nRF52811 ADC resolution and enob

[JsSenso]

Hi,

I am developing a custom board with nRF52811 where I will measure analog signals from op amp on ADC (single ended mode). Signals will be very small, so I need a good resolution. nF52811 and the board will operate on 1.9V supply, so I will need 12-bit resolution on ADC. This was actually the main reason for choosing nRF52 chip in the first place.

According to datasheet (SAADC Electrical Specification, page 281-282) the effective resolution (ENOB) is only 9bits, which is to low for me. There was already a similar topic on DevZone nRF52 ADC ENOB, but I still need some more clarifications regarding this topic.

• According to datasheet 9-bits ENOB is for differential mode with 1/1 gain. What will be the ENOB for single ended mode and gain 1/3?
• It was mentioned that the best ENOB for 12-bit can be 10.5bits (with oversampling), which is not much better for me compared with 10bit. What is ENOB for 10-bits (this parameter is missing in datasheet)?
• What is VOS for 12-bits (this parameter is also missing in datasheet)?

Thanks for any help, Jure

[JsSenso]

Hi,

can someone from the Nordic Tech Support team please provide the missing data noted above.

And also, I really need an information if it is possible to achieve 12-bit (effective) resolution on nRF52811 adc.

Thanks, Jure

[Jørgen Holmefjord]

Hi,

Sorry for the slow response. Here is the feedback from our designers:

1. We don't have numbers from measurement. But from simulation, We do not expect any degredation
2. For 10-bit resolution the ENOB will always be less than 10-bit, since there are other noise sources than quantization noise
3. A change in resolution is effectly a left-shift, so when you go from 10-bit to 12-bit, then you can multiply the offset numbers in the PS by 4. 

You will not achieve 12-bit ENOB with 12-bit resolution, that is not possible. 

You mention, however, that you're looking at very small signals from an op-amp. I would strongly suggest you look into some differential measurement if your input signal is small.

If however, you have a small signal with a very large unwanted component (for example 50/60 Hz), like an changing common mode, I would still investigate whether it's possible to low-pass filter those, and feed a differential signal to the ADC. For example.

Positive input: Large unwanted signal + high frequency wanted signal

Negative input: Low pass filtered, and thus only contains large unwanted signal

This way, the ADC only sees the wanted signal, although the common mode varies greatly.

If this is possible, then this would allow you to use large gain in the ADC, and thus you would achieve down to < mV precision.

 Best regards,
Jørgen

[JsSenso]

Hi,

thank you for the reply.

The understanding is much better now. Basically one cannot achieve better ENOB than 10.5-bit. I will try to include as much oversampling as possible and then see If I can achieve something better. I also assume that with 14-bit resolution things are exactly the same, because this is just a different format but the adc is still 12-bit?

The signal I am measuring is actually not small. It is amplified (large gain used) by an op amp to achieve the whole possible voltage range on adc (1V8 with 1V9 power supply). This signal is static or very slowly changing and I also don't have problems with noise or other unwanted signals. The only problem is that even after amplification, the changes in signal are still small (~1mV on ADC), which I need to detect. Basically I need to be able to effectively detect each change of 1mV on ADC. And this can be challenging, especially because ADC itself is generating some internal noise, which is I guess larger then 1mV (when using 10.5-bit).

Best regards, Jure

[Jørgen Holmefjord]

JsSenso said:

I also assume that with 14-bit resolution things are exactly the same, because this is just a different format but the adc is still 12-bit?

Yes, from the feature list:

• 8/10/12-bit resolution, 14-bit resolution with oversampling

When doing oversampling, you can have averaging of n 14-bit values, where the 2 LSB are zero.