As-found vs as-Left: What a Pressure Transmitter Calibration Certificate Must Prove

A calibration certificate is the only durable evidence that a pressure transmitter was inside specification on a particular day. The two data sets that make that evidence credible — as-found and as-left — are also the two that are most often recorded badly. This guide explains what each set actually proves, how to read a five-point table without being fooled by it, and what a pressure transmitter calibration certificate must contain before an auditor, a customer or a notified body will accept it.

The short version: as-found data is captured before any adjustment and proves how the transmitter behaved in service. As-left data is captured after adjustment and proves it is fit for the next interval. If the two look identical at every test point, the first thing an auditor will ask is whether the as-found set was really recorded before the trim.

As-found and as-left answer two different questions

Both data sets come from the same calibration run, but they support different conclusions.

As-found is the performance of the transmitter as it arrived at the bench, before zero, span or sensor trim. It answers a backward-looking question: was this device within specification while it was in service? That is the number that matters when a batch has to be released, a safety function has to be justified, or a drift trend has to be explained.

As-left is the performance after adjustment. It answers a forward-looking question: is this device fit to go back into service for another interval? It is the number that justifies returning the instrument to the loop.

The sequence matters more than most procedures admit. The complete as-found set has to be recorded and identified before the first adjustment, because once the zero or span is trimmed, the original evidence is gone. Modern calibrators handle this by locking the as-found block and writing it to the record before the adjustment step is unlocked. Paper-based workflows rely on discipline, which is why “as-found equals as-left at every point” remains one of the most common findings in calibration audits.

Two situations are worth separating from the normal case:

New instrument. There is nothing to drift from yet, so the as-found set is an incoming-inspection baseline. Record it anyway — it becomes the reference for every later drift calculation.
Verification without adjustment. Some regulatory checks require proof of condition only. That is a legitimate outcome, but the certificate must say so explicitly, for example “no adjustment performed; as-left equals as-found.” Silence on this point looks like a missing record.

Reading a five-point record: the arithmetic behind the table

A 4–20 mA transmitter with a 0–300 psi range (LRV 0 psi, URV 300 psi) is tested at five points. The expected output at any pressure is fixed by the transfer function, so the certificate can be checked line by line:

I_ideal (mA) = 4 + 16 × (P − LRV) ÷ (URV − LRV)
Error (% of span) = (I_measured − I_ideal) ÷ 16 × 100

The table below is the same record as a typical certificate, with the two columns most certificates leave out: the error expressed as a percentage of span, and an explicit pass/fail result against the stated tolerance.

As-found and as-left record, 0–300 psi / 4–20 mA transmitter, tolerance ±0.02 mA (±0.125% of span)

Test point	Applied pressure	Ideal output	As-found	Error (% span)	As-left	Tolerance	Result
0% (LRV)	0 psi	4.000 mA	4.02 mA	+0.125%	4.00 mA	±0.02 mA	Pass (at limit)
25%	75 psi	8.000 mA	7.98 mA	−0.125%	8.00 mA	±0.02 mA	Pass (at limit)
50%	150 psi	12.000 mA	12.03 mA	+0.188%	12.00 mA	±0.02 mA	Fail
75%	225 psi	16.000 mA	15.97 mA	−0.188%	16.00 mA	±0.02 mA	Fail
100% (URV)	300 psi	20.000 mA	19.98 mA	−0.125%	20.00 mA	±0.02 mA	Pass (at limit)

Three things in that table deserve attention, and none of them are visible if you only look at the as-left column.

1. The transmitter was genuinely out of tolerance

At 50% and 75% the as-found error is 0.03 mA against a 0.02 mA limit — 1.5 times the tolerance, not a rounding artefact. That is a real out-of-tolerance result, recorded before adjustment. Two further points sit exactly on the limit, which most quality systems treat as a warning rather than a comfortable pass. The as-left column does not erase any of this: a certificate with clean as-left data and unmarked as-found data tells half the story.

2. The error alternates in sign, which rules out simple drift

A zero shift moves every point in the same direction by the same amount. A span error grows with pressure. Neither pattern fits a record that reads +0.02, −0.02, +0.03, −0.03, −0.02 mA. That looks like scatter — and when the scatter is the same size as the tolerance, the problem may be the measurement, not the transmitter.

This is what the test uncertainty ratio is for. A working rule is that the reference standard plus the measurement chain should be at least four times better than the tolerance being judged (TUR ≥ 4:1). For a ±0.02 mA tolerance, the combined uncertainty of the standard, the pressure control and the mA measurement should be around 0.005 mA or better. If it is not, no one can say whether a 0.03 mA result is drift or noise.

3. As-left values that land exactly on nominal are a question, not a proof

4.00, 8.00, 12.00, 16.00 and 20.00 mA at every point is what a digital trim aims for, but a real loop still shows a few counts of scatter after adjustment. Values that are perfect at two decimal places usually mean one of two things: the readings were rounded hard, or they were copied from the setpoint instead of measured. Recording three decimals (4.001, 8.002, 12.000) removes the ambiguity and takes no extra effort on an automatic system.

