ADC128S102QML-SP: AtoD reference noise. Can they cause spurious outputs/sparkle codes on this part?

Hello Michael,

Initial thoughts are that we're going to find this is related to external components around the ADC and not related to the GIDEP sparkle code advisory.  Would you confirm more about the spurious disturbances in the outputs?  To answer the question about the reference directly, the AVDD voltage on the ADC is directly used as the reference voltage for the conversion.  Since the output code is Vin/(VREF/(2^12)), if there are variations on VREF (AVDD), they will directly impact the output code.  

Are there ceramic X7R type dielectric ceramic capacitors on the board in the signal chain?  If so, they're likely the cause of the trouble. Here are two good articles that describe the effect. 

https://e2e.ti.com/blogs_/archives/b/precisionhub/posts/stress-induced-outbursts-microphonics-in-ceramic-capacitors-part-1

https://e2e.ti.com/blogs_/archives/b/precisionhub/posts/stress-induced-outbursts-microphonics-in-ceramic-capacitors-part-2

Yes we have X7R capacitors on the board and on the AtoD supply. And I am aware they can cause voltage fluctuation in response to shock. However that effect on signal inputs to AtoD and AtoD output based on Vin/(VREF/(2^12)) should still be smaller than error I am looking for. Trying to explain a >1000 count jump (inferred from an error indicator in the logic) that happened once, though we can see some smaller disturbances when test repeated. As this AtoD can produce sparkle codes at certain boundaries I wondered if noise spike on VREF (from the X7R cap response) could cause corrupt output  (or sparkle code) if ref changed too much during conversion at other boundaries.

Mike   

Hi Michael,

If AVDD has voltage spikes on it during an active conversion it's hard to directly confirm what could happen as the charge injection and other things involved in the conversion process will likely be disturbed. Further, if the capacitors are in the ADC input signal chain, then gain and higher impedances around the caps could create larger voltages than expected. 

We can probably rule out the sparkle code if we examine the average codes you expected the output to have along with the value it jumped to.  Understood if this isn't possible if the effect was only observed, not logged.  If you have the information, would you share?  Also, if you're willing to share the schematic I could help review the analog signal chain for anything else that could be helpful.  We can start a private message and I'll share my e-mail so we can send the information privately.