This paper discusses Small Volume Provers (SVPs), used in the oil and gas industry to validate the performance of custody transfer meters. Recently Coriolis mass flow meters have been introduced for custody transfer; while these offer reduced maintenance requirements over traditional PD and turbine meters, proving Coriolis flow meters using small volume proving are challenging. This paper presents Small volume proving results for a Coriolis meter that matches or exceeds the most stringent requirements for custody transfer. This is achieved in part by using a novel signal processing technique that reduces the dominant component of the measurement noise, associated with so-called Coriolis mode vibration, with negligible loss of dynamic response.


In the oil and gas industry, custody transfer is a transaction in which ownership of a quantity of material is transferred from one operator to another, for example from the well operator to the first transportation operator, via pipeline, truck, or ship. Subsequent custody transfer transactions may take place as the crude commodity arrives at the refinery and as refined products leave. The high value of the material, and perhaps the payment of government duties, leads to stringent measurement requirements in order to verify the quantity of material involved in each transaction.

Current standards require flow meters used for custody transfer measurements to undergo regular verification of measurement performance; a process known as meter proving. Commonly this is carried out using a specialized piece of equipment known as a prover, which is based on the principle of filling a known volume. Examples include bi-directional ball provers and compact pistons or Small Volume Provers (SVP). Generally, large bi-directional ball provers are used for high-volume custody transfer operations. Small volume provers are typically part of a mobile truck-mounted system that travels from site to site providing meters on a scheduled basis.

To carry out proving, the Small volume proving system is connected in series with the meter under test. Fig.1 shows a typical SVP design. Flow passes through a cylinder of known volume. The prover’s piston contains a poppet valve which, when open, allows.

flow to continue irrespective of the piston location; specifically, this allows the piston to be pulled to the upstream position without interrupting the flow. With the piston drawn to the upstream position and the poppet valve closed, the piston moves with the fluid as it sweeps through the calibrated volume. At the start and end of each “pass”, high-precision position detectors generate electrical signals to indicate the passage of the piston. During the pass, the volumetric flow measurement from the meter is integrated, so that a comparison can be made between the known prover volume and the reported volumetric flow from the meter under test.

10:1 flow range. At the highest flow rates, a prover pass can be as short as 1.5s, raising such issues as the noise properties and dynamic response of the flow meter; consequently, the use of electronic or software damping of the measurement output is commonplace. It is also possible to define a “run” at a particular flow rate as consisting of one or more prover passes, where the meter output is summed over all the passes, to increase the effective duration of each run. The performance parameters of interest include the repeatability of the meter over a set of runs (each run consisting of one or more passes) at the same flow rate, and the linearity of the meter over the flow range of the test procedure, as explained in more detail below.

Mechanical meters, such as turbine or Positive Displacement (PD) devices, have traditionally been the most used in liquid custody transfer applications. Consequently, the most widely used meter output is the pulse output. This is generated using the blades of a turbine meter, or a gear within a PD meter, which are positioned near a Hall effect sensor coil. As the blades or gear teeth rotate and pass the pickup coil, a pulse is generated, the output rate of which is proportional to the rotation speed, and hence the flow rate of the fluid, with each pulse representing a discrete volume of fluid. As electronic meters have been introduced, the pulse output has been retained, the microprocessor generating pulses to represent the observed flow rate, thus maintaining compatibility with established standards, practice, and equipment.

One limitation in the use of pulse output is its discrete nature—it is only possible to detect an integral number of electronic pulses; this may place a significant constraint on the repeatability achievable overruns of short duration, given that it is unusual for pulse output frequencies to exceed, say, 10 kHz. An interpolation technique known as double chronometry or the double-timing method [2] has been developed to provide an estimate of the fractional number of pulses during a run, thus eliminating the issue of pulse discretization error.

