flowmeter calibrations using a small volume prover - · pdf fileflowmeter calibrations using...

Flowmeter Calibrations Using a Small Volume Prover

A Report for

National Measurement System Directorate

Department of Trade & Industry 151 Buckingham Palace Road

London, SW1W 9SS

Project No: OSDC53 Report No: 014/97 Date: October 1997

The work described in this report was carried out under contract to the Department of Trade & Industry (‘the Department’) as part of the National Measurement System’s 1994-1999 Flow Programme. The Department has a free licence to copy, circulate and use the contents of this report within any United Kingdom Government Department, and to issue or copy the contents of the report to a supplier or potential supplier to the United Kingdom Government for a contract for the services of the Crown. For all other use, the prior written consent of TÜV NEL Ltd shall be obtained before reproducing all or any part of this report. Applications for permission to publish should be made to: Contracts Manager TÜV NEL Ltd Scottish Enterprise Technology Park East Kilbride G75 0QU E-mail: [email protected] Tel: +44 (0) 1355-272096 © TÜV NEL Ltd 2003

National Engineering Laboratory

Project No: OSDC53 Report No: 014/97 Page 1 of 68

Flow Centre

Flowmeter Calibrations Using a Small Volume Prover

A Report for

National Measurement Systems Policy Unit

DTI, London S U M M A R Y This report describes the findings of a project undertaken by NEL for NMSPU, Department of Trade and Industry to investigate and quantify the performance of flowmeters calibrated under simulated ‘on-site’ conditions using a Small Volume Prover (SVP). The meters calibrated and tests undertaken were chosen from the results of a survey sent to users of SVPs and site calibration services. This report discusses the testing of a turbine, Coriolis mass, positive displacement and ultrasonic meters using a Brooks 18-inch SVP as the reference calibration instrument. All meters were subjected to an initial calibration using the NEL flow measurement reference gravimetric test facility and SVP simultaneously. This was followed by each meter undergoing repeatability tests entailing a number of SVP piston passes, SVP plenum pressure, one pulse per revolution of meters with rotating parts and test meter pulse output during SVP piston traverse between optical measurement switches. Overall conclusions are given on the performance of each meter type. Particular interest is given to the intercomparison of each meter type, pulse output mechanism and subsequent performance. A review of pulse interpolation methods has also been completed. Prepared by: Mr S Nicholson .............................................. Approved by: Mr R Paton .............................................. Date: 24 October 1997 for W Paton Director and General Manager



C O N T E N T S Page SUMMARY ........................................................................................... 1 1 INTRODUCTION .................................................................................. 4 2 INDUSTRY SURVEY ........................................................................... 5 3 SELECTION OF TEST METERS ......................................................... 5 4 TEST METER INSTALLATION ............................................................ 6 5 TEST PROGRAMME ........................................................................... 6 6 TEST RESULTS ................................................................................... 6 6.1 Brooks 3-inch Nominal Bore Parity Turbine Meter ............................... 7 6.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter ................................... 7 6.3 Smith 3-inch Nominal Bore Positive Displacement Meter .................... 8 6.4 Fisher-Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter . 8 6.5 Conclusions .......................................................................................... 9 7 NUMBER OF SVP PISTON PASSES VERSUS TEST METER REPEATABILITY .................................................................................. 9 7.1 Brooks 3-inch Nominal Bore Parity Turbine Meter ............................... 10 7.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter ................................... 10 7.3 Smith 3-inch Nominal Bore Positive Displacement Meter .................... 10 7.4 Fisher-Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter 11 7.5 Conclusions .......................................................................................... 11 8 PLENUM PRESSURE VERSUS TEST METER REPEATABILITY ..... 12 8.1 Brooks 3-inch Nominal Bore Parity Turbine Meter ............................... 13 8.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter ................................... 13 8.3 Smith 3-inch Nominal Bore Positive Displacement Meter .................... 13 8.4 Fisher-Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter . 14 8.5 Conclusions .......................................................................................... 14 9 PULSE OUTPUT FROM TEST METERS DURING SVP PISTON PASSES BETWEEN OPTICAL MEASUREMENT SWITCHES ........... 15 9.1 Brooks 3-inch Nominal Bore Parity Turbine Meter ............................... 15 9.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter ................................... 15 9.3 Smith 3-inch Nominal Bore Positive Displacement Meter .................... 16 9.4 Fisher-Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter . 16 9.5 Conclusions .......................................................................................... 16



C O N T E N T S (contd) Page 10 ONE PULSE PER REVOLUTION TESTS ............................................ 16 10.1 Conclusions .......................................................................................... 17 11 SUMMARY CONCLUSIONS ................................................................ 18 12 REVIEW OF PULSE INTERPOLATION METHODS ............................ 19 BIBLIOGRAPHY ................................................................................... 30 LIST OF TABLES ................................................................................. 31 LIST OF FIGURES ............................................................................... 32



1 INTRODUCTION Flowmeters are widely used in the downstream production side of the oil industry. This inevitably leads to ‘on-site’ calibration of meters. This enables meters to be tested in the working product at typical operating conditions and often without closing down a production stream. The latter is of obvious benefit to the meter operator. With the recent introduction of mass meters into fiscal fluid streams, their obvious advantages over existing volume meters and the future possibility of ultrasonic meters being introduced, it was considered valuable to test a cross-section of these meters using a Small Volume Prover (SVP). The equipment utilised during this project was a Brooks 18-inch SVP with standard Brooks compact prover electronics console (BCPE). Figure 1 shows a sketch of the Brooks Small Volume Prover. The key components are the prover cylinder and piston, poppet valve and the optical detectors. When the SVP is in standby (idle) mode or not being operated, the poppet valve is held open by hydraulic fluid and test fluid allowed to flow freely through the piston assembly. When a prover pass is initiated the poppet valve is closed pneumatically, using nitrogen from the plenum chamber, and seals against the face of the piston creating a solid piston surface. The flowing test fluid pushes the piston downstream through the cylinder sweeping out a volume of test fluid. A metal, invar, rod that has an optical ‘flag’ mounted on it is attached to the piston. As the piston sweeps through the cylinder volume the ‘flag’ passes through optical measurement switches. When the flag triggers the first measurement switch, pulses from the test meter are accumulated by a pulse counter. When the ‘flag’ passes through the second measurement switch the pulse counter is triggered to stop accumulating pulses. Finally the poppet valve is opened and hydraulic fluid is pumped into the actuator cylinder to move the measurement piston back to the initial standby (idle) position. This is one pass of the prover piston. The precise volume between optical switches is determined by calibration against a volumetric test measure or gravimetric weighing system. To provide acceptable accuracy the SVP uses a measurement technique known as double chronometry pulse interpolation. This technique will be explained in detail in Section 12, Pulse Interpolation Guidance Note, of this report but basically utilises two counters. One to measure the time between triggering of the measurement switches and the other to measure the time between the leading edges of the flowmeter measurement pulses. The ratio of the two times is used to determine the fractional flow measurement pulses that occur between the prover measurement switches. This method of pulse interpolation provides better pulse resolution which permits small volumes to be used for flowmeter calibrations. SVPs are generally the chosen ‘tool’ used for on-site calibration of flowmeters. This generally means turbine volume meters as they are the most popular meter used on high accuracy fluid streams. The introduction of new metering technology has lead to the availability of additional information from meter electronics, ie derived meter pulse output, temperature and density. Meter types using this technology, such as Coriolis mass and ultrasonic, have generally been considered unsuitable for use with SVPs due to their method of pulse signal output. This output is essential to the operation of the SVP electronics and final derived test meter K-factor. The pulse output from these types of meter are generated by software within each meter’s flow transmitter. This can lead to the pulse output lagging behind real time, output of pulse batches at different frequencies or the introduction of corrected pulses.



