ADC offset error calibration

[shuto kim]

Hi,

We are evaluating ADC functionality.(We use nrf52832 on a device with sdk 16.)

The evaluation was performed under the conditions with and without offset calibration (temperature 6 conditions), but the error was reduced under the conditions without offset calibration.

Offset calibration was performed only once when the power was turned on, and the power was turned on again at each temperature.

This result suggests that (1) it is not offset calibrated, or (2) it is offset calibrated, but an error is added.

About (1)
Please tell me how to calibrate the correct offset.
For example, is there a specification for the timing of offset calibration? How many seconds should be applied after supplying power to the microcomputer, or how many seconds should be applied.

About (2)
The error with offset calibration was stated as ± 2LSB, but what is the error without offset calibration? Also, is it possible that the overall error including the gain error will be larger if offset calibration is applied?

Best regards,

shuto

[Jared]

Hi,

There is indeed instances where the offset calibration might worsen results versus not using the calibration. This is often in instances where the SAADC is going to operate over a large temperature range. There is mainly two reasons for this:

1.  We know that the offset changes with temperature, so it might actually be very close to a tripping point between two DAC values.

2. 2) We know that there is noise in the system (inherent thermal noise) which introduces sufficient randomness to flip the calibration values back and forth.

The error with offset calibration was stated as ± 2LSB, but what is the error without offset calibration? Also, is it possible that the overall error including the gain error will be larger if offset calibration is applied?

It will not affect the gain error. If you see that the samples are better without offset calibration, then you should not use it. 

regards

Jared 

[shuto kim]

Thank you for your reply.

Could you give us a detailed explanation of (1)?

Or do you think we're offset calibrating correctly, even though we have a lot of variability?

[Jared]

Hi,

I think you're calibrating correctly, what you're observing is an effect of the calibration routine using an ADC with only 10 bit resolution. In your case it seems not using calibration would be better.

See this thread for more. 

regards

Jared