An orifice meter run is not flow-calibrated the way a turbine, Coriolis, or positive-displacement meter is. Its accuracy is inherited from three things: dimensional conformance of the primary element and tube to AGA Report No. 3 (or ISO 5167), traceable calibration of the differential pressure, static pressure, and temperature transmitters feeding the flow computer, and correct entry of the calculation constants in that flow computer. The deliverable that settles a billing dispute is therefore a documented acceptance package plus an uncertainty budget computed to ISO 5168 - not a shop flow test. If the client still insists on a physical proof, the only version that survives cross-examination is comparative proving against a master meter at line pressure, temperature, and gas composition.
Reading the Objection
The stated complaint - "the meter tube and flow computer will not measure accurately" - is almost never the real one. On a distribution header feeding several 2-inch, 600# customer skids off a single custody station, the client's actual exposure is the imbalance between the parent custody meter and the sum of the child meters. That imbalance always exists, and it is not evidence of a defective meter run. Separating the two questions is the first engineering task.
| Stated objection | What it actually tests | Correct response |
|---|---|---|
| "Manufacturer guarantees are not acceptable" | Traceability, not accuracy | Supply as-left calibration certificates traceable to a national standard, with as-found data, for each transmitter and the plate bore |
| "Calculations and simulations are not a real representation" | Whether the installation conforms to the empirical database behind the discharge coefficient | Dimensional inspection report against AGA-3 Part 2 tolerances: tube bore, roundness, straight lengths, tap location, plate bore, edge sharpness, flatness |
| "I am losing money" | Header balance, not meter performance | Uncertainty budget per meter to ISO 5168, plus an expected unaccounted-for-gas band derived from the combined uncertainties of parent and child meters |
| "Test it with gas" | Nothing, if the test fluid and conditions differ from service | Comparative proving at line conditions, or documented conformance in lieu of it |
Mechanism: Inferential Flow, Not Measured Flow
The term "orifice meter" describes an inferential device. Nothing in the meter run counts volume. The flow computer solves the AGA-3 mass flow equation from four inputs - differential pressure across the plate, absolute static pressure, flowing temperature, and gas composition - and multiplies by geometry terms and a discharge coefficient. The discharge coefficient is not measured on your tube; it comes from a regression fitted to thousands of laboratory calibrations of geometrically similar installations. That is why conformance is the calibration. If the bore, edge, tap set, and straight lengths sit inside the ranges the correlation was fitted over, the coefficient applies with the uncertainty the standard publishes. If they sit outside, no amount of shop testing recovers a defensible number without a full flow calibration of that specific tube.
Two consequences follow, and both matter for billing:
-
Square-root sensitivity. Flow varies with the square root of differential pressure, so a fractional DP error propagates at half strength:
dQ/Q = 0.5 x (dDP/DP). That sounds forgiving until you note that transmitter accuracy is specified as a percentage of span. At 10% of DP span, a 0.1%-of-span transmitter error is 1% of reading, giving 0.5% flow error. Turndown, not transmitter grade, is the dominant error at low flow. - Half-power sensitivity to P, T, and relative density. Because base-condition volume scales with the square root of the density ratio, a 0.27% error in absolute pressure (1 psi at roughly 365 psia) produces about 0.14% in flow; a 1 degF error at about 520 degR produces about 0.10%. Composition enters the same way. These are the terms a client's "lost money" argument actually lives in, and they are auditable line-by-line.
Governing Standards and What Each Settles
| Document | Settles |
|---|---|
| AGA Report No. 3, Orifice Metering of Natural Gas | The equation, the discharge coefficient, installation and specification requirements, and the uncertainty guidelines. Part 2 carries the dimensional tolerance and straight-length tables you inspect against. |
| AGA Report No. 8 | Compressibility and supercompressibility of natural gas from composition, used by the flow computer to get density at flowing and base conditions. |
| ISO 5167 | The equivalent primary-device standard: upstream and downstream pipework configuration, plate geometry, tap arrangements, flow conditioners. |
| ISO 5168 | The method for combining component uncertainties into a total system uncertainty. This is the calculation third-party consumers and regulators accept in lieu of proving. |
AGA-3 is the transmission industry's cash register in the United States, and its meters are verified at least annually. Quote that verification interval in the O&M documentation; a client who demands a factory flow test but has no periodic verification programme has the priority inverted.