This may cause pulse count time-frame errors for the BCPE within the time taken for the SVP’s piston to displace its calibrated volume. This obviously leads to an inaccurate calibration of the test meter. Until now there have been too few comprehensive tests utilising these meters to discover how they perform under various test conditions simulating a variety of typical problems encountered by SVP and meter users. 2 INDUSTRY SURVEY An initial industrial survey was undertaken to discover what problems have been encountered by British industry’s SVP operators and end users of small volume provers and their associated meters. To this end SVP operators were contacted by letter and fax with a survey form requesting details of meters calibrated using their SVP and what problems they had encountered while using each meter type, as well as problems they have encountered while using the prover mechanism itself. Seventy-five per cent of the survey questionnaires were returned. Out of the returned surveys two companies refused to supply information based on NEL’s possible entry into the on-site calibrations market in direct competition with themselves and the other on the grounds that they were the manufacturer of the most popular SVP and felt they could not give objective comments on operation of their product or divulge operating problems to NEL. The main aspects requiring investigation, based on survey replies, were: Number of passes per run required for accurate, repeatable proving. Effect of plenum pressure on test meter repeatability. Calibration of mass flowmeters (pulse collection problems). Calibration of ultrasonic flowmeters (pulse collection problems). Pulse output during piston operation between SVP optical switches. 3 SELECTION OF TEST METERS To enable the test criteria, determined by the survey conclusions, to be fulfilled, four types of flowmeter were procured. The meter manufacturers were chosen for their popularity in field use hence replicating the problems commonly found by those surveyed. All test meters were provided with pulse outputs. This enabled direct input to the SVP flow computer (BCPE) and therefore volumetric comparison could be made. The test meters chosen were: Brooks 3-inch Nominal Bore Parity Turbine Meter. Danfoss 4-inch Nominal Bore Sonoflo 3000 Ultrasonic Meter. Smith 3-inch Nominal Bore Positive Displacement Meter. Fisher Rosemount (Micromotion) 3-inch Nominal Bore CMF300 Coriolis Mass Meter.



4 TEST METER INSTALLATION Each test meter was installed downstream of the SVP as detailed in Figure 2 with the manufacturer’s recommended installation requirements followed. The Brooks SVP flow computer (BCPE) requires a pulse input to enable prover piston passes to be initiated. The use of the Brooks 3-inch turbine flowmeter, as shown in Figure 2, throughout the test series ensured operation of the SVP during tests not requiring direct pulse counting comparison between test meter and SVP, ie test meter pulse output between SVP optical measurement switches. All tests were carried out with the test fluid temperature of nominally 20ºC and test fluid line pressure of nominally 2.5 bar. The SVP plenum pressure was set, according to Brooks instructions depending on the exact line pressure of individual tests. The exception was the plenum pressure tests where the nitrogen plenum pressure was obviously altered to determine its effect on test meter performance. 5 TEST PROGRAMME The SVP used during the test series was initially water drawn to determine its calibration volume. This was completed using volumetric and gravimetric methods simultaneously and in accordance with procedures described in the Institute of Petroleum, Petroleum Measurement Manual, Part X Meter Proving, Section 3 - Code of Practice for Design, Installation and Calibration of Pipe Provers. All SVP calculated K-factors quoted in this report have been subject to corrections for fluid temperature and pressure, material temperature and pressure and invar and SVP tube material constant as described in the Institute of Petroleum, Petroleum Measurement Manual, Part X Meter Proving, Section 3 - Code of Practice for Design, Installation and Calibration of Pipe Provers. The test programme was split into five sections: • Initial calibration of the test meter using the SVP and NEL’s National Standard

single phase flow measurement facility. • Effect of SVP piston passes on test meter repeatability. • Effect of SVP plenum pressure on the test meter repeatability. • Pulse output of test meters during SVP piston passes between optical switches. • One pulse per revolution of turbine meters. 6 TEST RESULTS Initial Calibration Against SVP All test meters were initially calibrated against the SVP and NEL’s National Standard six tonne gravimetric test facility simultaneously. A reference calibration for each meter was thus obtained. Using the gravimetric standard two pulse collection methods are employed. The ‘standard’ method utilises a pulse counter gated from the weightank inlet valve. Pulses are counted from the start of flow until the flow is stopped coinciding with the closure of the weightank inlet valve. The ‘extended’ method used a pulse counter where the gate is opened prior to the weightank inlet valve opening and closed a predetermined time after the flow stops, ensuring any delayed pulses are collected.



6.1 Brooks 3-inch Nominal Bore Turbine Meter This was a ten blade meter supplied with signal amplifier and installed with twenty diameters upstream and ten diameters downstream of straight 3-inch nominal bore pipework. This type of meter is widely used by industry, in conjunction with Small Volume Provers, for simultaneous calibration of flowmeters and pipe provers at site locations. The meter was calibrated over the flowrange 5 to 35 l/s with the results shown in Table 1 and Figure 3. Each test point consisted of one run of the SVP, each run entailed ten passes of the SVP piston. Repeatability test points were completed at 10 and 30 l/s. The linearity of both SVP and NEL gravimetric calibrations, over this flowrange, was nominally 0.05% with repeatability of each set of results typically within 0.01% except at the lower flowrates where this figure increased to 0.06%. These results are typical of a turbine meter calibration. The ‘standard’ pulse counting method was used. The comparison between SVP and Gravimetric system results is generally within 0.015%. This falls within the expected comparison figures between the two systems. 6.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter This meter was a two path meter supplied with a Danfoss Sonoflo 3000 flow transmitter mounted on the meter casing. The meter was zeroed under operating conditions, as recommended by manufacturer, immediately prior to the first test point. Installation of the meter was completed according to manufacturer’s recommendations. The meter was calibrated over a 10 - 35 l/s flowrange with the results shown in Table 2 and Figure 4. Based on the comments received from our survey it was anticipated that problems may be witnessed during the calibration of this meter against the SVP. Industry opinion stated that they were not confident that the pulse output from this type of meter would be suitable for calibration against a SVP. To enable this to be quantified the meter was calibrated using the ‘standard’ and ‘extended’ pulse counting methods. The individual calibration results from the meter using the SVP and NEL gravimetric system, using both gating methods, are shown in Figure 4. The calibration curve from the NEL gravimetric system is typical of this type of meter, with meter K-factor increasing by nominally 2% as the flowrate decreases over the calibration range. These results can be compared with the SVP calibration curve which shows a systematic and almost parallel shift of 0.35% from the gravimetric system calibration. The test meter’s K-factor repeatability from the SVP and gravimetric system results is 0.2%. This is within the manufacturer’s specification. The systematic shift between test facility calibration curves can be explained by the possibility of the NEL gravimetric system missing pulses from the test meter. The use of both pulse counting systems enabled this to be quantified. Operator suspicions were confirmed when the use of the ‘extended’ pulse gating system demonstrated that the test meter was outputting pulses up to four seconds after the weightank inlet valve had closed.



Figure 4 demonstrates that data collected from the NEL ‘extended’ pulse gated system corresponds to the data simultaneously taken by the SVP. This shows that this type of meter is not suitable for calibration using standing start stop calibration systems utilising pulse gating directly from valve actuation. 6.3 Smith 3-inch Nominal Bore Positive Displacement Meter This meter is a sliding vane meter with a pulse signal mechanism attached to an extended gear shaft. This signal was amplified by an onboard amplifier prior to output to SVP and NEL gravimetric data acquisition systems. The meter was calibrated over the flowrange 3 - 30 l/s. The pulse counter used by the NEL data acquisition system was used in conjunction with a manually operated gating signal. This allowed the pulse counter gate to be opened prior to the NEL weighbridge inlet valve opening and to remain open following the inlet valve closure. This allowed all of the test meter’s generated pulses to be collected and alleviate the problem previously encountered with the Danfoss ultrasonic meter. The results are shown in Table 3 and Figure 5. The NEL gravimetric system results show K-factor repeatability figures of less than 0.02% at nominal flowrates of 7 l/s and 23 l/s. The K-factor increased by 0.1% between 30 l/s and 7.5 l/s before decreasing to a K-factor similar to 30 l/s at very low flowrates. This calibration curve is typical of this type of meter and is repeated on the SVP calibration curve although a drop in K-factor of nominally 0.03% is witnessed across the flowrange. This is within the expected accuracy of this type of meter. 6.4 Fisher Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter This meter is a ‘U tube’ design and operates using a 9739 remote flow transmitter (RFT). The pulse output signal from the RFT was used to calibrate this meter against the SVP and NEL gravimetric weighing systems. During initial set up of this meter the pulse counter was operated on ‘extended’ gating, where the test meter’s pulse output was witnessed to continue outputting pulse counts ten seconds after the weightank inlet valve had closed. This method of pulse gating was used for the remainder of this meter’s tests. The results of the simultaneous calibration of this meter using the SVP and gravimetric weighing system are shown in Table 4 and Figure 6. The calibration produced from the gravimetric system shows a repeatability and linearity of 0.04%. The SVP calibration data shows similar results at above 25 l/s but as the flowrate falls the spread of SVP K-factor results gradually increases to 0.7% at 5 l/s. This may be caused by delayed pulse output from test meter electronics. This will be investigated later in this project.



