OPV and low reduction levels

[archaeo]

I'm noticing that my OPV results are the most problematic on low-end reductions, those above 90%. It seems to be true for either CCE or HC. Is there any explanation for this that correlates a reduction level to output accuracy? I prefer to use only one pass since a two pass seems like overkill, and I’d rather use an encoder than a transcoder as the norm. But these low reductions are where I see the most variation. My experiences seem to show that the larger the reduction, the more accurate the OPV output. Anybody else notice this?

[pg55555]

Its called the granularity of OPV.

As OPV can just acept integers for Q, you get that for low reduction levels you need low Qs, and the change to the closer integer is proportionally larger..

Say you use a Q of 10 and you get an output of 4.40 G (a 0.7% oversize). So you increase Q to 11, a 10% increase, and assuming a linear behabiour (which it is not true, but to simplify), the size of the output is now 10% lower, ie 3.99 G = a 9.4% undersize vs DVD5.

If the original Q has been 5 (Lower compression) , when increased to 6 it is a 20% change, the output would have changed to 3.52, a18% undersize.

But if the original Q have been 20 (higher compression), changing it to 21 is just a 5%, so the output would be 4.18 G, just a 4.3% undersize.

[archaeo]

OK, so that's granularity. Thank you for the excellent explanation pg55555. So as the integers (Q) get smaller, there's less room to 'play' with reduction percentages. I assume that these low reductions also create difficulties in trying to tweak sector size for a better fit.