Calibration with 0,00% deviation? – Too good to be true!
Is a calibration report a reliable statement about the accuracy of measuring devices? – Most of us would like to answer this question with a resounding “Yes, of course!”.
This is why we end up removing our instrument from the process, incurring the cost of a replacement device or the downtime, sending it to a reputable calibration laboratory and paying a not insignificant amount of money to get it back with a new certificate, freshly calibrated and readjusted if necessary so that it now provides reliable readings within its specification.
Unfortunately, in our day-to-day business as an accredited calibration laboratory for flow measurement technology, we have had first-hand experience of how wrong this statement can be. Unfortunately, an official stamp often leads to the meaning of some rather strange calibration certificates not being questioned, partly due to a lack of knowledge on the part of the customer, partly – let me put it this way – due to certificates being issued that leave a lot of room for misinterpretation.
This may be acceptable for customers who use their measuring instruments in less critical applications and just want to complete their internal documentation. In this case it is also a pity for their expenses, but at least the deviations of the measurement results do not significantly affect their processes.
But there are critical applications where the development of a new product relies on accurate measurement results. For example, flow meters are used to control fuel consumption, optimize emissions or precisely dose expensive flavors into food or health-critical medical ingredients into medicines. Such customers are not helped by a nice certificate.
The following example is from such a sector. And this has prompted us to write an article to raise awareness of this issue.
Correct calibration should provide meaningful information about the performance of the tested instrument. It requires technical know-how, the necessary equipment and reliable handling.
“When purchasing a measuring device, trained and experienced engineers often invite several manufacturers to present their products and solutions. They weigh them up and test them carefully before deciding which equipment is best for their application,” says Costel Hanea, Calibration Engineer from TrigasFI Calibration Laboratories. “But when it comes to calibration, hardly anyone checks who and how they give their equipment to. We often receive devices in a package with a short note “Please calibrate”. If we then ask the important questions about process conditions, media, environment, installation, etc., then the device is sent to another laboratory for the next time, which doesn’t ask so many uncomfortable questions, but sends the measuring device back together with a nice “calibration certificate”. This is scary at times considering the applications these flow meters are used in. “
For example, one customer sent us a highly accurate turbine flow meter with a request to “calibrate to 0.01% accuracy” as it had apparently been done by another service provider during the previous calibration cycle.
“TrigasFI® is DAkkS / ISO 17025 accredited for the calibration of liquids with an uncertainty budget of 0.04%. In fact, our own uncertainty budget supports 0.03%, one of the best accuracies in Europe. An uncertainty of 0.01% would be a major challenge even for the best equipped national laboratories, such as PTB in Germany,” explains Hanea.
“However, there was another problem that made us suspicious. We know these types of flow meters usually can’t be much better than 0.1%. Even this accuracy can only be achieved in the application if the calibration is performed strictly under the same viscosity conditions that exist during actual operation, or if a certain viscosity range is calibrated and later compensated.
Therefore, we asked for the customer’s application data and the last calibration certificate to see what was done there to achieve such brilliant results… “
Here are some excerpts from this amazing certificate:
We were initially surprised to see in the header data that no measurement uncertainty was shown for the specified piston calibrator (MT50), which was given as a reference. The uncertainty for the calibrator or the master device should always be listed in a certificate, as it is crucial for the validity of the calibration. So why was it not stated? (The expected calibration uncertainty of such a piston calibrator is usually between 0.03 and 0.05% of the measured value).
Instead, a frequency generator and a voltmeter were listed with uncertainties of +/- 10 ppm / year and +/- 0.01% respectively, which are not normally required for a flow calibration. Obviously, however, it was implied here that the measurement uncertainty of the flow meter depends on these two devices, which is not the case.
The question now arose as to how this calibration data could be recorded at all. Was the calibrator used at all if no uncertainty was specified for it? And what were the frequency generator and voltmeter used for?
Let’s look at a normal set-up of a flow calibration:
The picture shows the calibration of a measuring chain of flow meters together with its evaluation unit (electronic signal conditioning / compensation unit) on one of the calibrators mentioned.
This table shows the test report, which is labeled “Measurement results”.
The first two columns obviously show calibration data: Flow rate in l/min, as it could have been measured by a calibrator, and the resulting frequency (Hz) of the tested turbine. However, this flow data was used to program it into the TLSV500 linearization electronics, with corrections for each point so that the 0…10 V analog output was accurately scaled to 0…10 lpm.