6.5 Conclusions Using the SVP as a calibration facility while simultaneously calibrating each test meter against the National Standard enabled data from both sources to be compared. The calibration of turbine meters using SVPs is well documented with the Brooks meter performing as expected with repeatability of better than 0.02% and comparison to the National Standard facility within 0.01%. Points taken at 10 l/s which had slightly higher repeatability may be due to a flowrate which equates to the typical transition period, where the turbine meter’s K-factor peaks before falling at the bottom end of the meter’s flowrange. By using the ‘extended’ pulse gating system all Danfoss ultrasonic meter pulses were collected by the National Standard facility data acquisition system. The comparison between the SVP calibration and the automatic gating system, taken from the National Standard facility weightank inlet valve operation, showed a parallel shift in calibration curve. When the manual gating system was utilised the calibration data overlapped the corresponding SVP derived K-factor. This information will be vital to future ultrasonic meter calibration work. The calibration data produced by this meter was typical of a meter of this type (see DTI project number OSDS37, Report No 35/96). The Smith positive displacement meter, when compared with the National Standard facility, showed a parallel K-factor shift of nominally 0.03%. This is within the expected uncertainty of both gravimetric and SVP calibration methods. The SVP repeatability data points showed a K-factor spread of 0.007% compared with the gravimetric facility spread of 0.03%. The Micromotion Coriolis mass meter produced a typical calibration when using the National Standard facility. The SVP data shows poor repeatability and a greater spread of K-factor results over the calibration flowrange. The pulse output from the mass meter required to be used on the National Standard manual pulse gating system due to the ‘run-on’ of pulse output after the weightank inlet valve was operated. 7 NUMBER OF SVP PISTON PASSES VERSUS TEST METER REPEATABILITY From the initial survey it was discovered that SVP users and operators had shown concern regarding the effect of the number of SVP piston passes on the repeatability of flowmeters. For site calibrations it is generally recommended that between five and ten passes of the SVP piston are carried out per run. This allows for an effective mathematical standard deviation to be computed on each set of results. Obviously the number of passes has a primary effect on the overall test time. It is therefore beneficial for both SVP operator and client to perform the correct number of passes, to enable accurate calculation of meter performance, within the shortest time. Short time scales also prevent the need for large fluid volumes, should the test meter require calibration in product. To enable this problem to be quantified each meter was tested against the SVP at a nominal flowrate of 25 l/s. This flowrate is representative of where this type and size of meter would be used in operation.



One, five, ten and twenty pass per run samples were completed with each increasing pass number running consecutively with the previous sample. This ensured that the test conditions remained constant throughout each meter’s test period. Six test points were recorded at each of one, five, ten and twenty passes per run with a SVP K-factor calculation being completed for each run. 7.1 Brooks 3-inch Nominal Bore Turbine Meter The test results from this meter are shown in Table 5 and Figure 7. A nominal flowrate of 28 l/s was set and the initial six test points, each point entailing one pass, were completed. This produced a 0.004% spread in K-factor results over the six test points. Test data at five passes per run provided a K-factor spread of 0.006%. This figure increased to 0.008% during the ten passes per run and remained at this level during the twenty passes per run test. This type of meter is widely used in conjunction with SVPs for site calibrations and hence the results provided are of a type generally expected. The criteria for acceptance of a turbine meter for fiscal use is five consecutive runs obtaining an overall spread in K-factor results of 0.02%. The results from the one pass per run tests demonstrate that this number of runs is statistically unsatisfactory. There is not enough data to provide a useable statistical analysis of the meter’s performance. A K-factor spread of 0.006% was obtained from the five passes per run test with the ten and twenty pass per run providing a K-factor spread of 0.008%. This demonstrates that, with a SVP operated at five passes per run, this meter could provide a satisfactory repeatability but statistically the use of ten passes per run is more satisfactory even with the slight increase in K-factor spread. Twenty passes per run would increase the test time to unacceptable levels with no great advantage over ten passes per run. 7.2 Danfoss 3-inch Nominal Bore Ultrasonic Meter A nominal flowrate of 29 l/s was set and the initial six test points at one pass per run were completed. This produced a 0.7% spread in K-factor results over the six test points. Test data at five passes per run provided a K-factor spread of 0.33%. This figure decreased to 0.24% during the ten passes per run test with a further reduction to 0.18% during the twenty passes per run test. The test results from this meter are shown in Table 6 and Figure 8. 7.3 Smith 3-inch Positive Nominal Bore Displacement Meter This test was completed at a nominal flowrate of 27 l/s and the results are shown in Table 7 and Figure 9. At one pass per test point a K-factor spread of 0.009% was obtained. This decreased to 0.007% for five passes per run, increased to 0.01% for ten passes per run with a final spread of 0.007% during the twenty passes per run tests.



Positive displacement meters are regularly used for service where repeatability is of prime importance. The results of this test demonstrate justification for this. 7.4 Fisher-Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter This test series was carried out at a nominal flowrate of 23 l/s. The initial six test points at one pass per run were completed. This produced a 0.12% spread in K-factor results over the six test points. Test data at five passes per run provided a K-factor spread of 0.04%. This figure increased to 0.05% during the ten passes per run test but reduced to 0.03% during the twenty passes per run test. The test results from this meter are shown in Table 8 and Figure 10. This indicates that this meter’s performance is unaffected by the number of passes provided that there is ample number to provide a statistical analysis, with five being a suggested minimum. 7.5 Conclusions The Brooks turbine meter repeatability was smallest at 1 pass per run with the repeatability figure getting gradually larger as the number of passes per run increased. The figure obtained from the twenty passes per run is half the fiscal metering acceptance criteria. Operating at ten passes per run would produce an acceptable repeatability figure and allow optimum time usage of the facility. At ten passes per run the Danfoss ultrasonic meter results concur with the initial repeatability figure of 0.24% at a similar flowrate. The repeatability value decreases from 0.7% to 0.18% over the 1 run per pass to twenty runs per pass suggesting that this meter would benefit from the increased number of runs. When comparing ten passes per run, at 0.24%, to twenty passes, at 0.17%, the reduction is considerable enough and could merit the increased time involved in completing the entire prover operation with a larger number of passes. It should be stressed that the repeatability figures obtained during this test suggest that this type of meter would be, at present, unsuitable for fiscal use. Positive displacement meters generally have good repeatability but experience has previously shown that the introduction of gear and linkage mechanisms can cause poor repeatability when this type of meter is calibrated using an SVP. The Smith meter used in this project did not demonstrate this problem. K-factor repeatability figures varied from 0.009%, at 1 pass per run, to 0.007% at twenty passes per run. A K-factor spread of 0.01% was achieved at ten passes per run but this was mainly due to one point, from six, producing a larger K-factor than the other five and hence increasing the overall spread. Without this point the K-factor spread would have been similar to the other pass per run tests. The Micromotion Coriolis mass meter showed improved repeatability when compared with the data produced from the initial calibration. K-factor repeatability figures were reduced to nominally 0.04% during the five, ten and twenty pass per run tests with the 1 pass per run test being 0.12%. The Smith PD meter produced the best repeatability figures of the meters tested, closely followed by the Brooks turbine both of which could be utilised for fiscal use. The repeatability of the ultrasonic and mass meters was above the 0.02% fiscal acceptance criteria, although the mass meter may achieve this level at higher flowrates. This may be due to the method of pulse output from these meters.



8 PLENUM PRESSURE VERSUS TEST METER REPEATABILITY Survey results showed SVP user concern regarding the effect of SVP plenum pressure on test meter repeatability. It was suspected that during site tests SVP operators may not set the SVP plenum pressure at the manufacturer’s recommended pressure, which is dependent on test line pressure. If plenum pressures were set at levels many times greater than the test line pressure the SVP piston could be launched past the first measurement optical switch at a greater velocity than the test fluid would normally allow. This could cause a pressure surge effect on the meter performance. If the plenum pressure flowrate was set too low the SVP poppet valve may not be fully closed at the time of passing the first optical measurement switch. Prior to operation the manufacturer’s recommended plenum pressure was based on a formula incorporating the test fluid operating pressure. The operation of the plenum was previously described in the report introduction. The plenum pressure is used to ensure that the piston poppet valve is closed prior to the piston passing the initial measuring optical switch with the test fluid pushing the piston through the calibrated volume in the SVP. Brooks Plenum Pressure formula:

Plenum pressureprover inlet pressure

psi= +50

60.

