Sample rates question

Hi all,

This is not the usual type of question I see here, but I don't have much faith in the kinds of groups that do discuss this. I trust the members here.

Anyway, I was doing some mixes last night in Pro Tools HD, 24-bit 44.1k sessions. The question is, is there anything to be gained sonically by bouncing to 24-bit 192k files rather than 44.1k? It seems like bigger is usually better in digital audio, but I have no idea how digital summing works.

Thanks

Mixing at 24/192 an then convert to 24/44.1 might be a slight improvement over 24/44.1 to start with. A lot will depend on the material.
if there are a lot of quite passages requiring a lot of detail you may notice an improvement. The quality od the AD/DA has more of an impact.

RonL

Thanks Ron.

agreed...

all that will basically happen when you move from 44.1 to 192 is that every 1 sample will become 4 samples at the 192 file.

There may be some clever dithering type maths that goes on... (i'm sure someone will correct me here... come on PRR, you know you want to :grin: )

Anyway, the guy that got me into all this once said to me... "you can't polish a turd". I translate that to be, you won't get better quality by converting up after you've recorded.

I'm told that one of the best ways to improve the quality of your recording is to get a professional wordclock input to your converters. (e.g. the M-Audio delta1010 sounds a lot better with a decent wordclock)

Cheeers

R

[quote author="rlaury"]Mixing at 24/192 an then convert to 24/44.1 might be a slight improvement over 24/44.1 to start with.[/quote]

I have to disagree with you on this one, Ron. By recording at 192KHz your antialias filter's slope is gentler on your material... but the sample rate conversion down to 44.1KHz itself is more detrimental to your audio, no matter how advanced the algorithm you use.

This is of course completely invalid if you're mixing using analog gear, in which case the multiple A/D-D/A conversions will not sound as bad if done at 192KHz, but if you're staying inside the box and are only doing standard audio 16/44 CD's, 44.1KHz is the way to go.

Of course with bit depth, it's a whole different animal, which brings us to:

[quote author="rlaury"][...]A lot will depend on the material. If there are a lot of quite passages requiring a lot of detail you may notice an improvement.[/quote]

Sample rate has nothing to do with how accurately signal levels are represented digitally. Quiet passages will sound better if recorded at a higher bit depth and dithered down to 16 bits, but 192KHz recording will give no improvement over the recording/reproduction quality of low signals over 44.1KHz - or even 8KHz for that matter.

Note that recording at 24 or 32 bit and converting down to 16 will only offer an improvement over recording straight to 16 if you use dither!

As a general rule, record at as high a bit depth as you can, but at the same sample rate as your final product.

Peace,
Al.

PS: I just re-read the original post and realized that you want to convert up to 192KHz from a 44.1KHz recording. Don't do it. There will be no improvement in quality. It can be argued that the result will actually be worse (from the computer "making up" 3.35 extra samples for each original one).

any type of SRC inside of Protools pretty much blows. stay with what ya got. 24/44.1k is fine. It'll get slimmed down to 16 for CD anyway. Don't do that inside of PT either! Use PT processing as sparingly as possible. Bad, bad math happening inside of there.

(Finally something on this forum I actually KNOW about! lol)

I just re-read the original post and realized that you want to convert up to 192KHz from a 44.1KHz recording. Don't do it. There will be no improvement in quality. It can be argued that the result will actually be worse (from the computer "making up" 3.35 extra samples for each original one).

Click to expand...

That's what I was getting at - when you bounce internally from a 44.1k session (multi-track file) to a 192k file is the computer combining the 44.1 tracks into a 44.1 file and THEN converting that file to a 192k file? It's just duplicating existing samples to generate more of them to fill in the spaces? I wondered if there was some logic according to which the computer extrapolates the additional files. Also, does the fact that the original session (the multi-track file) is clocked at 44.1k mean that when it's internally summed to a stereo 192k file there cannot be any different, intermediate data between the 44.1k steps in the 192k file?

I'll make a decision as to what I like better when the time comes, I'm not worried about that. I'm just wondering what actually happens when signals are combined in different ways.

Thanks

I have yet to hear an SRC that has not degraded the sound after conversion. I always cringe when I here people saying that they're going to record at higher sample rates only to SRC down to 44.1. My thoughts are, record at the sample rate that your final medium will be. Word length/bit depth is a different story.

If you really like working with higher sample rates, perhaps the best to maintain the sound without SRC is to layback the high rate session to 1/2 inch.

I just reread your first post and just realized what alk509 did.

Upsampling your 44.1K session will not give you better quality sound or "processing room" as some people lay claims to.

-E

Upsampling your 44.1K session will not give you better quality sound or "processing room" as some people lay claims to.

Click to expand...

