Problem of internal voltage reference

[Vyacheslav Azarov]

Posted on January 26, 2016 at 00:27

Voltage reference source in STM32F334 very depend from temperature. Internal temperature sensor can help compensate this. But not known the real value of the temperature coefficient of reference source. Voltage reference has calibrate at 30 C only. This make impossible mathematically compensate drift, without additional testing. And makes useless the internal reference. Sorry.

#vref #vref #vref

[waclawek.jan]

Posted on January 26, 2016 at 07:01

Please quantify.

Is the voltage reference's temperature drift more than the 10mV quoted in the datasheet?

JW

[Vyacheslav Azarov]

Posted on January 26, 2016 at 11:06

My application require 0.5% accuracy of measurement.

The value 10 mV is 0.83% from reference voltage, it's a lot. 

Total accuracy of ADC is only 0.12% from measurand, it suits.

Guaranty 100ppm/C is very great for precision reference and it not suits.

The real coefficient is unknown, what vary at production.

Therefore correction that based on temperature dependency is impossible.

Unless only you measure it yourself.

[waclawek.jan]

Posted on January 26, 2016 at 19:27

I understand your frustration, but unless you present a significant buying power it's unlikely ST will do anything about it.

You have probably considered external precision reference, haven't you?

JW

[Vyacheslav Azarov]

Posted on January 27, 2016 at 09:35

Yes. The simplest solution is to use an external reference source of voltage.

However, I would like to reduce cost in production. Our plans have small production volumes. And additional calibration of each chip in the production of undesirable. I would like to recommend the developers to add a second calibration point for the reference voltage, for example at 110 ° C, in the following implementations of this product as it is made for the temperature sensor.  If impossible make the temperature drift compensation in hardware.  In rest the STM32F334xx excellent product. Thanks.