The same number, five different meanings
Training material from Manila to Munich quotes the same figure for dissolved gas analysis (DGA) accuracy: ±15 %. That number is genuinely international — but the standards that surround it define accuracy in ways that are not interchangeable, and the error bands they actually enforce differ by a factor of five or more. For anyone writing a specification, or reading a laboratory report that says “compliant,” the distinctions decide whether two data sets can be compared at all. This article maps the accuracy requirements themselves — not the interpretation criteria — across the major standards.
The international baseline: IEC 60567
IEC 60567 is the international standard for sampling and analysis. Its ±15 % is an acceptance target drawn from round-robin testing against external gas-in-oil standards: the average result across participating laboratories. It includes extraction and degassing plus the chromatographic analysis, and it excludes sampling. CIGRE and IEEE apply the same target to laboratories and to online monitors alike, so the figure is a shared yardstick rather than a chromatograph’s precision specification. (We treat that decomposition in detail in the uncertainty budget.)
What each standard actually requires
The table below collects the published accuracy requirements. They measure different quantities — repeatability, reproducibility, inter-laboratory agreement, acceptance targets — which is precisely why they cannot be ranked as a simple league table.
| Standard | Scope | Published accuracy requirement |
|---|---|---|
| IEC 60567:2023 | International — sampling and analysis | Acceptance target ±15 %; includes extraction, excludes sampling; applied to laboratories and online monitors alike |
| ASTM D3612-02(2026) | United States — laboratory method | 95 % confidence limits of roughly ±25–48 % repeatability and ±38–82 % reproducibility (concentration-dependent); the headspace variant states no precision figure |
| IEEE C57.104-2019 | United States — interpretation and data quality | States candidly that while several laboratories meet the ±15 % requirement, others do not — with errors reported as high as ±60 % or more for some gases |
| РД 34.46.303-98 | Russia — mandatory error bands | Band by concentration: <10 ppm >50 %; 10–50 ppm ≤50 %; 50–500 ppm ≤20 %; >500 ppm ≤10 % |
| GB/T 7252 | China — laboratory analysis | Inter-laboratory agreement ≤30 %; within a single laboratory, ≤10 % above 10 µL/L |
| Q/GDW 10536-2021 | China — online monitoring devices | Class A: hydrogen ±2 µL/L or ±30 %; acetylene ±0.5 µL/L or ±30 % (whichever is greater) |
| Proficiency testing (iis) | Independent round-robin | Group reproducibility of 36–105 % of the reported mean — with the IEC targets described as very hard to meet |
Why the numbers diverge so widely
Three effects explain almost all of the spread.
- Different quantities. Repeatability (one laboratory, one instrument, same sample) is inherently tighter than reproducibility (different laboratories, different operators, different days). Quoting one as though it were the other flatters or damns a method unfairly.
- Concentration dependence. An error expressed as a percentage is meaningful only well above the detection limit. At 5 µL/L of acetylene, a ±10 % claim describes a capability no field instrument has.
- Scope. Some standards govern the analytical step; some govern the whole chain including sampling; some govern what an online device must demonstrate. A figure cannot be compared across scopes without unpacking what it includes.
The concentration trap
The Russian bands are the most instructive entry in the table, because they are the only one that refuses a single number. Requiring >50 % error tolerance below 10 ppm looks lax until you notice it is an admission: at trace concentrations, percentage accuracy degrades, and pretending otherwise invites false confidence. Every other standard in the table quietly depends on that same physics. This is why a specification that names one percentage without naming the concentration is not a specification — and why the correct question is never “what is your accuracy?” but “what is your accuracy at the concentration I care about?”
Sources
- IEC 60567:2023 (Ed. 5.0) — the ±15 % acceptance target: round-robin basis, includes extraction and analysis, excludes sampling.
- IEC 60599:2022 — interpretation of dissolved and free gases in mineral-oil-filled equipment.
- ASTM D3612-02(2026) — precision statements (repeatability and reproducibility, both concentration-dependent).
- IEEE C57.104-2019 — data-quality discussion, including measurement errors reported as high as ±60 % or more for some gases.
- РД 34.46.303-98 (Russia) — concentration-banded error limits, Table 1.
- GB/T 7252 (China) — inter-laboratory and within-laboratory agreement limits.
- Q/GDW 10536-2021 (State Grid, China) — accuracy classes for online monitoring devices.
- iis (Netherlands) annual DGA proficiency-testing scheme — group reproducibility of 36–105 % of the reported mean.
- Bräsel, E. & Sasum, U., “Genauigkeit der Gas-in-Öl-Analyse,” ew 111(14):50–55, 2012 — origin of the ±15 % figure.
What to put in a specification
- Name the standard and the quantity: repeatability, reproducibility, or acceptance target.
- State the concentration at which the figure applies, or require a banded table.
- Require evidence of participation in round-robin or proficiency testing, with results — not a self-declared compliance statement.
- For online devices, require the comparison protocol against laboratory reference analysis to be defined in advance: samples, timing, and the acceptance threshold.
Our IEC 60599 vs IEEE C57.104 comparison covers the interpretation side of the same standards landscape; this article covers only what they demand of the measurement.
PAS DGA and standards-comparable data
An online monitor earns its place by producing numbers that can sit beside a laboratory result without an argument. The PAS DGA DGA-900 reports nine gases plus moisture with a documented calibration chain, and the DGA-500 covers hydrogen and moisture for assets that need only those. Both are specified with the concentration-dependent thinking above in mind — absolute and percentage limits stated together, and hydrogen measured on a dedicated channel. Browse the product range, or contact PAS DGA to discuss the acceptance protocol for your next project.