or anything that might cause a side reaction. The international standard for water in insulating liquids points to the nitrogen-extraction method as the preferred one for the more viscous fluids.
The relevant standards to cite on a test report: IEC 60814 (the coulometric Karl Fischer method for insulating liquids and oil-impregnated paper and pressboard), and the ASTM equivalents D6304 and D1533. IEC 60814’s liquid method is meant for water concentrations above about 2 mg/kg, which covers everything you’ll see in service.
It helps to know where Karl Fischer sits among the oil standards. IEC 60296 governs unused oil, checked before the oil goes into the transformer. Once the oil is in the tank, IEC 60422 takes over for in-service monitoring and acceptance limits. Karl Fischer per IEC 60814 is the method that produces the water number both of them rely on, and breakdown voltage — the test water most directly affects — is run to IEC 60156.
Running it on transformer oil
The procedure itself is short. The discipline around it is what makes the number trustworthy.
- Sample from the bottom valve. Water and sludge settle low, so the bottom valve gives the honest picture. Flush a little oil through first to clear the dead volume in the valve.
- Keep air out. This is the one that bites people. Oil is hygroscopic — it pulls moisture straight out of humid air. Use a clean, dry syringe or a sealed sample bottle, fill it full with no air gap, and cap it immediately. On a humid day this is the difference between a real reading and a fake one.
- Let the instrument settle first. Run the cell to a stable baseline drift before you inject. A drifting cell adds water that isn’t in your sample.
- Weigh the sample. Coulometric KF gives you micrograms of water. To turn that into ppm you need the sample mass, so weigh the syringe before and after injection (or weigh the vial for the oven method).
- Read and convert. Water (µg) ÷ sample mass (g) = ppm, which is the same as mg/kg. That ppm figure is what goes on the report.
Record the oil temperature at the moment you sampled. You’ll see why in a second.
Reading the number
A ppm figure on its own isn’t an answer until you put it against a limit and a temperature. It helps to keep three quantities straight:
- W_abs — the absolute water content, in mg/kg (the same as ppm). This is what Karl Fischer gives you. It doesn’t depend on temperature.
- W_s — the oil’s water solubility: how much water that particular oil can hold at a given temperature. It rises with temperature, and old oil holds far more than new oil.
- W_rel — relative saturation, W_abs ÷ W_s, as a percent. If W_rel goes past 100%, the excess water is free — you’ll see it as haze or droplets.
Why ppm alone can mislead. The raw ppm number means little until you tie it to temperature, because W_s moves with temperature. As a rough orientation, dry or freshly processed oil sits in the single digits to low tens of ppm, and in-service oil is commonly kept below the 20–30 ppm range, with drying triggered above that. But treat those as orientation only — the actual acceptance limits live in IEC 60422, vary by voltage class (higher-voltage units get tighter limits), and are not a single blanket number.
Percent saturation is the better grid. Because it already accounts for temperature, relative saturation is the more reliable way to judge the insulation. The common interpretation bands (IEEE C57.106):
| Water saturation in oil | Condition of the cellulose insulation |
|---|---|
| < 5% | Dry |
| 5–20% | Moderately wet |
| 20–30% | Wet |
| > 30% | Extremely wet |
Mind the reference temperature. When moisture results are corrected to a reference for comparison, that reference is 20°C, and the correction is only valid at sampling temperatures of 20°C or above. Don’t confuse this with the 75°C reference used for resistance and load-loss work — they’re different references for different tests.
The dielectric-strength connection. Water and breakdown voltage move in opposite directions. A small rise in moisture drags the breakdown voltage down hard — the same oil that breaks down above 70 kV when dry can fall below 20 kV once it’s wet. That’s why moisture and dielectric breakdown voltage are read together: one explains the other. A low breakdown result with high moisture tells you the problem is water, not contamination or gas.
Most of the water isn’t in the oil. This is the part people miss. In an oil-paper transformer, the overwhelming majority of the water lives in the paper and pressboard, not the oil — new units are dried to roughly 0.5–1% moisture in the cellulose before filling. Only a tiny fraction is dissolved in the oil at any moment. What Karl Fischer measures is that small fraction — a window into the much larger amount sitting in the solid insulation.
