Close

Error sources

A project log for No battery NFC air pressure sensor

Measure tire or ball pressure with your smartphone

captmcallisterCaptMcAllister 08/06/2015 at 02:150 Comments

August 5, 2015: I've done a bit of thinking about error sources, and I think the few error sources that must be compensated can be done in software when the micro is programmed. A bit more analysis below:

First up is temperature. The NPP-301 has a temperature coefficient of 0.2% of the full scale reading per degree C. For the 700 kPa (101.5 psi) sensor, this is 0.2 psi per degree C. Assuming people won't want to be riding a bike in temperatures below -20C, that means the worst case temperature delta is about 45C, leading to a pressure inaccuracy of 9 psi. That's probably enough to ignore, but it can also be easily compensated using the internal temperature sensor inside the microcontroller. No hardware compensation should be necessary.

Next, the offset voltage. It's +/-10 mV/V. I figure the micro power supply will be about 2 V, so the offset will be +/-20 mV. This is a third of the full scale reading! The good news is that it should be pretty easy to compensate. When the device is first programmed, we can assume it is at about 14.7 psi, and the offset could be zeroed out in software without the use of an external pot. This isn't entirely accurate though, because the sensitivity spec is wide enough to drive a bus through...

Now the sensitivity is specified as 60 mV +/- 20 mV for the full scale pressure. That means 100 psi could show up as 40 mV or 80 mV. That's quite a range, and it's because these pressure sensors are not calibrated (that's why they're so cheap). Calibrated pressure sensors are generally more expensive, but also larger, which is very bad for a tag intended to go in a tire. To calibrate this out, two pressures will need to be measured at the time of programming. This isn't too hard to accomplish, but it is a bit of a pain. A pressure cell with a bike pump could do it. Fortunately, the micro is also programmable through RFID - it may be possible to put the completed tag in a sealed box and write the coefficients wirelessly in an automated process. Either way, there's still no need for extra hardware for compensation - the micro can take care of it all.

All the other variations are less than 0.2% FS, and therefore seem relatively minor compared to these factors. They will not be compensated. The amount of error is definitely allowable for this application.

Discussions