Over the last 10 years, Coriolis mass flow meters have gained acceptance in custody transfer applications, offering reduced maintenance requirements compared with mechanical meters due to their lack of moving parts. In 1996, Whitman described field proving of a Coriolis meter using a small volume prover [3]. Recently, the Weights and Measures Division (WMD) of the National Institute of Standards and Technology (NIST) published a series of articles about small-volume provers [4], including sections on Coriolis meters. The American Petroleum Institute (API) Manual of Petroleum Measurement Standards, Chapter 5, Section 6 on the measurement of liquid hydrocarbons by Coriolis meters [5], provides guidance for the appropriate number of proving runs as well as the number of passes per run for different types of provers, including SVP, in a table, reproduced as Table1. Detailed examination of these documents, however, reveals that guidelines and good practice recommendations are still open to numerous interpretations.

For instance, questions often arise as to exactly how many passes per run are permitted, what pulse output damping to use, and what constitutes a minimum acceptable run time.

An early (1998) and detailed examination of the issues associated with on-line calibration of flow meters using an SVP is provided in the NEL guidance note no 7[6], which describes trials of a range of flow meters, including PD, turbine, ultrasonic and Coriolis meters. Table 2, reproduced from [6], shows the effect of changing the number of passes per run in the NEL trials. This demonstrates the gulf in performance at the time between mechanical and electronic devices for single pass runs, with turbine and PD devices showing repeatability of less than 0.01%, while the ultrasonic and Coriolis meter could only deliver 0.66% and 0.12%, respectively. Another issue explored in[6] is the delay in pulse

output from electronic meters. For example, it proved necessary to delay the gating of the meter output pulses to accommodate electronic and software time constants. One advantage of a purely mechanical flow system such as a PD meter is that it exhibits negligible delay in the

generation of pulses. However, for electronically based devices, measurement noise, and microprocessor delay may be detrimental to meter performance, especially on short passes. Of course, the use of damping on the pulse output signal reduces noise at the expense of increased delay [7]. Accordingly, proving is considered by some in the industry to be something of a black art, especially when it comes to the setting of the damping factor for a meter that is difficult to prove in the field. Thus Vandiver [8], in offering guidelines on methodologies for Coriolis meter proving, recommends “The pulse output damping value should never be set higher than 0.8s”. This recommendation is illuminated by the Coriolis meter frequency output signal shown in Fig. 4 of [6], which shows that the pulse output may not have stabilized to a steady flow by the start of the proving run if the damping setting is too high.

The issues of pulse output delay and damping have become of sufficient concern within the industry that an API Task Group was formed to investigate the effect of microprocessor-generated pulse output delay on field proving of meters, including Coriolis and ultrasonic meters, using SVPs. It is hoped that the current paper will offer some insights into these issues.

The Invensys University Technology Centre for Advanced Instrumentation at the University of Oxford has developed Coriolis metering technology with improved performance characteristics, for example with respect to dynamic response. The reader is referred to the companion paper in this special issue on Coriolis mass flow metering [9] for a background on Oxford’s transmitter technology, including the use of Field Programmable Gate Arrays (FPGAs). The importance of pulse output generation has been addressed through the development of a patented algorithm [10] with high precision (to one part in 10 million averaged over one second [11]) and minimal delay.

This technology provides a variety of means for ameliorating what are perceived to be the principal difficulties when calibrating a Coriolis meter using an SVP:

