ΣSigmaStudio Ask AI Русский Log in

Parallel analog I/O for better performance?

SigmaDSP & SigmaStudio chmed Feb 26, 2017 00:00 2 replies Original on forum ↗
Subscribe
chmed Question
Feb 26, 2017 00:00

This is primarily with regard to the ADAU1701. Some of the high-end A/D convertors have an option to combine a pair of ADC together to improve SN and/or distortion performance and it has me wondering if the same would hold true for an all-in-one bug like the 1701. If I recall correctly from one of the datasheets, the improvement in SN is on the order of almost 3db over a single ADC which sounds like the self noise is basically random between the two channels. If the noise has some correlation (ps noise etc) then we would expect the advantage would be smaller. Unless! the two inputs/outputs were used in a differential configuration. It seems that this would be an ideal thing to do if there are inputs or outputs that would otherwise go unused. So, my question is if there is a good probability that performance could improve by paralleling inputs and/or outputs?

Audio Signal Processors

parallel

SigmaDSP Audio Processors

sn

adau1701

KJBob Reply
Feb 27, 2017 00:00

Hello chmed,

Likely the theory providing a 3dB improvement is correct. To make sure I measured the noise from the project shown below with no input. The Signal Merger includes a 6 dB loss to make up for the two DACs worth of signal, providing a direct comparison of noise levels. The 60 dB gain brings the noise up to a measurable level, and I included a homemade A-Weighting filter to see if it makes any difference.

dB ref to full output (60 dB gain accounted for):
Flat A-Weighted
One ADC -99 -104
Two ADCs -101 -107
Differential -101 -107

You may also wish to use two DACs to feed each following amplifier differentially -- this helps minimize distortion caused by the power amplifiers' ground return currents (which must be carefully accomodated in any case).

Best regards,

Bob

0a7a492f8c80a97d62d7a83e5372a8b8.png
0a7a492f8c80a97d62d7a83e5372a8b8.png
0a7a492f8c80a97d62d7a83e5372a8b8.png-620x259.png
0a7a492f8c80a97d62d7a83e5372a8b8.png-620x259.png
chmed Reply
Mar 8, 2017 00:00

As usual, KJBob, thank you for working on this! I recognize that it isn't a fully exhaustive test but I think it confirms the idea. I have a project that really only needs one output and I may try to mix them all together with some small value resistors instead of just using one output.