PCR Accuracy in mux'ed output

[gconnery]

Assuming that you create CBR transport streams as output (it sure looks like it), the accuracy of your PCR calculations leaves something to be desired.  On one sample I ran the rate computed from PCR to PCR varies from

Min Delta Rate:        16847720 (-29984)
Max Delta Rate:        16907891 (+30187)

The rate determined by subtracting the last PCR from the first is 16,877,704 while the bit rate I compute after looking at the first MB is 16,877,805.  These differ by 101bps, which seems excessive (its outside MPEG-2 spec tolerance anyway).

You should fix this.

[Vovs]

We know about such PCR behavior and are fixing this. But we are not sure that rate which can be compute from first MB should be equal with the rate determined by subtracting the last PCR from the first.  

Thank you for thoughtful observation that is really helpful for us.

[gconnery]

I just have a utility that looks at the rate by computing it from the first PCR and then looking at subsequent ones until it appears to have converged, meaning the rate from the first to the Nth PCR isn't changing anymore.  That rate typically converges within the first MByte if the PCR accuracy is any good at all.

[Vovs]

It seems you are right. I've investigated iso 13818-1 again and found that the PCR behaviour should be the same like you described.

Certainly this should be fixed.