( )

This equation was used throughout the test program with the exception of intentional resetting of the plenum pressure during this test phase. This test phase involved two individual tests. Each test concentrated on one aspect of the plenum pressure’s possible effect on the test meter performance. Initially each test meter was tested using the SVP as the reference instrument at a fluid inlet pressure of 2.5 bar and the plenum pressure set according to the above equation. Six test points were completed entailing ten passes per run. The plenum pressure was then adjusted to 7.5 bar, 10 bar and 2.5 bar with the process being repeated at each plenum pressure. Following each plenum pressure test the SVP was programmed to complete a number of piston passes. A National Instruments “Labview” software program was run simultaneously to establish the test meter output frequency versus time over the three passes. This was repeated at each plenum pressure and also with the SVP piston at idle and switched off. This enabled a graph to be plotted for each test meter detailing the effect of plenum pressure at all operating conditions on meter frequency output.



8.1 Brooks 3-inch Nominal Bore Turbine Meter The initial test results using the SVP are shown in Table 9 and Figure 11. The spread of K-factor results shown in Figure 11 indicates that this meter is not affected by plenum pressure below a value of 10 bar. Here the spread in K-factor increases from 0.004% at 5 bar to 0.007% at 10 bar. This effect could be reflected in the possible implications for the meter’s overall repeatability should the plenum pressure be set at a much higher level than the manufacturer’s equation suggests. If this situation occurred the meter’s ability to achieve the repeatability criteria for fiscal measurement could be impaired. The Labview analysis, Figure 12, shows that the frequency output remains constant, for all plenum pressure settings, during the SVP piston travel irrespective of prover piston passes. 8.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter The SVP reference test details are shown in Table 10 and Figure 13. These detail the effect of plenum pressure on the repeatability of the Danfoss meter K-factor results. Figure 13 shows the spread of K-factor results increasing from 0.15% to 0.21% as the plenum pressure increased from 2.5 bar to 5 bar. The spread of K-factor results at 7.5 bar plenum pressure was 0.35%, although this was exaggerated due to one point producing a K-factor 0.2% lower than the other five at this plenum pressure. The K-factor spread at 10 bar returned to 0.15%. The meter’s performance showed a slight plenum pressure effect especially between too low a plenum pressure and the manufacturer’s recommended setting. Figure 14 shows the Labview analysis of the meter’s frequency output versus time when the prover was operating over three passes and when the prover was in the idle position and switched off. The meter frequency output when the SVP is off and at the idle position, SVP on and piston ready to launch, show similar results. When the plenum pressure was set to 2.5 bar the frequency reduced by 3.6% and produced an increase of 2% during the piston’s travel between launch and return strokes. This result was similar to a plenum pressure setting of 7.5 bar. This represents a 2.5 bar increase to the manufacturer’s recommended plenum pressure setting, although there was an overall rise of 3% during the piston travel between measurement switches at this plenum pressure. At plenum pressures of 7.5 bar and 10 bar the results return to the same frequency as the initial piston idle frequency but produce a further increase of 2.4% during the piston travel between launch and return. 8.3 Smith 3-inch Nominal Bore Positive Displacement Meter The initial test results using the SVP as the reference are shown in Table 11. A nominal flowrate of 256 l/s was selected and six points were completed each entailing ten passes per point. Figure 15 shows these results in graphical form. The 2 bar plenum pressure results produced a spread in K-factor results of 0.01%. This reduced to 0.006% during the 5 bar plenum pressure results. The 7.5 bar and 10 bar plenum pressure results produced K-factor spreads of 0.01% and 0.007% respectively.



The results from this test show that this meter’s repeatability was slightly affected by the plenum pressure. The manufacturer’s recommended plenum pressure of 5 bar produced the smallest K-factor spread although the difference in K-factor spread throughout the plenum pressure tests was nominal. The data taken from the Labview software is shown in Figure 16. The test meter frequency output while the piston was at the idle position remained within 2% throughout this test series. The frequency data produced from the test meter during the SVP on, but piston at idle position, show that the meter frequency output increased by 0.5% from SVP off conditions. The meter’s frequency output during the 2.5 bar plenum pressure test remained at the same level as the piston idle frequency output although a noticeable rise of 3% was witnessed during the SVP piston’s launch and return phases. When the plenum pressure was increased to 5 bar the frequency output increased by 6%, during the period between piston launch and return, compared with the piston at the idle condition. The data taken from the 7.5 bar and 10 bar plenum pressures showed further rises of 7% and 9% respectively during the piston travel. 8.4. Fisher Rosemount 3-inch Nominal Bore CMF300 Coriolis Mass Meter This meter’s results are shown in Table 12 and Figure 17. The results from the 2 bar plenum pressure tests show a spread in K-factor results of 0.05%. This figure was repeated during the 5 bar plenum pressure results. The 7.5 bar plenum pressure results indicate an increase in K-factor spread to 0.09% before the K-factor spread returns to 0.05% at 10 bar plenum pressure. The data from the Labview software, Figure 18, shows the meter frequency output when the SVP was off and when the SVP was at the piston idle position to be comparable. When the piston was travelling between launch and return strokes, at 2.5 bar plenum pressure, the meter frequency output showed a rise of 2%. The rise in frequency at 5 bar plenum pressure was 3% although the time periods outwith launch and return strokes returned to the same frequency levels as previous tests in this series. The rise in frequency output during piston traverses between launch and return during the 7.5 and 10 bar tests was 5% although there was a systematic shift of 1.25% between the plenum pressure tests. 8.5 Conclusions The effect of plenum pressure on the repeatability of turbine meters was one of the main areas that our survey required NEL to investigate. The performance of the Brooks turbine meter indicates that the effect of plenum pressure would only be detectable, but not significant, if it was set at a much higher level than the manufacturer specifies. NEL’s data shows a plenum pressure of 10 bar, double the manufacturer’s recommendation, producing K-factor repeatability of 0.007% with a fiscal acceptance criteria of 0.02%. Fluctuations in the frequency output of the test meter throughout this test series were negligible. The effect of plenum pressure on the Danfoss ultrasonic meter showed improved repeatability at the higher, 10 bar, plenum pressure. The test meter frequency output analysis showed a definite frequency rise during the SVP piston travel time although this was not pressure related as the increase in frequency was constant throughout the plenum pressure tests.



K-factor figures from the Smith positive displacement meter suggest that the meter’s repeatability was not affected by plenum pressure settings. The frequency output did show continuous increase dependent on plenum pressure although this obviously had no effect on the meter’s performance. The Micromotion mass meter shows no significant effect of plenum pressure on K-factor repeatability with only a slight increase during the 7.5 bar plenum setting. The frequency output data does show that the test meter output during the SVP piston travel increased to higher levels dependant on the plenum pressure. The increase in frequency at each piston travel remains constant for each plenum pressure setting and therefore has no effect on individual pass repeatability. 9 PULSE OUTPUT FROM TEST METERS DURING SVP PISTON PASSES BETWEEN OPTICAL MEASUREMENT SWITCHES This test series utilised National Instruments “Labview” software to quantify each test meter’s pulse output frequency during four passes of the SVP piston. The voltage signal from the optical measurement switch, normally used by the BCPE pulse counter, was used to trigger a Labview monitor window. Another Labview window collected the pulse output from the test meter. Both monitoring windows operated simultaneously and on real-time mode. This allowed the SVP piston traverse to be monitored over the four passes simultaneously and the first measurement switch voltage output to be superimposed. Each pass of the SVP piston is indicated by the voltage output from the first measurement switch, both on the proving and return strokes. The actuation of the first measurement switch, and therefore SVP counters, is indicated by a voltage increase on the piston proving/return voltage line. All test data was carried out with a line pressure of 2.5 bar, plenum pressure of 5 bar and a nominal test fluid temperature of 20ºC. 9.1 Brooks 3-inch Nominal Bore Turbine Meter The results from this meter’s test is shown in Figure 19. This meters frequency output does not alter significantly during the SVP piston cycle. Although there is some fluctuation in the meter’s frequency output over the indicated test time, this is similar to that experienced during the plenum pressure test series. 9.2 Danfoss 4-inch Nominal Bore Ultrasonic Meter The Danfoss ultrasonic meter results are shown in Figure 20. This demonstrates the rise in meter output frequency coinciding with the launch of the SVP piston. The meter reached its maximum frequency output before the first measurement optic was reached, in three of the four passes, with the frequency reducing as the piston reaches the end of the calibrated volume as the effect of plenum pressure reduces. As the SVP piston returns, past the first optic to the idle position, the frequency output sharply reduced to a frequency similar to that indicated at piston idle.



