SV Sensor Accuracy and Error Budget

One of the cornerstones of Valeport’s philosophy is that our products should do exactly what we say they do. There are two reasons for this: firstly, it makes us more efficient – why would we want to put ourselves in a position where we have to waste time defending the indefensible, explaining the inexplicable, and justifying the unjustifiable to disappointed customers? Secondly, we find that telling the truth to our customers is a very good way of generating confidence in our products, and building the long-lasting relationships that are the focus of our business.

The quality of data you get from your Sound Velocity sensor underpins your whole survey, whether it is taken from a profile, a series of discrete samples, or a continuous input from a sensor at the transducer head. For that reason, we believe it is vital that you know exactly what errors your SV sensor is contributing to the system, and that such errors should be minimised. We therefore detail our Error Budget below, showing how we arrive at our stated specification:

Parameter

Note

Value

Precision

Clock Drift (max)

Calibration Equipment Error

Calibration Fit Error

Total

Stated Specification

1) We define Precision as Peak to Peak Signal Noise, i.e. under steady state conditions, every single reading from the sensor will be within a range of ±0.002m/s. What this means is that if the readings vary by more than this, then the Sound Velocity is changing; it is the Peak to Peak Precision which defines how good the sensor is at detecting variations in Sound Velocity. Beware specifications that quote standard deviation or RMS precision figures (or worse, do not tell you which method is used) – they are just a statistical manipulation that say most of the readings (63% of the readings in fact) will be within a certain range; they do not tell you what variations or noise you can actually expect to see from all your data.

2) We use a 1ppm clock, i.e. its performance over the operating range of the instrument will vary by a maximum of 1 part per million. In the context of a sound velocity sensor, this is a timing error of the order of 1/10th of a nanosecond, which equates to the SV error shown above. In contrast to Precision, this is an uncertainty in the overall accuracy, rather than an uncertainty in each individual reading.

3) Calibration equipment for Sound Velocity is basically pure water (<1ppm impurity – even the salt in a fingerprint on the sensor will disrupt it), baths in which the temperature can be controlled and maintained to better than ±0.001°C, and a highly accurate temperature measurement system. Calibration equipment errors take account of the accuracy of the temperature sensors and standards which are used in the calibration procedure. Again, to put this into context, we are measuring temperature to a total accuracy of around ±0.002°C during the calibration process, which equates to the sound speed figure given above.

4) We do not tolerate any errors in the mathematical derivation of the calibration equation from the data collected.

5) Our stated specification gives us a little headroom over what we can actually achieve; it is better to err on the side of caution than to state figures that are barely justifiable.
 
We would recommend that if you are considering an alternative to Valeport, you should ask for a similar breakdown of the error budget for comparison.