EVM430-FR6043: USS calibration between fluids with different SoS

Thanks for the reply.

There's a range of speeds of sound that we can expect in the field. It represents an important design constraint for us to be able to support different fluids in the field without having to replace a perfectly functional sensor or re-calibrate a sensor in-place. It's also not particularly practical for us to have to manage a portfolio of different SKUs and hardware for each type of fluid (which we may not even know when the sensors ship). It is also an expected use case that the fluids change -- both with respect to temperature and to their speeds of sound. 

For calibration, we start with an average USS_VOLUME_SCALE_FACTOR (670,000) and adjust per-sensor by comparing USS_Algorithms_Results.volumeFlowRate from USS_runAlgorithms to the flow rate from our reference meter (Coriolis). For example, if USS observes 3.9 LPM and the Coriolis observes 4.0, we adjust to 670,000 * (4.0 / 3.9) = 687,179. We are not currently using multiple calibration points.

With respect to the other questions, considering the physics, it appears to be algebraically possible to create a multiplicative correction factor accounting for the new speed of sound, specifically 

where 

• ToF[total] is the average (including up and down) absolute ToF (full path)
• ToF[solid] is the non-fluid ToF (transducer mounts, pipe wall)
• cal indicates the value seen at calibration time
• current indicates the value seen at measurement time

If we know the meter constant for a specific SoS, if the SoS changes, it seems a rather straightforward exercise to adjust the reported velocity (and subsequent VFR) accordingly. Given that the sensor and its algorithms operate based on flight time and arithmetic derivatives (e.g., differences), it would seem that such a compensation should work. Where is the flaw in this logic? The empirical results suggest there is one.

What kind of accuracy are you trying to achieve? This is a critical input to your query.

The only true way to get speed of sound is to measure it with a known distance. Unfortunately, thermal expansion/contraction causes this previously "known" dimension (and your cross-sectional area) to change over temperature. Pressure can also affect these as well if your mechanical design is not very rigid. Not only that, do you need to linearize/calibrate more than one point or is a single Meter Constant sufficient? If linearization is required, another thing to consider is that fluids of different viscosities will react differently through your meter body at the same VFRs.

The TI solution (and support structure) is good for water or a single fluid which can be characterized and corrected for. As soon as you need more than one correction curve implemented based on fluid, or get into deep fluid dynamics questions, you will have to implement your own enhancements to the calculations.

Thanks for the reply.

2% to start. In real deployment scenarios, changes due to SoS are >> than those due to temperature. Temperature is relatively stable.

It seems I am missing something fundamental with either the physics or the USS library, as accounting for SoS changes (assuming fixed geometry as a close approximation) should be straightforward... but it isn't in our case.

Agreed with Seth's comment. 

Even on a water meter, it usually requires multi points of calibration. For a better understanding of USS library, I would suggest you to check with this app note. https://www.ti.com/lit/pdf/slaaes9