9.3 Smith 3-inch Nominal Bore Positive Displacement Meter Figure 21 shows the results from this meter’s tests. The frequency output rate remains at the same level throughout the SVP piston’s travel. 9.4 Fisher Rosemount 3-inch Nominal Bore CMF 300 3-inch Coriolis Mass Meter The results are shown in Figure 22. This shows the rise in meter output frequency coinciding with the launch of the SVP piston. This meter did not reach its maximum frequency output until after the piston passed the first optical switch. This was demonstrated on all four passes. This shows the possible pulse output time lag from the meter’s software. The pulse output remains constant once the maximum output is achieved. As the SVP piston returns, past the first optic to the idle position, the frequency output sharply reduced to a value similar to that indicated at the piston idle position. 9.5 Conclusions The Brooks turbine and Smith positive displacement meter results show that both meter frequency outputs remained constant throughout the SVP piston proving cycle. This factor reflects the initial calibration of these types of meter. The Danfoss ultrasonic meter’s frequency output gradually increases following the SVP piston launch. The frequency then peaked before the piston reached the first optical measurement switch and remained at this level throughout the piston traverse of the prover’s calibrated volume. The meter’s frequency output, during the SVP piston traverse time, shows an expected gradual decline due to the plenum pressure having less effect as the piston reaches the end of its travel. The Micromotion frequency output shows a time difference, of nominally 1.5 seconds, between reaching its maximum pulse output and the SVP piston passing the first measurement optic. This demonstrates a pulse output time delay from this meter’s software and may be the reason for poor test results during the initial calibration of this meter. 10 ONE PULSE PER REVOLUTION TESTS The criteria for calculating the minimum number of pulses collected during SVP calibrations is described in the Section 12, Review of Pulse Interpolation Methods, of this report. It states that theoretical analysis by modelling has shown that the K-factor error introduced by pulse variations reduces in proportion to the number of raw pulses, in accordance with a mathematical square law. To enable this theory to be analysed, one pulse per revolution tests were carried out on two turbine meters. One with a proven repeatability and assumed intra-rotational non- linearity, the other with intentionally damaged rotor blades to provide poor intra-rotational non-linearity. Reducing the number of raw pulses per revolution without altering the number of meter revolutions would demonstrate that one pulse per revolution, from the meter with proven repeatability, should not be altered by the reduction of pulses per revolution. The meter with poor intra-rotational non-linearity would provide a greater spread in K-factor repeatability when comparing reduced pulse count per revolution to full blade pulse counting methods.



The Brooks 3-inch NB turbine meter was fitted with a ten blade rotor. During one piston pass between optical detectors the BCPE would record ten pulses per revolution from this meter’s pulse output. To enable the effect of one pulse per revolution to be quantified a pulse dividing mechanism was introduced. This was placed between the pulse output, after amplification by the meter, and the BCPE’s input. This allowed for real time pulse division. Three tests were completed at 10 l/s, 20 l/s and 30 l/s. At each flowrate the ten pulses per revolution test was carried out immediately before the one pulse per revolution test. The tests were completed consecutively at each flowrate therefore ensuring the test conditions were constant during each comparison. Each point consisted of one SVP piston run of ten passes with the average K-factor standard deviation, at each flowrate, expressed as a percentage of the average flowrate K-factor. The Brooks 3-inch NB turbine meter results are shown on Table 13. The flowrate averaged results from the ten pulses per revolution tests produced percentage standard deviations of 0.004%, 0.003% and 0.009% from 10 l/s, 20 l/s and 30 l/s respectively. The respective figures for the same flowrates during the one pulse per revolution tests were 0.004%, 0.003% and 0.005%. This shows the spread of K-factor results from the one pulse per revolution and 10 pulses per revolution tests to be of similar magnitude. The same test was completed using a Brooks 4-inch NB, twelve blade meter which had its rotor blades altered. To ensure a poor intra-rotational non-linearity each alternate pair of blades were pushed together ensuring that the pulsed output from the meter during the twelve pulses per revolution tests was at an unequal frequency. The pulse divider was then set to divide by twelve to repeat the previous test. The results are shown on Table 14. The flowrate-averaged results from the twelve pulses per revolution tests produced percentage standard deviation of 0.03%, 0.02% and 0.01% from 10 l/s, 20 l/s and 30 l/s respectively. The respective percentage figures for the same flowrates during the one pulse per revolution were 0.04%, 0.01% and 0.02%. This shows the spread of results from the one pulse per revolution tests to be similar to those produced during the twelve pulses per revolution. 10.1 Conclusions The results obtained from the 3-inch turbine meter show that reducing the number of pulse counts per revolution, on a meter with good intra-rotational linearity, does not affect the repeatability of the meter. The 4-inch turbine meter, with blades altered to simulate poor intra-rotational non-linearity should have shown an increased K-factor spread during the reduced pulse per revolution test. This was only evident at two of the three flowrates. Evaluation of the true time period between each pulse would have to be completed to enable the pulse reduction theory to be proved.



11 SUMMARY CONCLUSIONS Of the four meters tested the Brooks turbine and Smith positive displacement meter produced better repeatability throughout the test series. The initial calibration comparison using the PD meter utilising the SVP and NEL National Standard facility was only marginally less accurate, over the flowrange, than the turbine meter. Both meters performed within their manufacturer’s tolerances during plenum pressure and SVP piston passes per run tests. The turbine meter showed that reducing the number of passes per run from ten to five did not produce a significant difference in the meter’s K-factor repeatability. One pulse per revolution tests on turbine meters showed that K-factor repeatability was impaired when rotor blades were damaged, simulating poor intra-rotational non-linearity, but results from reduced pulses per revolution were inconclusive. This may be improved by further research into the true time between raw pulses over the flowmeters’ range. Research into the effect of one pulse per revolution from PD meters would also be advantageous as they are widely used in bulk loading operations. Once initial pulse counting problems were solved the Danfoss ultrasonic meter performed well throughout the test series. This type of meter is becoming more popular within the oil industry and may supersede the turbine meter in some applications. The pulse output from the Coriolis mass meter appears to be the cause of poor calibration and repeatability from this type of meter when calibrated using an SVP. Pulses appear to be output on a time delay, ie the meter pulse output lags behind real time. Present SVP electronics do not have the capability to cope with this type of pulse output. Alterations to the method of pulse output from this type of meter or to the BCPE package on the SVP may be required before this problem is overcome. With the exception of the turbine meter the remaining test meters demonstrated a rise in output frequency during the SVP piston travel. This rise was generally proportional to the plenum pressure. Increasing the number of SVP piston passes per run generally increased the repeatability of test meter K-factor results particularly when using the Danfoss ultrasonic meter. Alteration to the present operation method, to account for the possible improvement in repeatability, would have to be considered in conjunction with operating conditions at the time of meter proving. This would account for volume of calibration fluid available, proving timescale etc. For experimental purposes the analysis of pulse output during SVP piston travel, especially for ultrasonic and mass meters, could be improved with increased data sampling frequency. This would increase the 0.05 second sample rate available during this test series. Data analysis would also be improved with the addition of additional software voltage inputs to allow for the output of the second optical measurement switch. Further analysis of true pulse time intervals of meters utilising rotating parts could lead to revised industry standards regarding number of pulses required during one pass of the SVP piston.



