When a chromatograph measures a gas, the raw mole percentages it computes for each component rarely add up to exactly 100 percent. Physical composition must total 100 percent, so the analyzer scales each component to force closure, a step called normalization. This guide explains why the unnormalized total drifts off 100 percent, how normalization redistributes the shortfall or excess across the components, why an acceptance window on the raw total flags a bad run, and why forcing normalization can hide an underlying calibration drift that the raw total would have exposed.
GC Component Normalization in one line: GC component normalization is the step where a chromatograph scales each measured mole percent so the components sum to exactly 100 percent, since a real composition must total 100 while the raw measured sum rarely does. The raw, unnormalized total is compared against an acceptance window, such as roughly 99 to 101 percent, and a total outside that window flags a bad analysis. Normalization is necessary, but because it forces closure regardless, it can mask a slow calibration drift that the unnormalized total would otherwise reveal.
A chromatograph does not measure mole percent directly. It measures a detector response for each component as it emerges from the columns, and converts each response to a mole percent using a response factor derived from calibration. Every one of those conversions carries a little uncertainty, from the detector, from the calibration, from small run-to-run variation in injection and separation. When the individually computed mole percents are added up, those small errors accumulate, and the raw sum lands close to 100 percent but almost never exactly on it.
The physical truth, though, is unambiguous: the mole fractions of everything in the gas must sum to one, or 100 percent, by definition. So the raw unnormalized total is not a measurement of anything physical; it is a measure of how well the analysis reproduced a known constraint. A raw total of, say, 99.7 or 100.4 percent tells you the sum of the small errors in this particular run, which is useful diagnostic information precisely because the true value is known in advance.
This makes the unnormalized total a built-in check that a chromatograph gives you for free. Unlike most measurements, where you cannot know the right answer, here you know the components must total 100 percent, so any departure quantifies the run's error without a reference gas. That is why the raw total is watched, not discarded, even though the composition ultimately reported is the normalized one.
Because the flow computer and every downstream calculation expect a composition that sums to 100 percent, the analyzer normalizes the raw results before reporting them. The standard approach scales every component by the same factor, the ratio of 100 percent to the raw total, so that the scaled values add up to exactly 100 while each component keeps its share of the whole. A raw total of 100.4 percent, for instance, is brought to 100 by shrinking every component proportionally by that small excess.
Proportional normalization preserves the relative composition, the ratios between components, while adjusting the absolute total. This is the intended behavior for feeding heating value and compressibility calculations, which depend on the component fractions summing correctly. It also means the correction is spread across the whole composition rather than dumped on one component, so no single mole percent is distorted disproportionately by the closure step under normal, small departures from 100 percent.
The acceptance window is what keeps normalization honest. Before trusting a run, the analyzer or the measurement system checks the raw unnormalized total against limits, commonly a window such as roughly 99 to 101 percent, and a total outside that window flags the analysis as bad. The logic is that a small departure from 100 percent is ordinary measurement scatter that normalization can safely absorb, but a large departure means something went wrong, a leak, a bad injection, a component missed, or a calibration far off, and that run should not be normalized and used as if it were valid.
The hazard of normalization is that it always produces a composition that sums to exactly 100 percent, whether or not the underlying analysis is trustworthy. If a detector or a response factor drifts slowly over weeks, the raw total may creep away from 100 percent, but as long as it stays inside the acceptance window, normalization scales the run back to 100 and reports a clean-looking composition. The reported result hides the drift; only the trend in the unnormalized total reveals it. Watching the normalized composition alone, an operator sees nothing wrong right up until the drift is large enough to fail the window.
This is why the unnormalized total is treated as a health signal and not just a pass or fail gate. A raw total that is slowly walking toward the edge of the window, even while still passing, is early evidence of calibration drift and a cue to recalibrate or investigate before analyses start failing or, worse, before a biased-but-passing composition feeds the custody calculation. The point of normalization is to deliver a usable composition, but the point of monitoring the raw total is to make sure that composition is usable for the right reasons.
A cloud SCADA such as Merobix supports this by trending the unnormalized total over time and across analyzers, not just the reported composition. When the raw total is surfaced and trended, a measurement team can see a chromatograph's total drifting toward its limits and act before it fails, and can compare analyzers to catch one that is quietly degrading. Alarms on the raw total approaching or exceeding the window turn a silent, normalization-hidden drift into a visible, actionable condition, which is exactly the failure mode that watching only the normalized composition would miss.
Because a real gas composition must sum to exactly 100 percent, while the raw mole percents the analyzer computes from detector responses almost never do, owing to small accumulated measurement and calibration errors. Normalization scales every component proportionally so the total is exactly 100, giving the flow computer a composition that closes correctly for heating value and compressibility calculations.
It is a range, commonly around 99 to 101 percent, that the raw unnormalized total must fall within for the analysis to be accepted. A small departure from 100 percent is ordinary scatter that normalization can absorb, but a total outside the window signals a problem such as a leak, bad injection, missed component, or far-off calibration, and that run is flagged as bad rather than used.
Normalization always forces the composition to sum to 100 percent, so a slow calibration drift that pushes the raw total off 100 is scaled back to a clean-looking result as long as it stays inside the acceptance window. The reported composition looks fine while the drift is invisible in it. Only the trend in the unnormalized total exposes the drift, which is why that raw total should be monitored, not discarded.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.