View Single Post
Old 7th August 2005
[EDIT: Just ignore all the nonsense I spew below about sampler rate conversion between 'uneven multiples' being problematic. I didn't know what I was talking about.]

A higher sample rate is better, all things being equal.

But things never are, are they?

The obvious downside is relatively negligible (as I'm sure most experienced hands would agree the improvement is, as well): a roughly 10% increase in processing/storage overhead.

But the really nasty downside -- depending on how you work -- is what happens when you take your nice, sparkly 48 kHz mix down to 44.1 kHz for output to CD, Mp3s, etc.

If you do it in the digital domain (through SR conversion utilities/plugs) the conversion entails a rather brutish remapping of sample values across time.) [This is nothing like moving from one digtial word length [bit depth] to another, which is a fairly non-destructive process, given the givens.]

I should point out that this nasty remapping only occurs when you move from a sample rate that is an uneven multiple of the target. If you go from exactly double the target sample rate (88.2 kHz) down to 44.1 kHz, the hit is simply a reduction in sample density (resolution). But if you move from 96 kHz or 48 kHz down to 44.1 kHz, watch out, you're going to be re-aliasing your entire file (on top of reducing resolution).

I went into some depth on this issue in this post, complete with graphics:

Anyhow, bottom line -- if you record at 48 kHz and perform a sample rate conversion to get your audio down to the 'industry standard' of 44.1 kHz, you will very likely end up with a worse sounding product than if you started at 44.1 kHz in the first place.


This assumes, of course, you're keeping everything in the digital domain. If you mix 'out' through the analog outputs of your device (at 48) and record back into another device set to 44.1 kHz you won't have to perform that software SR conversion -- but you will be going through a whole 'nother D/A and A/D processing cycle.

Still, in the past, that's just what I did when going from 48 kHz (16 bit) DAT mixes into the computer. I first tried coming in via SPDIF and doing the SR conversion. And it really sucked. (This was a few years ago. The vendors will try to tell you that they've mastered SR conversion. I don't think so.)

So I ended up just taking the analog outs of the DAT and recording them into the analog ins on my computer rig. And it sounded a huge amount better. YMMV, to some extent. But 'uneven' SR conversion remains problematic.