A custody chromatograph is only trusted because it is proven against reference gas of a known composition. But there are two distinct jobs a reference gas can do, and doing them with the same cylinder quietly undermines the proof. The calibration cylinder is the blend the analyzer is adjusted against, the standard whose known values the GC is trimmed to match. The validation cylinder is a separate, independently certified blend used only to check the analyzer, run through it without changing any settings so the result reveals whether the GC is still reading true. Using one cylinder for both roles means the analyzer is checked against the very thing it was tuned to, which can only ever agree, so the check proves nothing. This guide explains why the two cylinders are kept separate, what happens when they are not, and how a monitoring system should treat validation results as a bias and drift metric over time.
Validation vs calibration cylinder in one line: A calibration cylinder is the certified reference blend that a chromatograph is adjusted against, so the GC's response is trimmed to match its known composition. A validation cylinder is a separate, independently certified blend run through the analyzer only to check its accuracy, without changing any calibration settings, so the result is an honest measure of whether the GC still reads true. Keeping them separate matters because checking an analyzer against the same cylinder it was calibrated to guarantees agreement and hides real errors, whereas a validation result against an independent blend can be trended over time as a genuine bias and drift metric.
Calibration and validation are separate steps that happen to use similar-looking bottles of gas. Calibration is the act of adjusting the analyzer: the GC runs the calibration blend, sees how its raw response compares with the certified composition of that blend, and updates its response factors so that it will report the certified values. After calibration the analyzer is, by construction, aligned to that particular cylinder. Validation is the act of checking the analyzer without touching it: a different, independently certified blend is run through the GC and the reported composition is compared with the cylinder's certificate, but no response factors are changed. The validation run answers the question calibration cannot answer about itself, namely whether the analyzer is still accurate on gas it was not tuned to.
The essential feature of a validation cylinder is that it is independent of the calibration cylinder, ideally from a different certified blend, and its known composition is not used to adjust the instrument. That independence is what gives the validation result meaning. If the analyzer reports the validation cylinder's composition within an acceptable tolerance without any adjustment, that is real evidence the calibration is holding and the GC is measuring correctly across the range. If the analyzer misses the validation cylinder, that is a genuine finding of drift or bias that a calibration against the calibration cylinder alone would never have exposed.
It helps to think of calibration as setting the answer key and validation as sitting a fresh exam. If you set the answer key from one test and then mark yourself against that same key, you will always score perfectly, which tells you nothing about whether you actually understand the material. Sitting a different test that you did not use to write the key is the only way to find out whether your knowledge generalises. A validation cylinder is that fresh test, and it is only informative because it is a different blend from the one used to set the analyzer's answer key.
When an operation uses a single cylinder for both calibration and validation, the validation step becomes circular and worthless. The analyzer is trimmed so that it reports that cylinder's certified composition, and then the same cylinder is run again and, unsurprisingly, the analyzer reports its certified composition. The check passes every time, not because the analyzer is accurate on real gas but because it is being asked to reproduce the exact number it was just tuned to reproduce. Any systematic bias in the analyzer's handling of a component that differs between the cylinder and the process gas, any nonlinearity, or any error in the cylinder's own certificate is invisible, because the same cylinder cannot reveal a fault that it helped to create.
The most dangerous version of this is a bad calibration cylinder. If the calibration cylinder's certified composition is itself wrong, the analyzer will be trimmed to that wrong composition, and it will then report process gas incorrectly by exactly that error. A validation run against the same cylinder will confirm the analyzer perfectly, giving false confidence, while the custody measurement is quietly biased. Only an independent validation cylinder, certified separately, can catch this, because it is unlikely that two independently prepared and certified blends carry the same error, so a disagreement between the analyzer and the independent cylinder flags the problem.
This is why custody measurement practice treats the independence of the validation blend as a control, not a nicety. The validation cylinder should be a separately certified blend, and it should be reserved for validation and not quietly rolled into service as the calibration blend when the calibration cylinder runs low, because doing so collapses the two roles back into one and the check loses its value from that day forward. Documenting which cylinder is the calibration standard and which is the independent validation standard, and keeping them distinct, is part of demonstrating that the analyzer's accuracy has been proven honestly rather than assumed.
A single validation run tells you whether the analyzer is within tolerance today, but the real value of validation appears when the results are trended over time. Each validation run produces, for every component, the difference between what the analyzer reported and what the independent cylinder is certified to contain. That difference is a direct measure of the analyzer's bias on that day, and a sequence of validation runs is a record of how that bias moves. Plotted over weeks and months, the validation error becomes a drift metric that shows whether the analyzer is stable, slowly wandering in one direction, or jumping around, long before the error grows large enough to matter to a custody statement.
A cloud SCADA platform such as Merobix is well placed to capture this because it can log every validation result as a data point per component and hold it as a continuous history alongside the analyzer's routine operation. Rather than treating a validation as a pass or fail event that is recorded on paper and forgotten, the platform can trend the signed error for methane, the heavier hydrocarbons, and the inerts across successive validations, so a gentle upward creep in the error on a single component is visible as a slope rather than only as an eventual out-of-tolerance alarm. That early visibility lets a measurement team schedule a recalibration or investigate a maintenance issue on their own terms instead of being surprised by a failed validation during an audit.
Treating validation results as a first-class trend also strengthens the audit story for a custody point. When a measurement analyst or an auditor asks how the analyzer's accuracy has been maintained, a documented history of independent validation runs with small, stable errors is far more convincing than a stack of calibration certificates alone, because it proves the analyzer was repeatedly checked against gas it was not tuned to and passed. Setting an alarm on a validation error exceeding a chosen limit, and a softer alert on a validation trend drifting steadily even while still in tolerance, turns the validation cylinder from a periodic formality into an ongoing, monitored guarantee that the custody measurement is honest.
A calibration cylinder is the certified reference blend the analyzer is adjusted against, so the GC's response factors are trimmed to match its known composition. A validation cylinder is a separate, independently certified blend run through the analyzer only to check it, with no settings changed, so the result honestly shows whether the GC still reads true. The calibration cylinder sets the analyzer, and the validation cylinder tests it.
Because the analyzer is trimmed to report that cylinder's composition, so running the same cylinder again always agrees and the check proves nothing. It cannot expose bias, nonlinearity, or even an error in the cylinder's own certificate, since the same cylinder cannot reveal a fault it helped create. Only an independently certified validation blend, one not used to adjust the instrument, can genuinely test the analyzer's accuracy.
Each validation run gives, per component, the difference between what the analyzer reported and the independent cylinder's certified value, which is a direct measure of the analyzer's bias that day. Logging that signed error from run to run turns validation into a drift metric, so a slow creep in one component's error shows up as a slope well before it becomes an out-of-tolerance failure. A cloud platform can trend these results continuously and alarm on both a hard tolerance breach and a steady drift while still in tolerance.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.