What sample rate are you recording at?

[waltzmastering]

From what I see coming in and also from a poll I conducted amongst mastering engineers not to long ago.. this is how it breaks down ime:

35% 44.1
35% 48
10% 88.2
10% 96
10% all others

[waltzmastering]

I'd use 88.2 and deliver the master to the mastering house at that rate. The conversion is much simpler math and is, again, less destructive than it would be going from 96 to 44.1.

On the issue of integer versus non-integer down sampling
This is a common myth.

From Daniel Weiss:
"The conversion from 96 kHz to 44.1 kHz can absolutely be as good as one that converts from 88.2 to 44.1 kHz.

[jonathanm]

On the issue of integer versus non-integer down sampling
This is a common myth.

From Daniel Weiss:
"The conversion from 96 kHz to 44.1 kHz can absolutely be as good as one that converts from 88.2 to 44.1 kHz.

What about going from 48 to 44.1? I've heard (though not experimented with it myself) that some conversion algorithms leave artifacts. Anyone have direct experience with this (i.e., heard it with your own ears), or have a quote from an authority like Bob Katz, etc.?

[mazz]

I guess the question in my mind is: are there better algorithms than others for doing sample rate conversions?

If in fact, the conversion from 96 to 44.1 works just fine, are all conversion schemes created equal or is there really only one conversion scheme anyway since it's "just" math? Obviously the inclusion of dither can make a difference but it would be interesting to know if there are different methods of conversion and if they're all created equal.

Hmmmm, time for some digging!

[waltzmastering]

Here is a graph/chart showing quality of conversion between different converters:
http://src.infinitewave.ca/ ..the faq and help page give tips on how to read it.

ime/o sr conversion has improved drastically over the years and comes down to what your using for the process.

For the money Izotope is one of the better performers.

[jonathanm]

That's good stuff, I think?

It'll take me a bit to digest, and I'm tired right now. It certainly makes it clear that not all sample-rate-conversion (SRC) algorithms are the same.

Given that, unless you're using the iZotope stuff, it may be best to record at the sample rate you think you're going to need it in, or a multiple of it.

Bottom line seems to be that down-converting from 48k (or a multiple of it) to 44.1k (or a multiple) is certainly going to introduce some artifacts. How much and what kind of distortion depends on the software.

Hmmm.

One other thought, this one regarding orchestral samples. At what rate were the samples you're using recorded? Obviously, if you're recording at 96k but using samples from a library recorded at 48k, there will be no benefit (except for new tracks you record, and possibly the quality of the fx plug-ins.)

Additionally, it seems that using 96k or 48k samples in a 88.2k or 44.1k project would incur some of the distortion artifacts demonstrated by Tom's link.