12 REVIEW OF PULSE INTERPOLATION METHODS What is pulse interpolation and why is it needed ? Pulse interpolation covers a range of techniques used to enhance the resolution of the pulsed output from an instrument and in the context of this document, a flowmeter. The technique is usually applied to allow calibration using a smaller volume of liquid than might conventionally be needed. Generally, flowmeters with a low resolution are accurate and provide more than adequate resolution to meter bulk quantities of product. If only enough fluid can be collected during calibration to generate say 100 pulses, the resolution error is 1%. For fiscal and high quality meters this is unacceptable and standards normally call for a minimum of 10000 pulses to be collected to reduce the resolution error to 0.01%. To achieve this calibration tanks or meter provers require volumes large enough to allow this 10000 pulse criteria to be exceeded. These large volumes lead to high cost through capital and the size and weight needed for the equipment are substantial. With the design of small volume provers, it was found that a small calibration volume could be measured very accurately even at high flowrates which left the problem of how to increase the resolution of the flowmeters to match the volume capability of the provers. One way of overcoming this problem is to re-design flowmeters to provide increased resolution. This is impractical for many technical and economic reasons. The second approach is to artificially increase the resolution of the meters by utilising pulse interpolation. This review outlines the current methods of pulse interpolation, gives their limitations and discusses their use with small volume provers. Potential new methods and enhancements to the techniques are also discussed. Although the review is written around current techniques for calibrating flowmeters using small volume pipe provers, hopefully it will not be restricted to this and it is left to the reader’s interpretation on how the techniques could be applied to other metrological requirements. Are we talking the same language? Starting with the title - Pulse Interpolation is not commonly used in conversation and therefore some definitions and glossary are required. Clock: Normally used to tell the time - but for this application it is the device used to generate a stable frequency, the period of which is used as a standard reference for time measurements. Detector signal: The signal which starts or stops the indicating device, pulse counters or clock totalisers. These signals are normally generated by micro-switches or optical detectors. They indicate the passage of the displacer, hence marking the calibrated volume. Rotational linearity: This describes the variation in pulse widths or spacing which occur in the output of a flowmeter caused by mechanical, electronic and flowrate variations. Although the terms were defined for rotating meters such as turbines and positive displacement meters, they have been adopted to cover variations in pulse widths from non-rotating meters such as vortex, Coriolis etc.



Intra-rotational linearity: Quantitative measure of the degree of regularity of spacing between the pulses, produced by a rotating meter at constant flowrate, generally expressed as the standard deviation of pulse spacing about the mean pulse spacing. This measure will include cyclic and non-cyclic measurements introduced by the meter mechanism. The pulse spacing is the time between the leading or lagging edges of consecutive pulses. Inter-rotational linearity: The regularity of measurement which repeats in a periodic or cyclic manner attributed to the rotation of the meter. Leading/lagging edge: Rising or falling voltage of a pulse signal generated by a detector trigger or ‘gate’ counter. Phase detector: Electronic circuit which detects a phase difference between two frequencies. Ramp generator: Electronic circuit whose output voltage varies in a controlled way with time. This variation can be linear or shaped in a controlled way although only linear versions are known to be applied to this technology. The generator can be single pass, repeating, or set to increase then decrease in a cyclic fashion as required. Repeatability (of a measuring instrument): Closeness of the agreement between the results of successive measurements of the same measurand carried out under the same conditions of measurement (VIM).

NOTE: The defined conditions of use are usually as follows:

• repetition over a short period of time; • use at the same location under constant ambient conditions; • reduction to a minimum of the variations due to the observer.

Resolution: Quantitative expression of the ability of an indicating device to distinguish meaningfully between closely adjacent values of the quantity indicated (VIM). Rotating meter: Meter, the measuring element of which has one or more rotating parts driven by the flowing fluid (eg turbine meters and displacement meters). Note: (VIM) denotes the definition is taken from the International Vocabulary of Basic and General Terms in Metrology. How do you do it? Three basic concepts are used to provide pulse interpolation: timing methods, pulse multiplication and computer methods.



Timing methods All timing methods are based on the principle that detector signals will occur part-way through an output pulse period. The first detector or gate signal occurs at the beginning and the second at the end of the calibration pass. It is also assumed that the period of a pulse is equivalent to a consistent and known volume but the pulse event is only recorded at one edge of the pulse. If the proportion of each pulse missed before or after the gate signal can be measured, a increase in resolution is derived. Two techniques are used with two variants of the second one. Double-timing or double chronometery This is by far the most common method employed in practice on commercially available flow computers and small volume provers. It was pioneered on ‘conventional’ (large volume) pipe provers in France where very low resolution turbine meters were used for bulk custody transfer and even conventional provers could not collect enough fluid economically. The principle of this method is shown below. The total number of complete meter pulses, n, generated during a proving run are counted as would normally take place in any proving operation. Two time intervals are also measured - T1 and T2. a) T1, is the time-interval between the first meter pulse following the first detector signal and the first meter pulse following the last detector signal. b) T2, is the time-interval between the first and last detector signals. This is the time normally measured to derive flowrate. The interpolated number of pulses is given by:

′ =n nTT

2

1

.

This method, due to the simplicity of the circuitry is by far the most common pulse interpolation method used in practice. A high precision clock, two totalisers for the time measurement and one totaliser for the meter pulses along with fairly simple gating is all that is required. In terms of performance when the pulse widths vary, the method is not quite as good as the quadruple-timing method but the difference is marginal and is normally not considered worth the extra circuitry involved in using the quadruple method as given below. It is much superior to the simple Quadruple timing method. Quadruple-timing. Two variants of the quadruple-timing method of interpolation are possible. The simple method described at the end of this section is not advised but included as it may be found in older systems. This, the first variant, provides slightly better performance than double chronometry while the second variant is significantly poorer. Both require more complex electronics but the first preferred method is by far the most complex as it requires anticipatory circuits.



The principle of operation of the first variant is shown below. The total number of complete meter pulses, n, generated during a proving run are counted as would normally take place. Five time intervals are also measured t1, t2, t3, t4 and T2. a) t1, is the time-interval between the first detector signal and the first meter pulse following that signal: ie the fraction of the pulse remaining after the first detector is actuated. b) t2, is the time-interval between the last meter pulse before the first detector signal and the first meter pulse after it: ie the width of the pulse present when the first detector switch is activated. c) t3, is the time-interval between the second detector signal and the first meter pulse following that signal: ie the fraction of the pulse remaining after the second detector is actuated. d) t4, is the time-interval between the last meter pulse before the second detector signal and the first meter pulse after it: ie the width of the pulse present when the second detector is actuated. e) T2, is the time between detectors to allow calculation of flowrate. This is not used in the interpolation process. The number of complete pulses, n, in the main pulse count is counted in the normal way by a counter gated by the detector signals. The interpolated number of pulses, n1, between the detector signals is then:

n ntt

tt

1 1

2

3

4

= + − .

This method does offer an advantage over the double timing method where intra-rotational linearity (pulse variations) becomes an issue. A counter, clock and five timer totalisers are required along with quite complex gating circuits. As the circuitry has to anticipate the occurrence of the detector signal to allow t2 and t4 to be measured, the problem is obvious. Although showing improved performance over double chronometry, most applications do not find the complexity gives any significant advantage to this method. The second variant is a simpler form of quadruple timing used to be utilised to avoid the anticipatory circuits. In this t1 and t3 are the times of the first pulses after detectors rather than the times of the pulses present when the detector is actuated. As this method clearly gives larger errors in the presence of varying pulse widths: ie poor intra-rotational linearity, this simple method is no longer advised. Pulse multiplication or phase-locked-loop? At first sight this method, being radically different from the timing methods, provides a simpler solution to increasing resolution. The concept is to multiply the output frequency from the meter by a factor to produce a measured pulse frequency suitably higher than the meter frequency. Calibration can then take place conventionally with no requirement for extra counters, timers and calculations.



The pulses from the meter are introduced to input 1 of the phase comparator. This device produces a voltage proportional to the difference in phase (frequency) of the two inputs. This output voltage is passed to the voltage controlled oscillator (VCO). This device generates a frequency which is proportional to the input voltage and is chosen to provide a frequency range larger by a suitable factor than the meter frequency. The output signal of the VCO is fed back, through a frequency divider, to input 2 of the phase comparator. As the output voltage of the phase comparator is proportional to the difference in phase (or frequency) between its two inputs, the output of the VCO is continually being servo-controlled to ensure that the frequency and phase of the two inputs are identical. As the frequency divisor has reduced the output frequency from the VCO by the given factor, this factor is in fact the multiplication factor for the system. The selection of frequency divisor, R, thus determines the pulse interpolation divisor. The interpolated number of pulses collected during the proving run is normally expressed as:

n nR

1 =*

.

