taking a project from 96 to 44.1

In the project settings, set samplefrequency to 44.1 kHz, then Convert all used audio files from 96 to 44.1 kHz, tick *Change file reference in Bin*.

The sound quality will be significantly degraded at 44.1kHz, downconverting to 48kHz will sound better.

This is all hindsight of course, and a downward conversion doesn't sound like it was part of the plan, but if you know your final format will be 44.1, choose 88.2 or 176.4.

Or better 352,8.

Christian

Andy Ry wrote:
The sound quality will be significantly degraded at 44.1kHz, downconverting to 48kHz will sound better.

This is all hindsight of course, and a downward conversion doesn't sound like it was part of the plan, but if you know your final format will be 44.1, choose 88.2 or 176.4.

Sorry, but this is simply incorrect. Digital audio is based on some rather complicated math, and things like this that appear to be common sense are often DEAD WRONG.

Myth #1: Increasing the sampling rate will capture a "smoother" or "more detailed" waveform. Wrong.

Myth #2: Sample rate conversion will sound better if the two sample rates are evenly divisible. Wrong.

The list could go on...

Not to pick on you, Andy, because I'm sure your reply was well-intentioned, but as general advice to everyone out there, please don't spread around "advice" about digital audio unless you're getting it directly from a reputable source.

-James

Well, I certainly welcome your opinion on the subject. Sounds like you might have debated this before - you seem very passionate about the subject.

My math has never been stellar, but I'm certain a 96kHz session converted to 48kHz will sound better than 44.1kHz - thats math I can handle! 😀 Clipping off that extra 4k might not show up on an ipod, but I'd wager you'd notice the difference on a set of decent speakers.

I noticed there, that you questioned my knowledge/reputation. Despite producing and mixing music for a living since 1986, I don't consider myself an expert on anything... that's for sure. And I don't think I would ever assume I was more knowledgable than someone I know nothing about. 😉

Guys like Bob Clearmountain, me, and many others decision to track music the way I described comes from listening critically to the difference. But we're all pretty humble guys, and sure, hey, we might all be wrong and you're right! No sarcasm intended. 🙂

Just go with my original advice. Hadn't seen this old myth about downconverting resurface here, and I agree with Jnash completely. Extra advantage: your CPU overhead will also go down considerably, as you're only sending (less than) half the data to be processed by plugins after the conversions.

regards, Erik.

Just to be clear, I am not a proponent of recording at sample rates above 48kHz, I am merely stating that down converting 96k to 48k will retain more information than 44.1.

In addition I am stating the fact that many recordings made by well respected engineers and producers who record at higher rates with CD's in mind have chosen 176.4 rather than 192.

But don't take my word for it; make up your own mind.

Set up a session at 24bit 96kHz. Record 5 seconds of white noise from a tone generator.

Bounce it out at 24/96, convert that file to 24/48kHz and 24/44.1kHz.

Listen and compare the 24/48kHz and 24/44.1kHz files.

If you hear a slight, but discernible dip in the frequencies above 16kHz with the 44.1kHz, then you might want to try 48kHz, if you don't, then go with 44.1kHz.

Do that, and you can make your own decision. 😀

Again, byzantine, hope that helps, and best of luck with your project.

Andy Ry wrote:
Just to be clear, I am not a proponent of recording at sample rates above 48kHz, I am merely stating that down converting 96k to 48k will retain more information than 44.1.

Theoretically true, in practice not. Nobody will hear the difference between a 44.1 or 48 file, both downconverted from the same source. Do you really think present computing power cannot handle the difference between 2 and 2,17687074829932?

In addition I am stating the fact that many recordings made by well respected engineers and producers who record at higher rates with CD's in mind have chosen 176.4 rather than 192.

Probably because the mental arithmatic is so much easier that way. And simple mathematics have a psychological and esthetic effect too, like foodcolouring improving taste.
Not too many pro's use that samplefrequency anyway. It is generally considered overkill - squared. I think it is a "monkeysandwich".

But don't take my word for it; make up your own mind.

Set up a session at 24bit 96kHz. Record 5 seconds of white noise from a tone generator.

Bounce it out at 24/96, convert that file to 24/48kHz and 24/44.1kHz.

Listen and compare the 24/48kHz and 24/44.1kHz files.

If you hear a slight, but discernible dip in the frequencies above 16kHz with the 44.1kHz, then you might want to try 48kHz, if you don't, then go with 44.1kHz.

That is the direct difference between the two samplingrates, and not the effect of the downconversion. The rolloff on 48 kHz is less steep, because, as we all know, its' nyquist frequency is slightly higher. If you would make recordings (not downconversions) at those frequencies, you would probably hear the same.

But, since I am over 40, I can't hear anything above 13 kHz anyway... 😉

Andy Ry wrote:
I am merely stating that down converting 96k to 48k will retain more information than 44.1.