Procedure: Assembling the Acceptance Package
- Pull the mechanical data book for each meter tube: certified tube internal diameter measured at the specified locations and temperature, roundness, internal finish, and the certified plate bore with its stamped bore, material, and reference temperature.
- Verify straight-length compliance against AGA-3 Part 2 for the installed beta ratio and the specific upstream fitting. Measure from the plate face, not from the fitting body. Record the downstream length to the pressure control valve and restriction orifice separately - downstream disturbance limits are shorter but they are not zero.
- Confirm the tap arrangement (flange taps versus pipe taps) and the static pressure tap side (upstream or downstream), then confirm both are entered identically in the flow computer. A tap-set mismatch is a silent multi-percent error.
- Calibrate DP, static pressure, and temperature loops five-point up and down (0, 25, 50, 75, 100% of span) with a calibrator at least four times more accurate than the device. Record as-found and as-left. As-found data is what the client is buying.
- Audit every constant in the flow computer against the data book: plate bore and material coefficient of thermal expansion, tube ID and material, base pressure and base temperature, local atmospheric pressure, gauge versus absolute static pressure, relative density or full composition, heating value, DP low-flow cutoff, and the integration and sampling configuration.
- Build the uncertainty budget to ISO 5168: root-sum-square of discharge coefficient, expansion factor, bore and pipe diameter, DP, static pressure, temperature, and compressibility contributions, evaluated at minimum, normal, and maximum customer flow. Publish all three numbers.
- Issue the package as a single signed document per skid, with the annual verification schedule attached.
Comparative Proving: Options and Cost
If the client funds a physical proof, the accepted methods are all comparative - the meter under test is run in series with a reference of known performance. Inline references in normal use are turbine meters, ultrasonic meters, a master orifice meter, or a bell prover for low-pressure work. A vendor shop equipped for this runs a closed loop with the meter tube installed in its production spool arrangement; a shop without that capability should subcontract rather than improvise.
The controlling rule: the closer the test conditions are to service conditions - gas composition, pressure, temperature, and Reynolds number - the fewer corrections you apply, and every correction you apply is another argument the client can open. A proof run on air at 30 psig and then corrected up to natural gas at 350 psig is a calculation wearing a test's clothing, and a client who rejected your engineers' calculations will reject it for the same reason. Costs on comparative proving are usually prohibitive for a 2-inch customer skid and only justify themselves when the metered volume makes the exposure large.
Why a Bottled-Gas Blowdown Proves Nothing
The proposed setup - a bank of nitrogen bottles, a manifold and regulator, a reference flow instrument at the skid inlet, venting to atmosphere - fails on four independent counts, any one of which is fatal.
- Flow rate. A bottle bank cannot sustain the mass flow the skid passes at 340-350 psig for long enough to integrate a meaningful volume. Regulator droop and bottle cooling as the bank empties make the source non-steady, so the reference and the meter are never seeing the same conditions at the same instant.
- Reynolds number and DP. At bottle-fed rates and low line pressure, the operating DP falls to the bottom of the transmitter span and the Reynolds number falls out of the range the meter was sized in. The discharge coefficient is Reynolds-dependent, so you are testing the meter at a point on the curve it will never run at.
- Traceability of the reference. The whole test rests on the inlet instrument. The first question is "prove to me that is the real flow rate," and answering it requires the same calibration certificate chain you were trying to avoid producing.
- Correction burden. Nitrogen or air at atmospheric discharge has different density, compressibility, and isentropic exponent from the billed gas. Converting the result back to service conditions reintroduces exactly the calculation the client already rejected.
Do not spend money on this test. It generates a number that is weaker evidence than the conformance package it was meant to replace.
Verification Checks with Expected Readings
- DP loop. Apply 0, 25, 50, 75, 100% of DP span from the calibrator with the manifold isolated. Expect the flow computer's displayed DP to track the calibrator within the transmitter accuracy plus input conversion error at every point, up and down, with no hysteresis beyond that band.
