inter-channel phase difference in MOTU 24 Ai

[ISchnell]

Hi,
I recently got some MOTU 24 Ai devices and I am now in the process of testing and trying out. I came across an unexpected behavior and I hope that somebody here in the forum could shed some light on the possible reason.

Test setup: sine signal generator output fed to two random channels out of 24. Inputs of both channels were bridged ( electrical connection between both + and both -, respectively). Device connected via usb as class-compliant sound card. Default routing 1-24 -> computer 1-24, no mixing, all trims at 0
Sampling freq. 48 kHz, internal clock.

Expectation: both digital signals received by the computer are exactly the same, no phase or level difference

Outcome: considerable phase difference as large as 20-30 degrees at 1 kHz, or up to 60 degree at 10 kHz, but this is different depending on which two channels we look at. This could be reproduced for 5 different MOTU 24 Ai devices, and also between devices when they are routed via AVB.

I am fully aware that for some applications this phase difference does not matter at all. Unfortunately, for the microphone array application I want to use it for, this is not tolerable. I order to come up with a compensation strategy, I would like to understand if this is due to tolerances in the analog AA filters or due to some strange synchronization problem between the AD converters. Has anybody some insight here?

Many thanks in advance!

[EMRR]

Well that’s disconcerting! I haven’t noticed similar with the older 16A, will have to look again. Initial tests looked at phase and frequency response and I recall pairs locked pretty well. There were small discrepancies in amplitude.

One recording artifact has been some sort of runtime difference in DP, don’t know if it’s DP or AVB or the 16A. The previous PCIe systems created equal length files in DP7. The 16A in DP8-11 have small differences in length at the end of the files. Could be a clue there…or not.