And this makes total sense, but the difference isn't due to sample rate conversion (SRC). 48k will preserve the highest audible frequencies with more accuracy than 44.1k--no doubt. But if you recorded your white noise test natively at 44.1k and 48k, you'd see the same results, independent of any SRC at all.

The comment I was responding to was

if you know your final format will be 44.1, choose 88.2 or 176.4.

This is the myth I was responding to. If I understand correctly, you were talking about comparing SRCs:

88.2k -> 44.1
96k -> 44.1
176.4k -> 44.1
192k -> 44.1

One might assume that the 1st and 3rd options would sound better because the math appears "simple," but the fact is that the math isn't simple for any of these cases (i.e. converting from 88.2k to 44.1k does not simply involve throwing out half of the samples--doing that would introduce some serious aliasing).

No disrespect, Andy, and to be honest, I did assume from your single-digit post history that you were a noob to recording. You've obviously got a ton of experience making music, and real-world listening always trumps "on paper." (Start comparing THD, frequency response, dynamic range, and one might conclude that virtually every mic sold at Guitar Center sounds better than a U47...)

For any particular converter, it's possible, for instance, that 44.1k might be the best-sounding choice, even if higher rates are available. On paper, 48k should sound better, but depending on how the converter is designed, anything's possible. Another example: Dan Lavry makes a compelling case why 192k can actually sound worse than 96k. The fact is, real gear has way more variables than the underlying sampling math, so even valid math won't necessarily lead to the right conclusions about what sounds best: I totally agree there's no substitute for actually doing the tests and listening.

Cheers, Andy!

-James

byzantine wrote:
well, caught in the polite (& passionately mathematical) crossfire which i appear to have generated i wonder where to go now

has anyone else any opinions to share on this one before i start digitally castrating myself?

Yeah, this thread is interesting, but not exactly helpful in solving your problem, eh ? Sorry to hijack.

I've got some suggestions for you, but need to run--I'll post again this afternoon. Hope I can help (but no guarantees... 🙂

-James

byzantine,

Could you explain what happened when you froze the tracks ? What issues still exist ?

If you're running VIs that simply won't work at 96kHz, downsampling may be your only option. But otherwise, it's probably possible to get your project working by freezing tracks, adjusting buffer settings, etc. (standard performance-tweaking stuff)

If all your tracks are now frozen, I'd be surprised if you have any CPU load issues, even at 96k. Possible that with a large number of frozen tracks you're running into disk speed issues, which might be fixed by spreading files across multiple disks, or defragmenting (although there are pros and cons to that--defragmenting audio files can sometimes degrade performance due to the way multitrack audio is read from disk).

More details about what's going wrong and what you've tried might help...

If you do end up downsampling, don't be at all concerned about your starting sampling rate (doesn't matter that it's 96k, not 88.2k). The ultimate question is whether to downsample to 48k or 44.1k, and there's not an obvious right answer. If you choose 48k, you'll probably get a slightly better-sounding mix. But... if your ultimate goal is a 44.1k master, you'll then need to do a second round of SRC to get there, and the SRC process probably does introduce subtle artifacts every time you do it. So, the choices are:

96k project -> SRC to 48k project -> mix at 48k -> SRC to 44.1k master
96k project -> SRC to 44.1 project -> mix at 44.1k

If you downsample to 48k, your mixing stage will have an advantage, but the result will get hit with a second round of downsampling. If you downsample to 44.1k, you'll take all of the sonic hit in one step, avoiding a second SRC step, but you'll have a slightly narrower frequency response at mixdown. Hard to guess which will produce the best results. Maybe someone out there has compared ?

Best, though, is if you can get the 96k project to mix as-is. Maybe if you explain what's going wrong in more detail, someone will have more solutions to that end. If there's a particular VI causing problems, maybe it's even possible to process that track standalone (outside Logic) ? Just an idea...

-James

Byz--so you froze all your existing tracks (including all MIDI tracks) ? And everything worked fine until you started adding more tracks ? What did your CPU meters show before you added the additional tracks ? (CPU usage should have been very low, if it wasn't, maybe some bussed plug-ins were bogging things down ?)

If the CPU usage is fairly low with the existing frozen tracks, then it spikes and you get the errors when you add the new instrument, it sounds like a CPU issue related to the Arturia stuff. You've tried raising the buffer, I'm guessing ?

What if you bounce your existing project (at 96/24), then start a brand new project (at 96/24), import that single stereo bounce, then start creating the new instrument tracks. If that works, finish the new tracks, freeze them, then hopefully you could combine the two projects for mixdown ?

Another way to go if none of the above works: convert your project to 44.1 or 48k (ideally 48k if the problem instruments run well at that rate), create your new tracks with the problematic instruments, then bounce them as individual audio tracks and import those bounces back into your original 96/24 project. That way, you wouldn't have to degrade your existing tracks with SRC, and you could still mix at 96/24.

I know this is all a big pain...

-James