OPA2188-Q1: how to select a proper refence driver for the ADC ref

Hi Eric, 

Eric Liu(SZ-FAE) said:

can you help me on how to select a proper ref buffer for the ADC ?

If the customer is using bandgap reference, I would suggest to follow the table below. REF35, REF34 are reasonable cost. Depending on your accuracy requirements over temperature, REF35 or equivalent may be required. 

https://www.ti.com/lit/an/snaa320b/snaa320b.pdf?ts=1730160306059

https://www.ti.com/power-management/voltage-reference/products.html?keyMatch=voltage%20reference&tisearch=universal_search

Eric Liu(SZ-FAE) said:

what is the key concerns of this selection?

For op amp, such as OPA2188-Q1 or similar to use as reference voltage, the concerns are about driving capacitive load (reference stability), accuracy and temperature drift over time.  

The following technique applies to OPA350 as voltage reference with large capacitive load. If you want to use OPA2188-Q1 to drive a reference voltage, you have to tell me how much capacitive load, and the op amp has to be compensated in order to drive it. (op amp can not drive large capacitive load directly without oscillation concerns). 

OPA350 as Reference 10292024.TSC

If you have other questions, please let me know. 

Best,

Raymond

Hi Eric, 

Here is open loop AC analysis for OPA350. 

OPA350 capacitive load AC Anslysis 10302024.TSC

Eric Liu(SZ-FAE) said:

we can do a math calculation：sqrt（1+5*5 ）=5.099 ，so the vref noise take the propotion as 1/5.1=0.19 ,it is too big for the total noise

Regarding to the noise issue relating to ADC, please submit a query for the the ADC team. The HVA team is supporting op amp related application issues. 

Best,

Raymond

Adding to Ray's comments, if you use one of the series voltage references shown above in Table 2.1, you do NOT need to use any op amps; these references are designed to drive directly 1uF to 100uF capacitive loads. Using a resistive divider with a buffer to drive a REF pin of 16-bit ADS should not be attempted as the error from supply voltage variation and resistor mismatch would greatly degrade the precision of the system (perhaps down to 10-bits).

As far as your question "why the VREF noise must be at least 5 times smaller than the ADC noise", it is because in such case it may be neglected.  As you calculated above, in case the VREF noise is five times smaller (or more) than ADC noise, its contribution to the total noise is negligible - {[sq-rt(26) -5]/5}*100% < 2%