An air calibration records how an ultrasonic gas meter behaves in air. A study of two G4 meters compared that baseline with measurements in four gas mixtures at matched flow points. Two results serve different purposes: the indication error against a reference and the change in that error from the air baseline. For one specimen at maximum flow, the 23% hydrogen mixture produced a reported indication error of −1.05%, while the pointwise change from air was −1.59 percentage points. This article explains both quantities, the study conditions behind them, and the information a buyer needs to assess whether a calibration applies to the intended gas.

A utility can receive gas from sources with different compositions over the life of an ultrasonic gas meter. An air-calibration certificate is useful for recording the meter's reference performance under one test condition. A reviewer also needs to know how the same meter reads when gas composition changes. Shi, Zhang and Song tested this question with an actual-gas reference system and two domestic G4 meters [1]. Their test permits two comparisons at the same flow. The first compares a meter reading to the actual reference volume in the selected medium. The second asks how the result in that medium has moved from the result in air.

The distinction is important when the air baseline itself differs from zero. A small absolute error in the application gas can coexist with a larger shift from air. A useful test report gives enough information to see both outcomes and to determine whether the media and operating conditions are relevant to the application.

In the source's simplified description, a pair of transducers measures the acoustic transit time downstream and upstream. L is the acoustic path length, cf is the speed of sound in the gas, θ is the angle between the path and the pipe axis, and vm is the mean velocity along the measuring line as defined in the paper.

S is the cross-sectional area in m² and vm is in m/s, so q is in m³/h under the stated relationship. The reciprocal-time difference cancels cf in the idealized equations when the same sound speed applies to the two directions. The result explains the measurement principle. Evaluating a finished meter also involves the transducers, the signal timing and the meter's flow geometry, so the equation alone cannot establish the error obtained with a different gas.

Figure 1. Transit-time geometry, redrawn in English from Shi, Zhang and Song [1], Figure 1 (PDF p. 2, print p. 255). The diagram is not to scale.

The authors tested one G4 meter from each of two manufacturers. The reference system had a stated standard volume of 120 L. The reported expanded uncertainty for the system was 0.25% at k = 2. The reported test conditions were 101 kPa and 20 °C. The temperature difference among the relevant locations and medium was kept within 1 °C under the test description. Each medium was tested at 0.1qmax, 0.2qmax, 0.4qmax, 0.7qmax and qmax.

The media were air and four gas mixtures. The study explicitly treated the blend with 23% hydrogen and 77% methane as outside the intended range of the tested meters. That blend is an exploratory case. Test results from it help explain the distinction between the reported quantities, while approval for a commercial application requires evidence for that configuration and gas.

For each case, the authors reported the percent indication error using the indicated volume Vi and the actual reference volume Vref.

E = [(Vi − Vref) / Vref] × 100% (3-1)

The table retains the reported indication errors. Each reading already compares indicated volume against the reference in that test condition.

For Meter 1 at 0.2qmax, air produced an error of −0.01% and the 23% hydrogen blend produced +1.71%. The difference is +1.72 percentage points. Meter 2 at qmax shows a different pattern. Its hydrogen-blend error was −1.05%, while the air value was +0.54%. Subtracting the air result gives −1.59 percentage points. This calculation shows how results on opposite sides of zero can make the shift larger than the absolute error in the application gas. Meter 2 at 0.7qmax gives another example, with −0.97% in the blend and +0.71% in air, a shift of −1.68 percentage points.

The methane-rich mixtures are also worth retaining in the table. At 0.4qmax, Meter 1 reports +0.80% in the 99.9% methane case while Meter 2 reports +0.16%. The corresponding air values are +0.05% and +0.73%. The resulting changes move in opposite directions. Treating the result as a single composition-dependent correction common to both specimens would obscure this difference.

The source calls Figure 3 a “maximum mean error offset.” The curves are consistent, to the resolution of the published plots, with the signed gas error minus the air error for the same meter and flow point in Tables 2 and 3. The article gives no separate equation or replicate data for how a mean or maximum was obtained. The redraw therefore labels the calculation as a pointwise difference. An acceptance report should state its averaging procedure and applicable flow bands.

Figure 2. Indication errors for G4 meters 1 and 2. English data redraw of source Figure 2(a–b), using the reported entries in Shi, Zhang and Song [1], Tables 2–3 (PDF p. 3, print p. 256).

Figure 3. Air-to-gas differences for G4 meters 1 and 2. English counterpart to source Figure 3(a–b) (PDF p. 4, print p. 257). The curves are pointwise differences recomputed from Tables 2–3, in percentage points. The plotted values agree at chart-reading precision with this calculation. The authors call the plot “maximum mean error offset” and do not document a separate averaging or maximization procedure.

The comparison shows why a qualification report should preserve both the absolute indication errors and the changes from the air baseline. A buyer comparing options can ask for the gas compositions, meter variants, flow points, pressure and temperature conditions, and uncertainty associated with the measurements. The study involved two specimens and one stated pressure and temperature regime. Further work could replicate the comparison across several units of each design and across the actual composition and temperature envelope in service. Raw repeat readings would help distinguish systematic shifts from scatter.

qmax means maximum flow. A pointwise shift is expressed in percentage points and calculated here as the gas-medium indication error minus the air indication error for the same meter and flow point. The source experiment and results apply to the reported test specimens. The standard referenced in the paper documents the original procedure.

[1] Shi, A., Zhang, J., and Song, S. (2021). "Experimental analysis of gas-composition effects on ultrasonic gas meter measurement" [Experimental analysis of gas-composition effects on ultrasonic gas meter measurement]. City Gas, 2021 supplement, pp. 254–257. Equations (2-1)–(2-4) and (3-1), Tables 2–3 and Figures 1–3. Figure 1 is an English schematic redraw. Figures 2–3 use the reported values in Tables 2–3, with the pointwise transformation identified in Figure 3.

Cover photo: National Institute of Standards and Technology Digital Collections, Gaithersburg, MD 20899. NIST engineer Robert Palla checking a test dryer. U.S. Government work, Public Domain Mark 1.0 per the source record. Editorial photo crop and title layout. https://commons.wikimedia.org/wiki/File:Engineer_Robert_Palla_of_the_Advanced_Measurement_Technology_Section_checking_one_of_the_test_dryers.jpg