Where n* is the number of multiplied pulses collected from the multiphase output and R is the selected divisor (or multiplication factor). In many ways this method seems at first sight to be by far the best solution to interpolation as the output frequency is simply a multiple of the input hence increasing meter resolution. Unfortunately, to achieve precise control of the feedback loop, it is necessary to filter the output of the phase comparator to avoid sudden VCO changes and hence instability of the output frequency. This filter, normally of the simple RC type, has the property of momentarily retaining the voltage required by the VCO to maintain R times the meter frequency between each phase comparison. Selection of the filter’s time constant is where the problem lies. If the time constant is too short, the output frequency can become unstable. If the time constant is too long, the output frequency will not track changes in the input frequency due to legitimate changes in the measured parameter such as a flowrate changes caused by the prover launch. It is this lack of precise control of the filter circuits across a range of frequencies, (flowrates and meter types) and in the presence of poor intra-rotational linearity that has lost this method popularity. Computer methods - the future? As yet no known methods are in production other than those given above but it is possible that other techniques could be evolved to utilise the advances in high speed computing and counting. With the rapid development of high speed computer data acquisition and calculation techniques, the potential of developing methods based on monitoring and correcting for the time of every pulse from the meter and therefore deriving the fraction of a pulse missed or gained relative to the average volume for that individual pulse becomes possible. Similarly the potential for producing a multiplied frequency output based on the input pulse frequency corrected for inter- and intra-rotational linearity becomes theoretically possible.



If multiple counters or a microprocessor based system was used to time the period of each consecutive pulse from the flowmeter and these times are collected in a controller along with the position of the meter prover switch operation in the time frame, full information about the volume/pulse time can be derived and a profile of the meter performance giving average volume per pulse or volume per revolution can be provided. From this data a high quality indication of the meter performance in terms of short and long term repeatability can be provided as well as potentially a characterisation of the prover or flow system being used. Clearly this is a very powerful technique but the hardware and software have not been developed to allow any commercial application of the techniques. No further guidance is given to this method’s application. How do you design and use the timing methods? a) Resolution The resolution of the final interpolated pulse count for high accuracy flow measurement calibrations should in all instances be better than 1 in 10000. As this criteria relates to the meter under test, its resolution and the volume of the calibrator, this is not an absolute criterion but one which must be looked at for each application. b) Number of significant digits for the interpolated number of pulses, n′

The interpolated number of pulses, n′, generated by the timing methods is not a whole number. For the timing methods which yield a fractional result, there will be a practical limit on the number of decimal places which are used for n′. The number of significant figures used to finally express n′ should not exceed the number allowed by the resolution of clock and counters used even though a greater number may be generated by the calculation. In practice, for fiscal metering of oils, ISO 7278/3 states that n′ should be rounded to five significant digits, not more and not less. c) Detector switch signal The switching edge from the detector should be well defined and repeatable (some mechanical switches produce signals with non-repeatable lagging edges due to switch bounce). It is necessary to define and use the same switching edge in all operations. d) Clock stability With modern electronic clocks and the relatively short times involved, stability should not be a concern. It is essential that during the design and use of the equipment that the chosen clock does actually have a stability commensurate with the required resolution. e) Timer resolution To ensure that the resolution of the interpolated pulses meets the 1 part in 10000 criteria, the clock frequency must be suitably high.



For double chronometry method, the period of the test, ie the time T2 , should be at least 20000 times greater than the reference period tc of the clock (ie the reciprocal of the clock frequency) used to measure the time-intervals.

That is:

T2 ≥ 20000tc therefore nf fm c

≥20000

therefore fn

fc m≥20000

where fm is the maximum meter test frequency,

fc is the clock frequency, and n is the number of pulses collected during the proving run.

For the Quadruple timing method the period of the test, ie the time T2, should be at least 40000 times greater than the reference period tc of the clock (ie the reciprocal of the clock frequency) used to measure the time-intervals. This gives by a similar analyses to the double chronometry

fn

fc m≥40000

Conditions of use and design criteria for the phase-locked-loop method a) Frequency (locking) range Any phase locked loop circuit will only operate over a limited frequency range. The operating frequency range should always fully cover the frequency expected from the meters being used. NOTE: A minimum frequency rangeability of at least 100 to 1 is recommended for any phase locked loop system to be used for pulse interpolation. b) Resolution To obtain a resolution better than ± 0.01% the count n* should be equal to or greater than 10000 pulses. The great unknowns The use of pulse interpolation is based on the assumption that there is not a significant variation in the frequency of the pulses during the calibration period. Any variations in frequency will degrade the accuracy. Variations in the pulse frequency come from many sources and the variations can themselves be random, short term repeatable patterns, longer term cyclic changes, longer term drift or sudden step changes. The causes can come from many sources. Flowrate instabilities, uneven spacing of the mechanical assembly of the meter, inconsistencies in gearing, correction changes in the flowmeter software generating the frequency are all examples. To complicate things in any one installation a number of different causes can exist at the same time.



Stability of flowrate Pulse interpolation methods are based on the assumption that the frequency from the flowmeter is stable during the time of the proving pass. Any change in flowrate will be shown as a change in meter frequency (and pulse period). Stability of flow can be difficult to achieve and measure. Clearly any fluctuations in flow are undesirable but for practical guidance the standards call for the fluctuations in the flowrate, during a pass of a prover displacer, to be less than ±2% of the mean flowrate. It presumed by current thinking that a slow drift in flowrate will be more acceptable to pulse interpolation than short duration changes which may be produced by pulsations or surging. One particular flowrate change occurs when the prover displacer is launched. The time between launch and the first detector must be long enough to allow the meter output to stabilise to the new flowrate before the first measurement detector is reached. This is not usually a problem for fast response turbine and PD meters but microprocessor based meters often have long time constants on the output even if the meter sensor responds quickly. Other sources of pulse fluctuation In an ideal meter, when operating at a constant flowrate, the emitted pulses will have a constant period. In practice, the spacing of pulses will be somewhat irregular, owing to intra- (and inter-) rotational linearity and other random fluctuations. Pulse spacing can vary in three distinct ways. These are not exclusive and are frequently present in the same meter. a) A random variation of pulse widths is often caused by meter bearing variations, output from a vortex meter or gearing backlash. Pulse widths can vary by 1 to 20% or more depending on the cause. b) A random variation of pulse widths but repeated during each revolution of the meter. This is the second level of fluctuation and is usually associated with turbine flowmeters where the spacing between the blades is uneven but clearly repeats as the meter rotates. In general the number of pulses in the repeating pattern is low (4-10) but could be greater in the case of rim type turbines. Variations in pulse widths tend to be in the 4 to10% region. c) A long cyclic variation of pulse width often associated with PD meters where the rotation of the meter produces a cyclic increase and decrease in pulse width during the rotation of the meter rotor or gearing. Variation of the pulse widths occurs with a period of tens or hundreds of pulses. Variations are seen in the 5-30% level. In practice these distinctive features can be superimposed and are of very different magnitudes.



What does this mean? Variations in pulse widths reduce the accuracy of a pulse interpolation system. The greater the degree of irregularity, the more random error is introduced in the interpolated figure and this can only be counteracted by collecting more raw pulses during a proving run. Even this is not exactly true because it is found that the relationship between the number of pulses, the pulse variation and the random error is asymptotic and as a result at high pulse fluctuations, no significant improvements can be made by increasing the number of pulses collected and in fact lower error can be achieved by counting raw pulses. It has often been noted in practice that not always does poor repeatability on a calibration using pulse interpolation indicate a poor meter. The same meter may give good repeatability when calibrated against a large volume with no interpolation. What do we do? Clearly the minimum number of pulses to be collected to provide good proving conditions has been at the heart of debate on the use of interpolation since its inception as a flowmeter calibration tool. Even now not enough is known about the effect to lay down mandatory rules, but if the guidelines given below are followed then the errors resulting from pulse spacing irregularities are considered unlikely to be very significant. As well as not being able to lay strict guidelines, in practice, the magnitude of pulse variation is not generally known, not easily measurable in the field and certainly not quantifiable in terms of pattern without specialised equipment. Theoretical analyses by modelling and by statistical analyses have shown that the error introduced by pulse variations reduces in proportion to the number of raw pulses in accordance with a square law. It also shows that the number of pulses required to provide an acceptable proving run is very much smaller for random fluctuations, more for repeating patterns and very many more for cyclic variations. Very little published work exists to relate carefully measured pulse variations to the number of pulses collected and the repeatability of calibration in practice. What does exist supports the conclusion above, but for all practical test work it appears that better repeatability is found than is predicted by the theoretical or simulation work. To assist the user, guidelines have been produced within ISO 7278/3 to advise on the minimum number of pulses required to give an acceptable calibration with a spread of results within ±0.02%. If the pulse intervals scatter in a random manner as described in a) above, the following equation gives an estimate of the minimum number of pulses required.

nm = 500(σ1)2 where: nm is the recommended minimum of pulses;

σ1 is the standard deviation of the pulse time intervals expressed as a percentage of the mean pulse interval.



