ADC resolution 12 vs 16 vs 24 bit

[Kevin A]

Posted on August 09, 2017 at 02:36

Regarding ADC resolution, 12 vs 16 vs 24 bit, is this software or hardware dependable? 

Looking at STM32F37x family, seems like they have an 'effective number of bits/ENOB' of 14 bits, using sigma delta ADC they advertise it as 16 bit ADC? but able to oversample to achieve 24 bit?  The reason I'm asking is because I'm debating between STM32373C-EVAL and STM32H753I-EVAL, the later doesn't mention about the ADC resolution? can I just assume I can get 24 bit resolution from it?  My application is acquiring a signal in microVolts with maximum sampling frequency of 2KHz. 

Lastly, is there even such a thing as true hardware 24 bit ADC? Texas instrument is advertising their ADS1299 chip as 24 bit, but doesn't mention over sampling? is that just a marketing scheme or a true 'hardware' 24-bit?  

##stm32h753i-eval-# ##stm32h753i-#stm32373c-#adc-#biosignal-#resolution

[Tomas DRESLER]

Posted on August 09, 2017 at 10:47

Hi, F373 has 16-bit sigma-delta and you can extract 16-bit data.

ENOB is a mathematical concept explaining how many bits are actually usable due to inherent noise (effects of quantization of input voltage etc.)

H7xx has 16-bit SAR ADC with possibility to oversample (or filter) the result further.

You can think of L4 with DFSDM and external modulator for more precise conversion, too.

Beware that anything around the MCUs will get its share of noise due to HF operation and communications inside and outside of the chip, so separate ADCs for high precision, shielding and perfect PCB design are the minimum.

[S.Ma]

Posted on August 09, 2017 at 13:52

If looking at microvolts, sounds like a weatson bridge similar to pressure gauges.

There is a quick overview of ADC there:

https://community.st.com/0D50X00009XkW3SSAV

If an instrumental ADC is required, depending on the signal amplitude, required resolution, noise from environment, thermal noise, etc... will tell what is needed or not. Using a low noise external op-amp to amplify/shift the signal maybe necessary. Running differential would probably help remove environmental parameters to the first order. Some homework is needed to know if it may fit an embedded MCU ADC and which type. Careful analog signal management falls into signal integrity field.

For each additional oversampled ADC bit you might increase the time (samples) by the square/exponential, so it might hit a limit pretty quick. Digital filtering may help there if the noise is well characterized.

Sounds like some additional homework digging needed to conclude.

[Kevin A]

Posted on August 10, 2017 at 03:01

yes im looking into 24 bit ADC from analog devices or maxim, unlike TI they seem to have all their API and software openly available, which is what I need given I have zero hardware code experience. Also shielding are you refering to active or passive? I was reading on active shielding which seems to produce great results, based on faraday cage shielding concept? seems like just a wire with outter metal mesh shield that you run electricity through?