- Static pressure loop. Same five points. Expect the displayed value to match the calibrator, and confirm the sense: if the transmitter is gauge, the flow computer must be adding the entered atmospheric pressure to reach absolute. Verify that entered atmospheric value against site elevation, not the 14.73 psia default.
- Temperature loop. Substitute a precision decade resistance box for the RTD at three points spanning the operating range. Expect agreement with the resistance-to-temperature conversion for the configured RTD curve within the flow computer's stated input accuracy. Confirm degF versus degC.
- Zero-flow check. Close the block valves and open the equalizer. Expect DP at zero within the transmitter zero tolerance, computed rate at zero, and the accumulator to stop advancing. Record the DP low-flow cutoff value that is actually enforcing that zero.
- Plate verification. Measure the bore with a bore gauge at several diameters, record the metal temperature, and inspect the upstream edge for sharpness and the plate for flatness. Expect the measured bore to match the stamped bore within the standard's tolerance table and to match the flow computer entry exactly.
- Log integration. Pull one hourly record and hand-integrate the averaged DP, P, and T against the reported volume. Expect agreement to the resolution of the log. A mismatch here exposes averaging or sampling misconfiguration.
Recurring Failure Modes on Small Solar Metering Skids
Plate orientation is the most common single cause of a real billing error: the sharp square edge faces upstream and the bevel faces downstream, and in a dual-chamber orifice fitting the plate can be seated correctly in the carrier but the carrier inserted reversed. Check the paddle stamping direction physically, not from the data book.
Sampling method inside the flow computer is the second. Flow is proportional to the square root of DP, so the device must extract the square root at the sampling interval and average the result. Averaging DP over the interval and then taking the square root under-reads whenever flow fluctuates, and it under-reads in the seller's favour, which is a difficult conversation to have after the fact.
Turndown is the third. The downstream restriction orifice sets each customer's flow band, so the operating DP may sit permanently low in the transmitter span. Verify that minimum billed flow still produces DP in the usable part of the span; if it does not, re-range the transmitter or stack a second DP element rather than accepting the error.
On solar-powered installations, low-voltage events during winter or extended cloud cover can interrupt logging and reset accumulators. Confirm the battery autonomy against worst-month insolation, and confirm that the flow computer's audit trail flags power interruptions so lost intervals are visible rather than silently zero.
Finally, if the flow computer runs on a fixed relative density and heating value from a single lab sample, seasonal composition drift becomes an uncorrected bias on every invoice. Establish a sampling frequency and a procedure for updating composition, and put it in the O&M document alongside the annual transmitter verification - that combination answers the client's real question about where the money goes.
Frequently Asked Questions
What happens if the orifice plate is installed backwards?
The bevel presents to the flow instead of the square edge, the discharge coefficient no longer matches the AGA-3 correlation, and the meter typically under-reads by several percent. Nothing in the flow computer flags it - only a physical plate pull finds it.
What happens if I test the meter run with nitrogen bottles at low pressure?
You test the meter at a Reynolds number and DP it never operates at, with a non-steady source, and then correct the result back to service conditions - which reintroduces the same calculation the client already rejected. It buys no additional confidence over the conformance package.
What happens if the flow computer's atmospheric pressure constant is left at the default?
Absolute static pressure is biased by the difference between the default and true site barometric pressure. Because base volume scales with the square root of density, a 1 psi error at roughly 365 psia produces about 0.14% flow error - constant, one-directional, and applied to every invoice.
What happens if the customer's flow drops below 10% of the DP span?
The transmitter's percent-of-span accuracy becomes a large percent of reading, and the square-root relationship halves it into roughly 0.5% flow error or worse. Below the configured low-flow cutoff the flow computer records zero and the gas passes unbilled.
What happens if the parent custody meter and the sum of the customer meters do not balance?
They never will. Each meter carries its own ISO 5168 uncertainty, and the readings are taken at different times with line pack in between; publish the expected imbalance band computed from the combined uncertainties, then verify it by re-running the ISO 5168 budget at minimum, normal, and maximum flow for every skid.