#180 Higher Precision with Speed.

[Radikal Den store]

The speed function in Sox, seems to be quite limited, and not as accurate then it probably can be.

For example:

150431808 samples with a ratio of 1.00048321 should become = 150504498.15394368 (but we will cut the decimals making it 150504498).

But, in Sox that will become 150359153
while in Avisynth, it will become: 150504493

Sox is very far from the precise number as you can see.

[robs]

The parameter to the speed effect is a ratio of speeds, not times.

If you have a ratio of times then you must take the reciprocal, e.g.twice the time = half the speed.

In your example, 1/1.00048321 = 0.999517023379, so:

$ sox -n -n trim 0 150431808s speed 0.999517023379 rate -q stat
Samples read: 150504498

[Radikal Den store]

I see, thought speed ratio was the same as when changing the factor of samples.

But then , which is the correct approach, changing samples, or speed?

What iam trying to accompish is, i get a factor , for example, it runs 24ms less in 10 hours. Meaning "Duration"-24ms.

I am Not talking about trim here.

How exact can it be done?
I can get a very precise calculation of the ratio, but how precise will Sox allow?