How to identify with oscilloscope noise or ripple that affects ADC accuracy

[Red_Micro]

The circuit attached is a 1.65V buffer that has the following purposes:
1- It provides a 1.65V DC offset to four op amps to measure AC signal with STM32 ADC.
2- It is directly sampled by the MCU also.
3- It provides a reference for other op-amp circuits

Both circuits work, but the one with the shunt reference makes the AC measurements more accurate, thus ADC measurements more accurate. So it seems that there's some factor around the 3.3V supply that it's eliminated by the shunt before going through the buffer. The buffer seems quite robust so I thought the shunt would not be needed. You can see that in the circuit with the shunt reference I had to change the upper resistor of the divider to 5.1K to keep the 1.65V since the shunt provides 2.5V.

I tried to watch with the scope what's happening around the 3.3V and 1.65V points but could not catch anything. I set the probe to 1x, AC coupling, 5mV division but did not see anything other than the scope noise floor. I even got rid of the scope ground clip and got the ground loop as short as possible. Either my scope is not suited to see such things or I'm doing something wrong. At the end I want to explain myself what is the shunt reference eliminating that makes the ADC measurements more accurate.

[David Hess]

It is very likely that the noise level in either circuit is lower than a typical oscilloscope can measure.

[harerod]

What David said.
In addition:
- you supply the OPAMP somehow? Interesting CAD system, I am curious about the layout.
- Some STM32 ADCs show several bits of periodic noise from other subsystems (e.g. ART on STM32F407). Best build-in ADC performance can only be reached by disabling those subsystems during conversion.