And this is where the connection with the actual flow measurement ends. The remaining tests were obviously carried out without the flow meter – and therefore also without taking into account the decisive influences of mechanical and thermal effects on this sensor in reality.
Instead, a signal frequency was fed into the TLSV electronic unit using the frequency generator mentioned above. The values of this frequency were set so that they exactly matched the frequencies recorded during the original flow calibration and which had been programmed into the electronic signal conditioner at the time. And because the frequencies corresponded exactly to the cardinal points programmed into the TLSV electronics, the analog output naturally showed exactly the corresponding values in V.
Electronics programmed in this way and fed with a frequency of 8.850 Hz will naturally generate an output of 0.220 lpm, within the resolution accuracy of the built-in digital-to-analog converter.
And indeed, the displayed deviations of 0.00 to 0.01% from the final value are to be expected with a 13 -14 bit D-A converter. This conversion does not exist for the frequency output, so that the deviation here is 0.00% throughout.
However, the way the data is presented in this certificate implies that 0.01% is the uncertainty of the calibration point, which is clearly not the case.The data displayed is nothing more than a functional test of the electronic evaluation unit without taking into account the actual performance of the flowmeter.This is what a setup of such a test looks like:
As if that were not enough, a closer look reveals further questionable information. For example, the deviation of this measuring device is specified in % of the final value (EW) instead of % of the measured value (MW). The specification of the accuracy in % of EW is common in specifications of flow meters that have a lower accuracy; for high-precision turbine flow meters, they are very unusual. In addition, they are not relevant for a calibration, as calibration points and their deviation (from the measured value) are to be recorded here. The indication of the results from the final value is obviously an attempt to make the results appear better. In fact, even the values recorded in the manner described above should have looked like this:
Flow rate: 0.220 l / min
Expected voltage: 0.220 V
Actual voltage read: 0.219 V
Deviation: 0.001 V
% error: 0.001 / 0.220 = 0.45% of MW (of the measured value)
However, the report gives values between 0.00 and 0.01% of the final value, while the actual measurement uncertainty is 45 times higher. Furthermore, the specification of the viscosity range at which the calibration was supposedly carried out raises questions (of which no results are available here, but which should have been listed individually in order to properly determine the programming values for the sensor).
It can be assumed that the actual flow calibration was carried out with the specified calibrator at ambient temperature with a calibration medium of approx. 75 mm²/s +/- 10%. (The viscosity variability of +/- 10% is due to the fact that the temperature is not controlled during the calibration process).
The statement about 40°C therefore does not refer to the actual flow calibration, but to a temperature simulation.
“Now we were curious to see what the actual accuracy of such measuring equipment was and how it would correspond to the certificate provided,” recalls Mr. Hanea. “To illustrate this, we calibrated an identical flow meter once without and then with temperature compensation as a measuring chain (flow meter and electronics together)”.
The red line shows the unrealistic performance curve (red line) generated by the frequency injection.
It can be seen from the results that despite the values of 0.01% or better shown, the customer’s expectation of such a measurement uncertainty cannot be fulfilled in practical application and under real conditions of temperature and viscosity influence.
The statement of the present certificate, which only concerns a test of the D-A converter of the evaluation electronics, will hopefully not be interpreted by the customer as an uncertainty of his flow measuring device. Misleadingly, however, this report is titled “Calibration certificate”. And even the values of the electronics test should actually have been shown correctly at approx. 0.45% of the measured value.
Summary:
Customers who have high expectations of their flowmeters intended for use in critical applications should insist on transparent, unambiguous calibration certificates with data relevant to their process requirements. Such certificates must clearly identify the calibrator used with its uncertainty of measurement to ensure that it is fit for purpose.
Calibrating only a part of the measurement chain (either only the calibration of a flow meter or only a simulation test of the electronic amplifier, as we have in this case) does not provide a complete picture of the accuracy of the measurement chain. The correct way to establish a comprehensive uncertainty budget for the measurement process is to calibrate the entire measurement chain under conditions as close as possible to those of actual operation.
“The results may not be as brilliant, but they are relevant. And only if we know how high the deviation of a flow meter really is can we try to take the relevant factors into account and compensate for them,” says Costel Hanea with a twinkle in his eye. “It’s like in real life, we should always be careful when something is too good to be true.”
Erdinger Straße 2b
D-85375 Neufahrn
Germany
Telefon: +49 8165 6472-0
Fax: +49 8165 6472-29