Transformer oil sampling best practices exist for one reason: a DGA result is only ever as good as the container that reached the laboratory. The same oil, drawn twice from one unit on the same afternoon, can read as a healthy transformer and as an arcing one. The laboratory is rarely at fault.
Why the sample, not the analyser, sets the error budget
If the oil was already altered at the valve, no instrument quality recovers the original gas content. IEC 60567:2023, the sampling and analysis standard, sets an overall acceptance target of roughly ±15 % for the two together, and a syringe drawn through the wrong tubing can move a single gas by more than that.
Air ingress adds oxygen and nitrogen and almost no acetylene or ethylene, so it cannot manufacture a fault but can bury one. Degassing loss does the opposite: hydrogen, the least soluble diagnostic gas, leaves the oil for the headspace and the unit looks quieter than it is. One is a false alarm; the other a missed warning.
Six sampling faults that manufacture gas
The signature column is the part worth keeping.
| Fault in the chain | Signature in the result | How to rule it out |
|---|---|---|
| Air drawn in at the valve | Oxygen and nitrogen rise together, near the 3.7:1 ratio of dry air | Compare nitrogen against oxygen; repeat from a second valve |
| Headspace in the container | Hydrogen falls first, then methane; totals drift down | A drawn-metal bottle filled to the brim leaves none |
| Wrong valve or compartment | Acetylene appears although nothing is arcing | Confirm the sample point; judge tap-changer oil under IEEE C57.139-2015 |
| Reused syringe, PVC or rubber tubing | Methane and ethane with no matching hydrogen trend | Single-use containers; PTFE or metal tubing |
| Long, warm transport | Hydrogen, methane and carbon dioxide creep upward | Keep it cool and dark; note the hours to analysis |
| Direct sunlight on the container | Hydrogen rises alone, reading like partial discharge | Re-sample, and transport in the dark |
Moisture rides the same chain: a wet fitting or a headspace reports a water content that describes the container, not the paper (IEC 60814, Karl Fischer).
What are common causes of false positives in DGA results?
Five recur in field work. Air drawn in at the valve raises oxygen and nitrogen and never acetylene. Ultraviolet light on a clear container generates hydrogen. A warm delay lets microbial activity add hydrogen, methane and carbon dioxide. Tubing and reused syringes contribute methane and ethane. The fifth sits inside the tank: stray gassing, where mineral oil produces hydrogen and methane at low temperature with no fault present, the reason ASTM D7150 exists. Handling practice: oil sampling guide, sampling procedure.
How do you rule an air leak or a sampling error out?
Work in this order, and never on one container alone.
- Read nitrogen against oxygen. A ratio near 3.7:1 with flat hydrocarbons points at ingress, not a fault.
- Ask what moved. A real fault raises gases in a pattern; a handling fault lifts one or two.
- Audit the chain. Container, flushing volume, tubing, light exposure, hours in transit; one link usually explains it.
- Re-sample from another valve. A properly taken second sample is the cheapest discriminator in DGA.
- Watch three readings, not one. Against the ±15 % band of IEC 60567:2023, one pair cannot settle a borderline case.
- Re-check the interpretation. Confirm each gas sits above its detection limit before trusting a ratio scheme.
Not every surprise is a sampling error
Writing off a genuine fault as contamination is the more expensive mistake, and no set of transformer oil sampling best practices replaces judgement. Stray gassing is benign until it is not; a tap changer can pass decomposition products into the main tank; an incipient thermal fault can start as noise. Before dismissing a result, ask whether the rise continued into the next sample and whether the unit has a history of similar readings. Two guides carry that decision further: what monitoring catches before failure and running the programme across a fleet.
What does continuous monitoring change about sampling?
An on-line analyser reads the oil in service, at the valve, with no container, transport or handling step. That removes the chain described above for the gases it covers, and it gives readings often enough to show a rate of rise rather than two points a year. It does not retire the laboratory: the full gas set still comes from a drawn sample, which anchors the trend. The trade between the two routes is set out in online versus offline DGA. PAS DGA covers the on-line half, from the DGA-500 dissolved hydrogen sensor to the DGA-900 nine-gas analyser; send the team your gas list and unit count.
Sources
- IEC 60567:2023 — sampling and analysis of gases in oil-filled equipment; acceptance target of roughly ±15 %.
- IEC 60475 and ASTM D3613 — sampling insulating liquids. ASTM D3612 — gas-chromatographic analysis.
- IEC 60599 — interpretation of dissolved and free gases. IEEE C57.139-2015 — tap-changer oil evaluation.
- ASTM D7150 — stray gassing potential. IEC 60814 — water by Karl Fischer titration.
- Dry air (standard value): 78.08 % nitrogen, 20.95 % oxygen by volume, giving the 3.7:1 ratio above.