And it moves with temperature. Water shifts between the paper and the oil as the transformer heats and cools. A hot transformer drives water out of the paper and into the oil, so the same unit reads a higher ppm when warm and a lower ppm when cold. Two samples from the same transformer at different temperatures can look like a change that isn’t real. Measure the oil temperature in the flowing stream as you sample, note it on the report, and lean on relative saturation rather than raw ppm.
Where it goes wrong
A Karl Fischer result fails quietly. It hands you a confident number that happens to be wrong. The usual causes:
Moisture picked up during sampling. The single biggest one. Air contact, a damp syringe, a slow transfer on a humid day — any of these add water that was never in the transformer. Seal it, fill it full, work fast.
A leaking cell. A bad septum or a worn seal lets room air seep into the titration cell. The drift climbs and your low-ppm readings inflate. Check seals and septum if the baseline won’t settle.
Exhausted reagent. The coulometric electrolyte has a limited capacity, set by its sulfur dioxide content. Once it’s spent, results drift off. Watch the reagent’s age and condition.
Side reactions in direct injection. Some additives, or certain contaminants, can react in the cell and report as water when they aren’t. If you suspect this, switch to the oven method — vapour transfer leaves the troublemakers behind in the vial.
An unmixed sample. If the oil wasn’t homogenised, free water can separate out and you’ll either miss it or over-read it. Mix before you draw.
Reading drift as result. If you inject before the cell has stabilised, the baseline drift gets counted as sample water. Always establish the blank first.
Where it fits in your oil testing
Karl Fischer titration doesn’t stand alone. It’s one leg of a routine oil panel that usually also includes dielectric breakdown voltage, acidity, interfacial tension, and dissolved gas analysis. Moisture is the test that ties several of them together: it explains a sagging breakdown voltage, it accelerates the acidity and aging you see elsewhere, and tracked over time it tells you whether the insulation is drying out under maintenance or slowly getting wetter.
Run as a quick injection with a careless sample, it produces a number that occupies a line on a report and means nothing. Run with a sealed sample, a settled cell, a recorded temperature, and the right method for the oil, it’s the most reliable way there is to know how much water is hiding in a transformer.
FAQ
What does Karl Fischer titration measure in transformer oil?
It measures dissolved water content, reported in parts per million (ppm), which is the same as milligrams of water per kilogram of oil. It’s the reference method for moisture in insulating oil.
Volumetric or coulometric for transformer oil?
Coulometric. Transformer oil carries trace water — single digits to low tens of ppm — and only the coulometric method, where iodine is generated electrically, is sensitive enough at that level. Volumetric is for higher water contents.
What’s an acceptable water content in transformer oil?
As a rough orientation, single digits to low tens of ppm for new or dried oil, and below the 20–30 ppm range in service, with drying triggered above that. But ppm alone is temperature-dependent — the more reliable grid is relative saturation: under 5% is dry, 5–20% moderately wet, 20–30% wet, above 30% extremely wet. Use the voltage-class limits in IEC 60422 for the actual acceptance values.
Why does temperature matter when sampling?
Because water migrates between the paper insulation and the oil as the transformer heats and cools. A hot unit reads higher ppm, a cold one lower, for the same actual condition. Record the oil temperature at sampling, and consider relative saturation rather than raw ppm for condition assessment.
How does water affect dielectric strength?
Strongly and in the wrong direction. A modest rise in moisture can roughly halve the oil’s breakdown voltage. That’s why moisture and breakdown voltage are read together — moisture usually explains a low breakdown result.
What standards cover the test?
IEC 60814 for coulometric Karl Fischer on insulating liquids and oil-impregnated paper, with ASTM D6304 and D1533 as the common equivalents. In context: IEC 60296 covers unused oil before filling, IEC 60422 covers in-service oil and its limits, and IEC 60156 covers the breakdown-voltage test that water most affects.