Extreme DAX sample rate weirdness (low-level)

Andrew Rodland

So I'll begin by saying that I'm using my own implementation of the TCP API and DAX for Linux to talk to my 6400M (you can see more about that in my post history). I don't have a Windows or macOS machine in the house (although I can run a Windows 10 VM as required). I've had a few bumps along the road, including dealing with timebase skew to make sure that my audio buffers don't underrun or overrun... but I thought I had things reasonably well sorted out.

However, today, I absolutely could not the audio to quit stuttering. I tried rebooting the Flex, downgrading SmartSDR on the Flex, upgrading SmartSDR on the Flex, using a different PC, using a different version of my DAX driver, and connecting with the Windows version of SmartSDR in hopes that it would clear out some hidden setting I don't know about — all to no effect.

So I decided to fire up wireshark and do a packet dump of the incoming VITA49 packets. Everything looks neat, there are no gaps in the sequence numbers, the packets come at a predictable interval... only problem is that interval is about 5743 microseconds when it should be 5333. During a span of 100 seconds, I received 17,308 packets (should be 18,750). In other words, the Flex is giving me about 44.31 ksps instead of 48k!

Why in the world is this happening? My PCs are synchronized to NTP stratum-1 using Chrony, my Flex is locked to a ppb-accurate 10MHz reference, there's no tuning error whatsoever on the Flex, and the audio out of the Flex's speaker is fine.

Andrew Rodland

I can provide a PCAP dump file if that would help at all.