There are two threads in sampling that are often looked at separately: where the sample is taken, and when it is taken. The position decides which part of the oil in the tank is drawn, the moment decides what state that oil is in. Both change the result, yet the site record usually carries the date, the sampling point and the test values, and the moment of sampling is often left blank.
Running and standing are not the same measurement
While the machine runs, the oil in the tank is in motion: new oil comes in and treated oil goes out all the time. The sample taken then represents the oil that has just passed through this part of the process, carrying the temperature it was heated to and the condition it had when it left the separation section, with the gas content at the level left after degassing.
After the machine stops, the oil stands in the tank. The temperature falls gradually, the gas that was dissolved under heat re-establishes its balance with the oil phase, free bubbles near the surface rise and dissipate, and components of different density may separate into layers. A sample taken now represents the oil in its standing state.

Both states can pass and both can fail, but they answer different questions. A sample taken while running answers what this part of the process is achieving at that moment; a sample taken after standing answers what this batch of oil is like after it has been stored. When the two are recorded together, the two sets of data are no longer comparable.
Temperature and bubbles rewrite the breakdown voltage
Breakdown voltage is the indicator most clearly affected by oil temperature and bubbles. Its test basis is GB/T 507 and the acceptance line is ≥65 kV. When the oil is hot, the breakdown voltage reading of the same oil is not the same as in the cold state; bubbles left in the sample form a gas gap between the electrodes, the dielectric strength of that gap is far below that of the oil itself, and the reading is pulled down. Take a sample without allowing enough standing time and the result may be a figure lower than the true one.
The same applies to water content, whose basis is GB/T 7600 and whose limit is ≤3 ppm. Water exists in two forms, dissolved water and emulsified water, and they behave differently in the oil; the distribution between them changes as the temperature changes. A different oil temperature and a different standing method at the moment of sampling can put the reading in a different place.
| Indicator | Limit | Factors most affected by the moment of sampling |
|---|---|---|
| Water content in oil | ≤3 ppm (GB/T 7600) | Oil temperature, the split between dissolved and emulsified water |
| Gas content in oil | ≤0.01% (GB/T 423) | Whether the machine is running, how tight the system is |
| Breakdown voltage | ≥65 kV (GB/T 507) | Oil temperature, residual bubbles, standing time |
| Cleanliness | NAS class 5 (NAS 1638) | Foreign particles introduced during sampling |
Fix the sampling rule
- State the moment: taken while the machine runs, or taken after it has stood. Only one of the two may be used, and it has to be used consistently.
- State the standing time: where a sample is taken after shutdown, define how long the oil shall stand as a clear requirement.
- State the oil temperature at sampling, or record the temperature at the moment of the test. Do not leave the field blank.
- State the specification of the sample container and the way it is sealed, so that different people do not use different bottles.
- Treat the same batch of oil with the same machine by the same method wherever possible, so that the data remains continuous.

Once the rule is fixed, the test data begins to carry diagnostic value. The margin above the acceptance line and the trend have been discussed before, and a trend only holds if the data is comparable. The baseline for that comparability is the moment of sampling. Putting it on the first page of the log is more reliable than trying to remember it later.
If the moment of sampling is not written down, one of the lines that connects two sets of data is missing.