The constant 500 was derived from theoretical work and field experience and is no doubt a compromise figure. Where large degrees of cyclic intra-rotational non-linearity are present, this equation may underestimate or overestimate the number of pulses required. As the repeatability of the calibration is a complex function of the standard deviation related to the number of pulses in the repeating cycle, the nature of the cycle, and the pulses collected in the proving run, no more detailed guidance can be given. In all cases it is recommended that more than 100 pulses should be collected in a proving pass and calibration should be with more than one meter rotation or intra-rotational cycle. The above guidance is only that - guidance. Many practical results show acceptable calibrations with fewer pulses than recommended above; others require more. What else can be done? When using small volume provers it is common practice to effectively increase the number of pulses collected by averaging a number of proving passes to provide one result. It is common to average 5 passes, increase up to 10 if good repeatability is not found but the IP advises increasing only to a maximum of 20 if it is known that pulse variation is the problem. It has also been reported that where large amounts of intra-rotational linearity are present, with a low level of purely random component, eg a good turbine meter with poor blade spacing, good results are obtained by counting only one pulse per revolution of the meter. This reduces the number of pulses collected by the number of blades on the turbine (typically 6 to 10) but also very significantly reduces the standard deviation of pulse widths, hence an improved result is obtained. Pulse width variation reductions from 10% to 1% can be achieved by this technique. The above recommendations are based on practical and theoretical experience. In each particular application where poor calibration repeatability is found, poor meter repeatability or effects from pulse interpolation errors may be the cause. Separation of the two effects can be difficult but looking at the pulse spacing from the meter would be the first diagnostic tool. Increasing the number of collected pulses may improve the result and this is done normally by increasing and averaging a number of meter prover passes. How do you measure pulse widths? Currently, in field calibration situations the pulse width variation is not measured and the equipment and staff to take the measurements are not normally available. This section provides some techniques to both correctly measure the intervals or to provide some idea of the intervals using readily available equipment. Electronic timers can be used to measure the time interval between two pulses. Unfortunately due the time taken to read the result, by hand or by data logger, they can not be used to measure every pulse. They can be used as a simple useful tool by triggering and logging, by hand or computer, a large number of pulse widths. If using computer logging, it is advisable to modify the logging rate to avoid synchronising the measurements with the meter revolutions.



Counter boards within computers with suitable acquisition software can be used to record every pulse. The actual pulse time may not be accurate due to time delays but as long as it has a consistent error the variation in pulse widths can still be recorded. Older storage oscilloscopes, with screen storage, can show an image of the pulses if switched to ‘store’ and ‘free run’. This provides a very rough guide to the spread of pulse intervals. Modern digital storage oscilloscopes can store the incoming pulse signal and play it back on to the screen or download it to a computer. As this information is the pulse voltage and the time from a start point, extracting the time between pulses can be difficult. The capabilities of individual instruments would have to be examined to assess their ability to give the required performance. Visual interpretation of a digital storage scope display can also be used but this is time consuming. Possible use of computers with data acquisition boards and programmed to simulate an oscilloscope can also be considered. Logic analysers, used to test the operation of complex electronic systems, will time and record the data. Most analysers have multiple channels, only one of which is required for interval timing. Each channel has the capability of timing many hundreds of events, either in real time or relative to the last event. The collected data can be displayed on the instrument’s screen or in many cases transmitted to a computer or printer for calculation and display. Usually signal processing will be required as analysers are designed to operate at normal electronic logic levels of 5 V or 12 V with square edged pulses



BIBLIOGRAPHY 1 ISO 7278-2: 1987, Liquid hydrocarbons - Dynamic measurement - Proving

systems for volumetric meters - Part 2: Pipe provers. 2 ISO DIS 7278-3: 1997, Liquid hydrocarbons - Dynamic measurement -

Proving systems for volumetric meters - Part 3: Pulse interpolation. 3 International Vocabulary of Basic and General Terms in Metrology -

International Organisation for Standardisation, 1993. 4 Petroleum Measurement Manual Part X Section 1 field guide to proving meters

with a pipe prover. Institute of Petroleum, London. 5 Prediction of flowmeter calibrations repeatability using compact provers. R Paton,

NEL, East Kilbride, Glasgow. 6 Test Report on irregularity of pulse spacing for PD meters. A Takada Oval

Engineering Japan. (Private correspondence.) 7 Pulse interpolation. Revision of ISO 7278/3. R Paton, North Sea Flow

Measurement Workshop, NEL 1998. 8 Simplified statistical analyses of the effect of pulse spacing variation on the

repeatability of meter proving where pulse interpolation is employed. A T J Hayward, SGS technical report, June 1988.

9 Influence on the irregularity of output pulses from flowmeters in the pulse

interpolation method. Y Ogawa (Private correspondence.) Oval Engineering Japan.

10 Ultrasonic meters for oil flow measurement. G Brown, NEL, East Kilbride, DTI

Report, November 1996.



LIST OF TABLES 1 Calibration of 3-inch Brooks turbine meter against SVP and 6T weighbridge

simultaneously 2 Calibration of 4-inch Danfoss Ultrasonic meter against SVP and 6T weighbridge

simultaneously 3 Calibration of 3-inch Smith PD meter against SVP and 6T weighbridge

simultaneous 4 Calibration of 3-inch Micromotion mass meter against SVP and 6T weighbridge

simultaneously 5 Brooks 3-inch turbine meter. Results of tests on the effect of number of SVP

passes per run versus test meter repeatability 6 Danfoss 4-inch Ultrasonic meter. Results of tests on the effect of number of SVP

passes per run versus test meter repeatability 7 Smith 3-inch positive displacement meter. Results of tests on the effect of number

of SVP passes per run versus test meter repeatability 8 Micromotion 3-inch Coriolis mass meter. Results of tests on the effect of number

of SVP passes per run versus test meter repeatability 9 Calibration of a 3-inch Brooks turbine meter against SVP. Plenum pressure versus

test meter repeatability 10 Calibration of 4-inch Danfoss U/S meter against SVP. Plenum pressure versus

test meter repeatability 11 Calibration of 3-inch Smith positive displacement meter against SVP. Plenum

pressure versus test meter repeatability 12 Calibration of 3-inch Micromotion CMF300 mass meter against SVP. Plenum

pressure versus test meter repeatability 13 One pulse per revolution - Brooks 3-inch turbine meter 14 One pulse per revolution - Brooks 4-inch (damaged) turbine meter.



LIST OF FIGURES 1 Small Volume Prover schematic 2 Test facility schematic 3 Calibration of Brooks 3-inch turbine meter against SVP and NEL reference 6T

weighbridge simultaneously 4 Calibration of Danfoss Ultrasonic meter against SVP and NEL reference 6T

weighbridge 5 Calibration of Smith PD meter against SVP and NEL reference 6T weighbridge 6 Calibration of Micromotion mass meter against SVP and NEL reference 6T

weighbridge 7 Brooks 3-inch turbine meter. SVP passes versus repeatability 8 Danfoss 4-inch Ultrasonic meter. SVP passes versus repeatability 9 Smith PD meter. SVP passes versus repeatability 10 Micromotion CMF300 mass meter. SVP passes versus repeatability 11 Brooks turbine meter. Plenum pressure tests 12 Brooks 3-inch turbine meter. Frequency output during SVP operation at various

plenum pressures 13 Danfoss Ultrasonic meter. Plenum pressure tests 14 Danfoss Ultrasonic meter. Frequency output during SVP operation at various

plenum pressures 15 Smith PD meter. Plenum pressure tests 16 Smith 3-inch PD meter. Frequency output during SVP operation at various plenum

pressures 17 Micromotion CMF300 mass meter. Plenum pressure tests 18 Rosemount CMF300 mass meter. Frequency output during SVP operation at

various plenum pressures 19 Brooks turbine meter. SVP optical measurement switch tests 20 Danfoss Ultrasonic meter. SVP optical measurement switch tests 21 Smith PD meter. SVP optical measurement switch tests 22 Micromotion mass meter. SVP optical measurement switch tests.

flowmeter calibrations using a small volume prover - · pdf fileflowmeter calibrations using...

Documents