I'm using a MS5803-15BA (datasheet) to make pressure measurements. The datasheet contains this graph of pressure error vs. pressure at page 17.

enter image description here

The datasheet also specifies on page 1 that the accuracy when reading from 0 to 6 bar and operating from 0°C to +40°C is +- 20 mbar.

1) Does this means that if I'm using the sensor at 25°C, to read a pressure of around 4000 mbar, my reading will be something close to 10 mbar more than the real value? Or if I'm using the sensor to read a pressure around 12000 mbar also at 25°C, my reading should be corrected to 40 mbar more than observed pressure?

2) If this is the case, will it worth to make a script that uses the information in this graph to compensate the reading and yield a more accurate pressure measurement? Something like, if sensor is reading 2000 mbar at 25°C, subtract 5 bar to find a more accurate pressure measurement. Is there a way to obtain the data used to make this plot in order to code this script?

3) In case question 2 answer is also positive, how can I assess the accuracy of my reading after that compensation?

  • \$\begingroup\$ The chart shows typical figures; there is no guarantee associated with that. If you want better accuracy then you will have to calibrate the sensors individually. And with a typical long term stability of -20 mbar/yr you may need to calibrate several times a year if the application is that critical. \$\endgroup\$ Jun 10 '17 at 21:45
  • \$\begingroup\$ this says it has 2nd order compensation enabled. IF you recall how to convert data to a 3rd order equation or find a program to do it then you can correct with coefficients for 3rd order error. And it that is not good enough try a 5th order equation then error correct. 3rd order correction is common in TCXO's to compensate AT crystals with temperature which you can buy for a couple bucks in volume and not all the R&D costs. I have done this for <1ppm -40~70'C using a temp sensor and varicap. for <$1 with code. You have find your own solution and calibrate \$\endgroup\$ Jun 11 '17 at 1:22

Your main problem is that the graph is headed 'Typical'. There are no guarrantees associated with this graph.

The fact that the pressure output varies by more than 20mBar over temperature at zero pressure ought to alert you to the fact that there are things (like temperature) that will affect its accuracy. These will include excitation voltage, age, and mechanical errors like hysteresis, retrace and stiction.

This means the pressure sensor itself is simply not stable enough to be worth attempting to correct to better than 10 or a few tens of mBar, at least not for absolute pressure, and not over a long time.

Establish what the stiction error (minimum change of pressure required to change the output, after a previous change in the same direction), hysteresis error (minimum change of pressure required to change the output after a previous change in the opposite direction), and retrace error (difference in output at a pressure of p0, after visiting a pressure of p1) before you decide whether it's worth the effort to try to calibrate out the remaining few tens of mBar error.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.