Accuracy of timestamps of acquired data

[lowestpass]

I have several DAQmx tasks running asynchronously, and I would like to be able to accurately align their data based on timestamps afterwards (for my purposes, alignment within 1 ms would be good enough). I know from this article that the waveform timestamps will not be perfect, but it makes it sound like on a PXI that sub ms accuracy should be achievable. But in my testing so far I seem to be doing much worse than this.

For my current test I have a PXIe-6375 sampling at 5 kHz and a PXIe-4464 sampling at 50 kHz. I also have a PXI-6683H timing card in the system with PXI_CLK10 disciplined to it. I am running a common pulse to both devices to compare timing. Comparing the waveforms, the 6375 is consistently about 10 ms ahead of the 4464. Realizing that the DAQmx timestamps may be flawed I also tried getting a timestamp using NI-Sync instead, and am getting very similar results. 

Can anyone explain why I would be getting this 10ms gap between the different tasks, and what I can do to get more accurate timestamps?

[ZYOng]

Did you compensate for the group delay from the DSA module?

Synchronized Data Delayed When Using NI DAQ Devices with Delta-Sigma-ADC

[lowestpass]

I hadn't been aware of this effect, so thank you for that link. I think what I'm seeing is something different though. First, following that documentation RemoveFilterDelay should default to true for the 4464. And even after explicitly setting it to true I'm still seeing the delay.

After continuing to play around with it with several different cards, I think it's related to differing sample rates. When the sample rates of the different tasks are the same, I get very good sync. But when I have different sample rates, it seems that the faster task ends up having a delay. So the DSA cards end up being before the regular AI cards if they're sampling slower. Would a faster sampling rate create a delay between the ADC and the buffer that isn't being accounted for?

[lowestpass]

Well, turns out my problem wasn't with my data at all. I had been using the tdms viewer Scout, and it appears to be a problem with that application. When I read the exact same TDMS data myself using the LabVIEW VIs my data was perfectly in sync. I guess this is a lesson in trusting 3rd party applications.