VRX Buffer Underflow

[ZeroDawn]

Hi,

I have been testing with 1080p59.94 video streams and our analyser was showing that the VRX buffer was underflowing especially when running more than one stream. I ran a pcap from the RxTxApp through the EBU LIST analyser and this also showed a non-compliant VRX buffer:

I noticed there was a LATENCY_COMPENSATION constant that has been tweaked occasionally, its currently set to 3, but reducing that to 1 does improve seem to improve it.

I've attached the json config here: tx_1v_1080p59.json

I can also provide the pcap if its useful.

Many Thanks,
Matt

[ZeroDawn]

I just wanted to add a bit more info here.

As mentioned above, adjusting the LATENCY_COMPENSATION did help somewhat in our test, but its my understanding that reducing that value means you see negative latency measurements. I tried to reproduce this and have some questions/observations.

I am running the RxTxApp without user_timestamps/user_pacing set.

With LATENCY_COMPENSATION set to 0, I see:

and with it set to 3:

I assume this is what you are observing in your tests?

One thing I did note is the RTP Offset value in these tests is measured at 622us:

In all of our testing, that offset is always zero and the latency is up at ~620us. (Fwiw, this is the same as what we see with a different, non-MTL based 2110 stream as well)

The same values are in EBU LIST analysis well. The RTP Offset is shown under the graphs section.

All this to say, I'm just wondering if the latency issues you are seeing are more to do with the fact that it seems the RTP timestamps aren't being aligned to the 'epoch aligned start time of the frame period'?