Sample rate conversion question.

Karyn


There will be timing errors,  but only at the sample level and thus so small (<44 thousanth second) that even you couldn't notice.

alexoosthoek


Thanks Jon.

bapu


So, how would one conduct a null test of a 24/48 and a converted 24/44.1?

Or better still, if one could conduct a null test (the mechanism escapes me since a Cakewalk project can only ever be one sample rate for all tracks) would it be truly null?

Karyn, assuming a straight 1/4 note snare hit, would each hit be off relative to true 1/4 note or would the 1,000,000th hit now be considerably off. Theoretical I know but can you see what I'm getting at over a 5-10 minute song?

the number is actually closer to 60 trillionths of a second.

drewfx1


Well it depends, the less loud you play (for a given bit depth) the timing resolution goes down somewhat.

So, as always, you'll want to play as loud as possible.

bitflipper


Just to be clear: neither sample rate nor sample rate conversion affects timing. Every note will still be exactly where you left it.

Jonbouy

Ed per second is per second whatever way you look at it.  So if you song is 186.34521124556674222 minutes long it's still going to be the same length and all the time points along the way are going to be identical. 

It likely won't null perfectly because of the altered sample rate and any aliasing will show up as Karyn said only at the sample level, but it ain't going to be fret worthy.

bapu

Jonbouy

Ed per second is per second whatever way you look at it.  So if you song is 186.34521124556674222 minutes long it's still going to be the same length and all the time points along the way are going to be identical. 

It likely won't null perfectly because of the altered sample rate and any aliasing will show up as Karyn said only at the sample level, but it ain't going to be fret worthy.

The first concept I now "get".


 I think you're right about the second concept but I'm still hard pressed to figure out how to null test a 24/48 and 24/44.1, since my sound card and SONAR can only one or t'other but never both simultaneously. 

drewfx1

bapu

Jonbouy

Ed per second is per second whatever way you look at it.  So if you song is 186.34521124556674222 minutes long it's still going to be the same length and all the time points along the way are going to be identical. 

It likely won't null perfectly because of the altered sample rate and any aliasing will show up as Karyn said only at the sample level, but it ain't going to be fret worthy.

The first concept I now "get".


I think you're right about the second concept but I'm still hard pressed to figure out how to null test a 24/48 and 24/44.1, since my sound card and SONAR can only one or t'other but never both simultaneously. 

Convert to and back and then null with the original. For 24bits, it might or might not be "perfect" depending on the SRC used. But regardless, I suspect the differences wouldn't be audible to people who listen with their ears and not their imaginations.

You see charts illustrating the effects of different SRC's here:

http://src.infinitewave.ca/

bitflipper
However, a null test wouldn't be the best way to check timing. The notes (or, more accurately the several thousand samples that make up the initial transient and therefore the onset of the note) are going to be in exactly the same place regardless of whether the two files null. 

bapu

bitflipper
However, a null test wouldn't be the best way to check timing. The notes (or, more accurately the several thousand samples that make up the initial transient and therefore the onset of the note) are going to be in exactly the same place regardless of whether the two files null. 

And so it may be said that the transient timing of the converted wav will always be as accurate (in time) as it was recorded?