The accuracy of open channel ultrasonic flowmeters

Real world working conditions on water sites mean that ultrasonic level sensors used on open channel flumes and weirs can be significantly in error, because of ambient changes: is there a solution?

Ultrasonic level measurement techniques have been used for around 25 years for open channel flow measurements in the water industry, working with primary devices like flumes and weirs.

The ultrasonic measurement, being non-contact, was attractive in that it offered minimal need for regular maintenance, and could work directly in the main flow stream, ie did not need the stilling well previously used for float-based level measurement systems.

But an ultrasonic level measurement device measures the distance from the transducer face to the target, the water surface.

Since the flow is derived from the height of the water above the weir, or base of the flume, or point of the V-notch, all the measurements are referred to this reference point, which is say 500mm away from the transducer face, in a small system.

How accurately will an ultrasonic system measure this distance? Obviously the flow measurement reading at very low flows is going to be relatively inaccurate, because the flow measurement is derived from the difference between two relatively large distance readings, and the instrument accuracy at say 1% gives a possible measurement error of 5mm.

Some level measurement equipment specifications can offer a better measurement repeatability, such as 0.25%, but usually specify a minimum accuracy value of around 3mm, related to the step size in the waveform, dependent on the wavelength of the ultrasound being used.

The significance of this large error at very low flows is not important, when computing the daily total flow, assuming there are some periods of high flows, where the discharge flow is at a higher level in the channel.

These higher flows can be measured hopefully to a better accuracy, and since they represent the largest contribution to the total daily flow, the MCERTS effluent discharge requirement of an overall total daily measurement accuracy of +/-8% is achievable (

The MCERTS testing of the performance of open channel level measurement devices is very much a laboratory exercise, and the practical conditions, particularly outside in the open, on a discharge to a stream for example, can be very different! This has been shown up on various water industry evaluation tests over the years.

Let’s look at some of the on-site problems that can be found.


This varies with the temperature of the air that the sound is passing through.

Every 6C change of air temperature changes the speed of sound by 1%.

Most manufacturers put a temperature sensor in the ultrasonic transmitter head to make an automatic correction.

This does not really work very well in any conditions of fairly rapidly changing temperatures, and also is really upset by the summer sun shining on the sensor head, making it rise significantly above the air temperature.

In extremes there might be up to 24C error or 4% on the 500mm range measurement, ie 20mm.

Several manufacturers, and some users, add sunshades to their sensors, to reduce this problem.

The speed of sound also varies a little with gas composition and humidity, but for open atmosphere measurements these will not be significant.

Humidity only affects the speed of sound in tanks where the liquids are heated above say 80C.

What no theory tells us, and what might only be observed on site, is how the speed of sound might be affected by the water droplets in an early morning mist, in cold conditions, rising off (relatively hot) liquids!.


The answer to such temperature problems was originally sought by the use of reference pins, positioned at around 300mm from the transmitter face, giving a small reflection to enable measurement of the speed of sound over this known distance.

The variety with the tube created an ideal place for spiders to live, and interfere with the calibration, and the variety without the tube was an ideal shiny perch for passing birds, flies, butterflies et al.

Plus these reference pins have a problem with the calibration changes caused by water droplets, dirt, ice and snow, to various degrees.



With changing temperatures throughout a 24 hour day of maybe a total of 20C in the air, there is a lag in the temperature sensor and the accuracy correction.

A reasonable estimate might be 6C for 3 hours at each end of the day.

So this gives an error of 1% in the speed of sound used.

Take the V-notch weir used in the example quoted on the MCERTS website, with a zero level 500mm from the transducer face.

Say the flow is running at 120mm above the notch, giving a flow of 1.446 LPS.

The sensor to surface range is 380mm, so a 1% error is 3.8mm, and the indicated flow at 123.8mm is 1.563 LPS.

(Note that this example is for the temperature in the sensor head being 6C lower than the air temperature, leading to a higher flow reading than should be recorded).

The V-notch weir is the most extreme example, because of the 5/2 power law, but this flow error resulting from a 3.8mm inaccuracy in measurement is just over 8%.

This error is on top of any other errors in electronics timing, the weir itself, and the set up of the unit, and it is at a reasonably high flow rate, therefore significant in the daily total calculations.

What is the conclusion? Inevitably, the conclusion is that for accurate measurement there is a need to look closely at how the speed of sound is calculated, particularly in relation to the use of sun shields, the positioning of the temperature sensor, and also, the choice of flume/weir, with the ultrasonic sensor as close as possible to the flow.

But it is still not necessarily going to give the accuracy that might be needed.


This has become identified as a significant problem, even in UK installations, and even when sunshades are used: a sunshade is not normally the complete answer.

The temperature reading from a sensor in the ultrasonic transmitter, positioned in the full glare of sunshine above the flow stream, can be maybe 20-25 degrees hotter than the actual air.

This solar gain effect, fed into the open channel flow calculation, produces an error in the monitored flow rate and totals: the recorded flow value is typically much lower than the actual flow, (with a sensor reading a high temperature) and the under-reading error can exceed the 8% considered the limit for MCERTs installations by a long way.

Apparently such discrepancies have been reported in sunny conditions at several installations in Anglian and South West Water, and have led to some evaluation trials to quantify the problem, and test the possible solutions.

A display of the temperature being used by the flow calculation is relatively easy to see on most micro based units via the standard keyboard.

Alternative approaches are being proposed to solve this problem, but a sunshade seems a sensible first step!.


A new approach has just been announced by Pulsar Process Measurement, that uses two sensors, as a variant of the reference pin system.

One sensor is mounted higher up than the other, at a known extra distance from the liquid.

Both sensors transmit a pulse towards the liquid surface at the same time.

Only the lower sensor acts as a receiver.

This sensor detects and times the arrival of its own pulse, and then does the same for the pulse from the higher sensor.

The time between these two pulses allows the unit to calibrate for the speed of sound in the air space between the two sensors, at that moment.

Given good air mixing, all the worries of temperature changes through the day are solved.

The system is on trial in various water authorities where their applications suggest that a more accurate system is needed.


The original Mobrey MSP90 system, in use for open channel flow metering since the late 1980s, had the capability to accept a separately installed air temperature sensor, wired into two terminals in the sensor head.

(I have to admit rather detailed knowledge of this from working for Mobrey at the time and specifying the original MSP90 unit).

Water industry evaluation trials discovered that by adding this temperature sensor on site, and positioning it in a shaded or North facing area within the flume, improved the temperature, and therefore the flow measurement accuracy considerably.

Using this technique, Anglian Water was able to upgrade the accuracy of the data available from a large number of their installations, in all weather conditions.

More recently, further trials were organized in South West Water, and these included an evaluation of the new MSP900FH unit, designed to be used with the separated temperature transmitter.

Robin Lennox of SWW confirmed that there had been some concern about the accuracy of open channel metering systems across the seasons, and initial investigations suggested that the effects of solar gain on the temperature sensors embedded in the ultrasonic transmitters were significant.

It was necessary to find a way of improving the accuracy of the measurement data.

Trials with the Mobrey Measurement MSP900FH showed that the use of a separated temperature sensor provided the improvement sought .

Results on their test rig were generally within ± 0.5% on distance measurement (for typical installation dimensions) when the sensor was sited to monitor air temperature from a properly shaded position.


This could be the next step: but the current lower cost free to air radars currently seem to have around a 10mm accuracy minimum – but maybe there could be a further development to satisfy the application, if the demand grows.

Radar from Vega Controls has already been applied to tide height measurements on the Humber Estuary, it just needs scaling down a bit!.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: