As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
Because the numbers that matter in gas detection live at the bottom of the scale. Take an instrument spanning 0 to 100 ppm of hydrogen sulphide with an accuracy of 5 percent of full scale: the error band is plus or minus 5 ppm, flat across the range, and a widely used low alarm set point is 5 ppm. The band is the entire set point. The instrument cannot distinguish the alarm condition from clean air, and no feature anywhere else on the specification sheet repairs that. The effect is structural rather than a defect, and it gets worse as the span gets wider.
An alarm set point can sit entirely inside the error band
Alarm thresholds for toxic gases are deliberately low, because the point of the instrument is to warn before harm rather than to describe a hazard already present. The MSA ALTAIR 2X single gas detector in its low range hydrogen sulphide version publishes a low alarm at 5 ppm and a high alarm at 10 ppm. Those two numbers are what the instrument is for. Everything else on its specification sheet exists to make sure they fire when they should.
What the bullets publish alongside them is traceability and endurance rather than uncertainty: over three hundred logged alarm events, over eight hundred and fifty hours of recorded data including functional check and calibration records, a green indicator flash every fifteen seconds confirming a passed check, a red flash for a failed or expired one, an end of sensor life warning, an expected sensor life of four years, over moulded rubber armour, survival of repeated drops onto concrete, and certified dust and water tight construction to IP67. The accuracy figure and the measuring span are not stated in the listing bullets, which means the arithmetic in the opening paragraph is an illustration about this class of instrument rather than a claim about this one.
That distinction is worth keeping clean. A low range variant exists at all because the manufacturer recognised that a wide span is the enemy of a small threshold, and naming the variant for its low range is a quiet admission of the whole problem this article describes.
Oxygen depletion arithmetic on a wide span
Oxygen monitoring shows the same effect with even less room to spare, because the interesting range is a narrow slice near the top of a percentage scale rather than a trace near zero. The BW Clip single gas oxygen detector publishes a low set point at 19.5 percent and a high set point at 23.5 percent by volume, along with an automated self check, an LCD and IP66 and IP67 ingress ratings. Its span and its accuracy are not stated in the listing bullets.
Oxygen instruments of this type conventionally cover 0 to 25 percent by volume. Apply a 5 percent of full scale figure to that span and the band is plus or minus one and a quarter percentage points, which lands as follows.
| Quantity | Value | Consequence |
|---|---|---|
| Oxygen in normal air | About 20.9 percent by volume | The reference the instrument is zeroed against |
| Published low alarm set point | 19.5 percent by volume | The deficiency threshold that triggers action |
| Gap between the two | 1.4 percentage points | The entire margin a monitor has to resolve |
| Band from 5 percent of a 0 to 25 percent span | Plus or minus 1.25 percentage points | Almost as wide as the gap it has to detect |
| Band from 5 percent of reading at 19.5 percent | Plus or minus 0.98 percentage points | Narrower, but still a large fraction of the margin |
| Band needed to resolve the gap confidently | Better than plus or minus 0.3 percentage points | Why real oxygen cells are specified far tighter than 5 percent |
The bottom row is the point. Nobody ships an oxygen monitor with a 5 percent of full scale error band, because the application would be impossible. Working the arithmetic backwards from the alarm thresholds tells you roughly how good the sensor has to be, and that is a useful thing to be able to do when a listing declines to tell you.
Zero offset and span error are separate problems
A sensor’s total error decomposes into parts that behave differently, and the full scale notation collapses them into one number that hides the structure. Offset is a constant added to every reading, from electronics, from a drifting baseline, from a cell that has aged. Gain is a multiplier slightly away from one, so it contributes nothing at zero and most at the top. Linearity is the curvature between the points where the instrument was adjusted. Drift is any of the above moving with time or temperature. Response time governs how far from the settled value a reading taken too early will sit.