Why five points, and why the direction matters

A two-point check at 0% and 100% proves span. It says nothing about linearity, hysteresis or repeatability, which is where most transmitter faults actually appear.

Five points (0/25/50/75/100%) give a linearity claim and localise a fault to a region of the range.
Up-and-down runs at each point expose hysteresis — the difference between the reading approaching from below and from above. A transmitter that passes on the way up and fails on the way down has a mechanical or diaphragm problem that a rising-only test will miss.
Three pre-conditioning cycles over the full range, before any data is recorded, settle the diaphragm and remove the effect of storage.
A dwell time per point long enough for the reading to stabilise, plus a leak check at full scale, is what separates a valid record from a fast one.

None of these steps shows up in the certificate. That is exactly why the certificate should reference the procedure and state the sequence — up/down, number of cycles, dwell time — rather than only listing five numbers. For the mechanics of venting, pressurising and trimming a loop, see our step-by-step guide to how to calibrate a pressure transmitter.

What a calibration certificate must contain

The list below is what a reviewer, an auditor or a customer will look for on a pressure transmitter calibration certificate. The right-hand column is where certificates most often fall short.

Required content of a pressure transmitter calibration certificate

Field	Why it is asked for	Commonly missing
Unique certificate number, page x of y	Makes the record traceable and prevents page substitution	Pagination on multi-page certificates
Instrument identity: manufacturer, model, serial number, tag / loop number, range, output, accuracy class	Proves the certificate refers to this device, not a similar one	Tag number, accuracy class
Condition as received	Distinguishes a calibration from a repair; supports the out-of-tolerance assessment	Almost always absent
Date of calibration and ambient conditions (temperature, humidity)	Reference values are only valid inside a stated environment	Humidity, actual recorded values
Method or procedure reference	Shows the test points, sequence and acceptance criteria were defined in advance	Procedure number and revision
Reference standards: ID, serial, range, uncertainty, certificate number and calibration due date	This is the traceability chain; without the certificate number the claim is unverifiable	The standard’s own certificate number and due date
As-found and as-left data at each test point, with units	The core evidence, and the input to every drift and impact decision	Units, test direction, pass/fail marking
Measurement uncertainty statement with coverage factor	Without it, a result close to the limit cannot be judged	Coverage factor (usually k = 2)
Statement of conformity and the decision rule used	Simple acceptance and guard-banded acceptance give different verdicts on the same data	Decision rule, in almost every certificate
Adjustments performed (zero, span, sensor trim) and any repairs	Explains the difference between as-found and as-left	Which trim was applied and by how much
Technician name, signature, date; reviewer or approver	Accountability, and a second pair of eyes on an out-of-tolerance result	Reviewer signature
Calibration due date or recommended interval	Drives the recall schedule; without it the certificate has no forward effect	Interval basis

The decision rule deserves a sentence of its own. When a reading sits exactly on the tolerance limit — as the 0% and 100% points do in the table above — a simple acceptance call is not defensible if the measurement uncertainty is a meaningful fraction of that limit. Either apply a guard band (accept only if the error plus the expanded uncertainty stays inside the limit) or state plainly that simple acceptance was used and why it is acceptable for this instrument. Either answer is fine. Having no stated rule is not.

Traceability: what “NIST traceable” has to mean

Traceability is an unbroken chain of comparisons, each link documented with its own uncertainty, ending at a national or international standard. On a certificate, that chain is only real if a reader can follow it:

The reference standard is identified by serial number and by its own calibration certificate number.
That certificate is within its validity period on the day of the calibration.
The laboratory that issued it is named, and the statement says how the chain reaches a national metrology institute such as NIST or an equivalent body.
The uncertainty at each link is small enough to support the tolerance being judged.

One distinction prevents a lot of confusion in procurement. ISO/IEC 17025 is a standard for the competence of testing and calibration laboratories. It is not a specification for transmitters, and holding an accredited calibration does not make an instrument more accurate. What accreditation adds is that an accreditation body has assessed the laboratory’s methods, uncertainty budgets and competence, and that the certificate carries an accreditation symbol with a defined scope. A manufacturer’s factory certificate can be perfectly valid and fully traceable without being accredited — but the two documents answer different questions, and regulated buyers usually want to know which one they are holding.

When as-found data is out of tolerance

An out-of-tolerance as-found result is not a paperwork problem; it is a quality event with a time window attached. The sequence that survives audit looks like this:

Tag and quarantine the instrument. No further use, no quick trim back into service before the record is complete.
Quantify the drift. Compare the current as-found set with the previous certificate and calculate the error per interval, not just the error today.
Bracket the suspect period. The device was probably inside specification at the last calibration and outside it at this one. Everything measured with it in between is potentially affected until evidence says otherwise.
Assess the impact. Which batches, safety functions or custody-transfer readings depended on this loop? This is the step that turns a calibration record into a quality record, and it is the first thing an auditor asks to see.
Close the loop on cause. Overpressure events, pressure spikes from a pump, moisture in the impulse line, diaphragm damage, a blocked impulse line or simple electronics ageing all produce different drift patterns. The pattern in the as-found data usually points at the cause — which is one more reason to keep all five points.
Decide the next interval. A device that drifted out in six months should not come back in twelve. Shortening the interval is a documented, defensible outcome; quietly recalibrating and moving on is not.