  • The first, which is now rare, is the production of irregular pulse streams from the meter. This was found in many early Coriolis meters that were changing frequency or missing and adding pulses was the technique used as part of the pulse generation software. This gave rise to uneven or intermittent pulse mark-space ratios. Uneven pulses give rise to poor repeatability when interacting with pulse interpolation, even to the extent that better repeatability can be achieved without using interpolation. In the Oxford transmitter, the 40MHz clock signal to the FPGA ensures a smooth 50% duty cycle: even at a high pulse output frequency of 10kHz, there are 2000 clock ticks between edges. This ensures that pulse interpolation techniques will improve repeatability. Of course, this technology is entirely immune to the pulse irregularities arising from wear in mechanical meters.
  • The second issue arises from damping and time constant. The dynamic response of the meter is pivotal: as previously mentioned, the meter needs to generate a stable output frequency after the initial drop-in flow rate caused by the launch of the piston, but before the first detector triggers the start of the pass. Previous trials on the Oxford transmitter [12] have demonstrated that for the bent tube design used in these trials, a very rapid (4ms) step change in flow is indicated via the pulse output from the meter within 16 ms of the start of the physical change in flow and that the step change is completed on the pulse output signal within a further 20ms or less. For a high-frequency straight flow tube, the companion paper [11] in this special issue reports a delay in the change to the pulse output of less than 4ms and an insignificant extension of the duration of step change. Of course, in both these experiments no damping is applied to the pulse output signal as it is considered unnecessary.
  • A third issue is also response-time-based. Many SVPs show a drop in flow during a pass due to the gas plenum pressure dropping as the piston moves, which must be successfully tracked by the meter. This issue is perhaps best considered as a low-frequency flow modulation, as opposed to a step change in flow discussed above. However, as

Applied to the proving problem, the Oxford transmitter connected to a commercial bent flow tube has passed the most stringent interpretation of current proving standards at a commercial testing laboratory. Specifically, over an 11:1 flow range, covered in eleven steps, using five repeated prover runs each of a single pass only, the meter achieved a repeatability of less than 0.05% over the range, with an average repeatability of 0.017%. This was achieved without applying any damping on the pulse output signal.

The Oxford research transmitter collected high bandwidth data during the proving runs, and this paper provides a detailed analysis of the data and describes the signal processing techniques used to provide the measurement performance. Section 2 describes the proving trials and results obtained by the testing laboratory. Section 3 provides the corresponding data collected within the Oxford transmitter and demonstrates that they are essentially consistent with those of the testing laboratory. Section 4 provides a more detailed analysis of one set of prover runs and explains how the signal processing techniques employed can simultaneously reduce noise with little loss of response time so that pulse damping is unnecessary. Overall, the aim of the paper is to demonstrate that the latest advances in Coriolis meter signal processing can provide reliable technology that is able to match the stringent proving requirements of the oil and gas industry.

Oxford transmitter data records

There are two main types of detailed records that can be collected by the transmitter. One records the raw data from the two flow-tube sensors along with flow-tube temperature information in a so-called “sensor” file, updated at 12kHz. This file type allows the most fundamental reconstruction and analysis of the measurement calculations but provides no information on other aspects of meter operation such as flow-tube control or external inputs. The second type of file stores up to 60 channels

Signal analysis

Having demonstrated that broadly speaking the internally recorded data are consistent with those observed via the pulse output, in this section an analysis is provided of the raw sensor data and the signal processing algorithms used to generate these results.

In virtually all Coriolis mass flow meters, the raw sensor data are generated from two sensors (usually velocity-sensing coils), which provide essentially sinusoidal analog signals derived from the motion of the flow tube. While the primary

Concluding Remarks

demonstrated that the latest generation of Coriolis mass flow meters are able to match those requirements. Insight has been offered into the signal processing issues associated with providing high measurement precision and fast dynamic response, most specifically with the relationship between the drive and Coriolis modes of vibration, and their influence on the raw flow…


The authors are grateful to Steve Whitman of Coastal Flow Measurement, Inc. and Richard Paton of NEL for detailed technical comments on earlier drafts of the text.


M. Henry et al.

Response of a Coriolis flow meter to step changes in flow rate.

Flow Measurement and Instrumentation (2003)

International Standards Organization (ISO). Liquid hydrocarbons–Dynamic measurement–Proving systems for volumetric…

International Standards Organization (ISO). Liquid hydrocarbons–Dynamic measurement–Proving systems for volumetric…

Whitman SK. Operational experience proving mass flow meters with small volume provers. In: The energy week conference…

Lee GD. Part 1: Small volume provers–identification, terminology, and definitions, March 2005. Part 2: Small volume…

The American Petroleum Institute (API), Manual of Petroleum Measurement Standards, Chapter 5, Section 6: Measurement of…