At the bottom of a span, offset dominates everything else, which is exactly why an absolute band bites hardest there. The remedies are specific. Zeroing in genuinely clean air removes the offset at that instant and does nothing for gain. A functional check with a certified concentration proves the chain from sensor to alarm responds, and adjusts nothing. A calibration with certified gas at a known concentration corrects the gain term, which is the only one of these three procedures that changes the numbers the instrument reports. Confusing the second with the third is the most common error in gas detector care, and the difference matters more on instruments whose sensing element is inherently drifty, such as the semiconductor types.
Why manufacturers quote full scale in the first place
There is a legitimate reason and a commercial one. The legitimate reason is that a sensor has a noise floor, and a noise floor is fixed in absolute units, so a guarantee expressed as a fraction of the span is the only form the physics supports near zero. Promising a percentage of reading at one hundredth of span would require the manufacturer to control offset and drift to a level that a low cost cell cannot hold between calibrations.
The commercial reason is that one absolute number is cheap to guarantee and reads as a tight percentage. It also conveniently describes the part of the scale nobody uses, since the readings that drive decisions cluster in the lowest tenth of the range.
Below that there is a third tier, where no performance figure appears at all. The Testo gas leak detector listing publishes no feature bullets of any kind, and the Inficon combustible gas detector listing carries only the product class, the country of manufacture, the manufacturer name and the package dimensions. Neither absence says anything about how the instruments perform. Both say a great deal about how much a buyer can learn from the page, which is the honest thing to record.
Using a detector whose accuracy you cannot pin down
Treat the instrument as a trigger rather than a meter. Rely on the alarm, which the manufacturer has set and checked, in preference to the displayed concentration, which may be anchored to nothing you can verify. Set any adjustable threshold with margin below the level you actually care about, since late alarming is the failure mode an unquantified error band permits. Zero in air you have reason to believe is clean, outdoors and upwind rather than in the plant room. Run a functional check against certified gas to a schedule and keep the records, because an instrument that logs those checks makes the schedule auditable.
Then go and find the real figure where a decision depends on it. Manuals and datasheets routinely publish the span, the error band, the response time and the cross sensitivities that a retail listing omits, and for an entry decision or an exposure record that document is the only acceptable source. The instrument classes most likely to publish it, and the differences between single and multiple sensor designs, are covered in what a multi gas detector is, the four gas detector guide and the hydrogen sulphide detector guide.
Frequently Asked Questions
Does a bump test prove the instrument is accurate?
No. A bump test applies a concentration above the alarm threshold and confirms that the sensor responds and the alarm activates. It is a functional check of the whole chain, from sensor to buzzer, and it is the right thing to do before each use. It does not adjust anything and it does not establish that the number on the display is close to the true concentration.
What does zeroing in clean air actually fix?
It removes the offset error at the moment you do it, by telling the instrument that the present condition is zero. It cannot correct a gain error, because gain only shows up when there is a signal to multiply, and it cannot stop the offset drifting again afterwards. Zeroing in air that is not actually clean writes a negative offset into the instrument, which is worse than not zeroing at all.
Is a narrow range detector always better for low concentrations?
Under a full scale specification it is better at every concentration both instruments can read, because the absolute band is a fixed fraction of the span. The trade is headroom. A narrow range device saturates in a large release and stops telling you how large, which is why some monitors carry a low range sensor variant specifically for trace work.
Why do alarm set points matter more than the displayed number?
Because the set point triggers an action and the number informs a judgement. A detector whose error band straddles its own threshold will alarm late or early by an amount nobody can quantify, so the action it triggers happens at an uncertain concentration. Knowing the band tells you how much margin to build into the set point you choose.
What is cross sensitivity and how does it interact with this?
It is a response to a gas other than the target, which adds an apparent concentration that the accuracy figure does not describe at all. An electrochemical cell can respond to several interfering species at different efficiencies, so a reading near the bottom of a span can be mostly interference. This is one more reason a small number from an unspecified instrument is not evidence.
Do fixed and portable detectors differ in how accuracy is published?
Portable instruments bought for personal monitoring more often publish alarm set points, expected sensor life and ingress protection than they publish an error band, because the buying decision is framed around compliance and durability. Fixed installations are usually specified by an engineer against a document, so the uncertainty figure tends to live in that document rather than on a retail listing.