Electronic records, audit trails and signatures

Paper certificates fail in predictable ways: a missing page, an illegible transcription, an as-found column filled in after the fact. Electronic records solve those problems only if the system is built for it.

Time-stamped, computer-generated audit trails that record who changed what, when, and why — and that cannot be edited or deleted by the person who made the change.
Locked raw data. The as-found set is written once, before adjustment, and cannot be overwritten by the as-left set.
Signatures linked to the record, with printed name, date, time and meaning (performed, reviewed, approved), so a signature cannot be copied onto another record.
Unique user accounts. Shared logins destroy the evidential value of an electronic signature faster than any technical failure.
Retention and export. The record has to be readable years later, in a human-readable form, on a system that may no longer exist.

In pharmaceutical and medical-device environments, 21 CFR Part 11 sets the baseline for electronic records and signatures: validated systems, secure time-stamped audit trails that do not obscure prior data (§11.10(e)), signature manifestations that show name, date/time and meaning (§11.50), and signatures linked to their records so they cannot be excised or transferred (§11.70). Even outside the FDA’s scope, the underlying principles — attributable, legible, contemporaneous, original, accurate — are the ones a customer audit will test.

Building the record into the calibration itself

Most record defects are created by the workflow, not by the technician. Hand-pumping to a point, reading a dial, writing a number and later typing it into a spreadsheet introduces three opportunities for error per test point. An automatic pressure calibration system removes them by making the record a by-product of the calibration:

Pressure is generated and held automatically to each set point, so the dwell is repeatable and the up/down sequence is defined by the program rather than by habit.
The device under test is powered and its mA output measured by the same system, so the reading is captured, not transcribed.
The as-found block is stored and locked before the adjustment step, which is the single most useful control an auditor can see.
Error against tolerance, the pass/fail verdict and the uncertainty budget are calculated by the software, and the certificate is generated with the reference standards’ IDs and due dates already attached.
The record lands in a database with an audit trail, a due date and a recall schedule — which is what turns calibration from a yearly scramble into a controlled process.

The right hardware depends on where the work happens. Laboratory benches handling high volumes of transmitters, gauges and switches need automated pressure generation and switching; field work needs a portable automatic pressure calibrator that still produces the same data structure; single-loop checks need a handheld pressure source that still feeds the same records system. If the loop uses HART, digital trim should be performed through the protocol and the analog output verified afterwards — a certificate that only shows the analog result cannot prove the digital path was left correct.

Calibration record checklist

As-found data recorded and time-stamped before any adjustment
Five points minimum, up and down, with the sequence stated
Three pre-conditioning cycles and a full-scale leak check
Ambient temperature and humidity recorded, not assumed
Reference standards identified by serial number, certificate number and due date
Error expressed against a stated tolerance, with pass/fail per point
Uncertainty statement with coverage factor and the decision rule
Adjustments and repairs described, with the trim values
Technician and reviewer signatures, both dated
Next due date set, and the interval justified by the drift observed

Frequently asked questions

Should the as-found and as-left readings be identical?

No. If a transmitter arrives in tolerance and needs no adjustment, the two sets will be close but rarely identical to three decimals. Identical values at every point are possible, but they are also the classic signature of an as-found column that was filled in after the trim. If your records look like that consistently, the workflow — not the instrument — is what needs checking.

Is a manufacturer’s calibration certificate the same as an accredited calibration?

Not the same, though both can be valid. A manufacturer’s certificate states the results and the standards used and is usually traceable. An accredited calibration adds independent assessment of the laboratory’s competence and uncertainty claims, and carries an accreditation symbol with a defined scope. Regulated industries and many tier-one customers specify the accredited version; general industrial users often do not need it.

How often should a pressure transmitter be calibrated?

There is no universal interval. It should follow the criticality of the loop, the duty (pressure cycling, overpressure, temperature, corrosive or wet media) and the drift actually observed in previous certificates. Six to twelve months is common for critical loops, longer for non-critical indication only. The defensible approach is to set an initial interval, then adjust it using recorded as-found drift instead of leaving it untouched for years.

Does calibrating the transmitter calibrate the whole loop?

No. A transmitter certificate covers the transmitter and its output. The receiving card, wiring, isolator and PLC analog input are separate links and are proven by a loop check, not by a transmitter calibration. When a reading is wrong, the certificate tells you the transmitter was not the cause — which is only useful if the rest of the chain has its own records.

The record is the deliverable

A calibration that produces a number but not a defensible record has solved today’s problem and created next year’s. The as-found set is the only evidence of how the instrument behaved in service; the as-left set is the only evidence that it is fit to go back. Capture both, capture them in the right order, and state the tolerance, the uncertainty and the decision rule that turned the numbers into a verdict. Everything else on the pressure transmitter calibration certificate exists to make those three things checkable by someone who was not in the room.