Temperature affect some MSP430AFE253s’ Measurement Accuracy .

Hanson He said:

customer require the error ＜5%

I assume this should read <0.5% :)

Hanson He said:

We just exchange the MSP430AFE253 between Meter17(NG) with Meter14(PASS)

So it is a problem with the internal reference or the converter calibration.

However, the reference is +-5%, so 1% error is still within range. If you need more precise results, you'll have to calibrate the device by yourself, best to do it on different temperatures, so you can interpolate. It makes the calculations more complex, but pays. Especially if your operating condition is not the default 30° 3V VCC.

Sorry for my mistake, it is 0.5% accuacy.

Could you tell how to calibrate device in different temperature? Could you share calculation method? This is urgent case. Thx.

We test 130pcs, but just less than 10pcs could meet less than 0.5%, other devices more than 0.5%, some chips even more than 1%.

Why same package chips have different result?

Hanson He said:

Could you tell how to calibrate device in different temperature?

Basically, you need to heat the MSP to the operatign temperature (best tools is a climate room), apply a known voltage to the inputs and see the result. Then you can calculate a correction factor for the measured value to the real value.

Actually this is what your test tool already does. Just add the 'error' as additional factor into the "ADC reading -> input voltage" conversion formula.

Thx, but I have question, as you see above testing data, each MSP430AFE253 have different accuary result at different temperature on customer board, if every chip calibrate its value at different temperature, this is very large work load, if customer have more than 10,0000 devices, each need to calibrate, this is very hard for them,

Whether have other suggestion for it, many thx.

If the device operates on changing temperatures, then a two-point calibration is necessary. e.g. 30° and 80°. Between the two, a linear interpolation is usually sufficient, as you can then calculate a static offset and a (assumed linear) gain by temperature. Of yourse, then you'll need to measure the temperature too. Which too requries a two-point calibration and interpolation. But many MSPs have factory-provided calibration values. I'm not sure what is provided with the 253, but the 5438 already has calibration values for 30° and 80° for the reference voltage and the temperature sensor. So it comes down to only more math at runtime.

Yes, high precision is highly difficult (or very expensive, since someone else then has done the difficult part for you). :)
And 0,5% error is high-precision, definitely.

Jens-Michael:

      Many thanks, could you help confirm MSP430AFE253 could add factory-provided calibration value later?

      How to get MSP430F5438 calibration algorithm?

      I just make below method, if any suggestion for it?  Different temperature have different ADC output. So I think ADC output proportional to temperature vary. So it have below formula.

Hanson He said:

could you help confirm MSP430AFE253 could add factory-provided calibration value later?

???

Hanson He said:

Different temperature have different ADC output.

Obviously, when talking about the internal temperature sensor reading :)
Your formula is correct. However, both, temperature sensor and reference change with temperature. But if both are assumed to make a linear change, both are combined into gain.

However, when adding the change of gain (for external signals) based on temperature to the formula that calculates the voltage form the external signals ADC input, you may be able to pull some parts out of the formula, directly working with the known temperature channel reading minus temperatur offset value, a constant factor and apply this to the adc input value.

You may even just read the temperature once a while, and calculate a simple factor you multiply the ADC reading with, so you get the signal voltage in mV or so.
It's a lot of math, especially if you transform this into plain integer arithmetic with still a reasonable precision. But it pays.

For my ADC, I don't use the temperature drift of the reference. It's still precise enough for my applicaitons. So the fomula is simpler.