I'm trying to nail down in my mind if there's a difference between upsampling a 44.1 k file to a 192k file as opposed to bouncing together several 44.1k files to a 192k file. Is that a different process that approaches the new 192k file as a unique thing, or is it just two sequential steps - combining to a new 44.1k file and then upsampling?[/u]

[quote author="jrmintz"]

I'm trying to nail down in my mind if there's a difference between upsampling a 44.1 k file to a 192k file as opposed to bouncing together several 44.1k files to a 192k file. Is that a different process that approaches the new 192k file as a unique thing, or is it just two sequential steps - combining to a new 44.1k file and then upsampling?[/u][/quote]

The only thing it's good for is wasting some time. Seriously, Admin nailed it IMHO. And Bouncing to Disk in PT is probably the worst thing you could do to a mix. Better off coming out digitally to something like a Masterlink.

STAY AWAY FROM ANY TYPE OF PROTOOLS SAMPLE OR BIT PROCESSING!!!

High sample rates are great for the intial recording, as higher rates will preserve more detail that exists in the sound to begin with. But once you've captured something at 44.1k, you're not gonna enhance it by up-sampling to anything. Even the most minimal amount of digital processing will be more of a detriment to the file.

[quote author="JPrisus"][quote author="jrmintz"]

I'm trying to nail down in my mind if there's a difference between upsampling a 44.1 k file to a 192k file as opposed to bouncing together several 44.1k files to a 192k file. Is that a different process that approaches the new 192k file as a unique thing, or is it just two sequential steps - combining to a new 44.1k file and then upsampling?[/u][/quote]

The only thing it's good for is wasting some time. Seriously, Admin nailed it IMHO. And Bouncing to Disk in PT is probably the worst thing you could do to a mix. Better off coming out digitally to something like a Masterlink.

STAY AWAY FROM ANY TYPE OF PROTOOLS SAMPLE OR BIT PROCESSING!!!

High sample rates are great for the intial recording, as higher rates will preserve more detail that exists in the sound to begin with. But once you've captured something at 44.1k, you're not gonna enhance it by up-sampling to anything. Even the most minimal amount of digital processing will be more of a detriment to the file.[/quote]

JP,

I understand what you're saying, and I don't doubt it. My question is, why? I'd like to understand the process to the extent I can.

Thanks

jrmintz --

Forgive me if I've misunderstood your question, but I may be able to explain why any kind of mixing down in PT blows goats.

The hardware inside the protools rig is a 24bit Fixed point processor. The fact is that when your doing any kind of mixing or effects in audio, it's preferable to actually work in 32bit Floating point (or even 40). The main difference between fixed and floating (for the layman) is that floating point gives a far bigger dynamic range, and tends to sound a bit more 'musical'. There are many opinions on this, some say there's no difference, and some won't even touch fixed point processors.

Anyway, PT try to compensate for this lack of dynamic range by using 'double precision' - essentially taking the processing to 48bit fixed point, at the sacrifice of processing power.

Either way, due to the nature of fixed point processing... it still blows goats.

Proper mixing will only begin to 'sound natural' once they make the jump to a floating point platform... be it Analog Devices, TI or a PC's internal processor. For now, the only thing keeping those Mot's alive is the years of code already written for it (... which blows goats! :wink: )

Sorry, rant over now

Because the sound was initially captured at 44.1k. That means for every second of time that passed during the duration of the signal, there were 44,100 snapshots of it taken. What you're talking about is taking those 44.1k snapshots per second and dividing them into 192,000 snapshots per second, basically subdividing each little sample by 4 and change. You're subdividing bigger pieces into smaller ones, that's all. Imagine having a pizza cut into 8 slices. You could cut those 8 slices into 4 new slices each, for a total of 32 slices per pie. But you don't have more pizza. Just smaller slices.

I like that

What a *great* analogy I'll bear it in mind next time someone asks me that one

Thanks Rochey, I may patent it someday lol

Just for the record, I'm not saying there's no advantage to recording at higher sample rates. I'm still at 24/44.1k personally, because making the leap to 96k hasn't been justified for me yet. Maybe someday! But upsampling is utterly pointless, considering you're just gonna downsample again to put it onto CD anyway. Stay where you are, and avoid the shitty PT math at all costs.

I own and use Alsihad (ProTools).
Haven't had much requests for 96 or 192 yet but have need to convert files for various reasons. Bits, Sample rates and Files types and conbinations. I've done it with software and with Analog bouncing and every combination in between.

Heed advice given above BUT more than anything do what the client wants. Sure, you can give advice client all the advice you like but balancing time money and the clients wishes is the business. Managing the business is what will make or break you.

... or is this job for pleasure ?

an alsiha eh???

You gonna be buying the book too?

Thanks guys - the pizza analogy brings it home for me. Dammit, I want more pizza, not just more slices! And I like goat cheese on pizza, BTW.

:green: :guinness: :guinness:

BUT more than anything do what the client wants.

Click to expand...

Kev, in this particular case I'm the client. The buck stops here. :shock: Ultimately I'll let my ears decide. I'm tired today so I'm not listening critically until tomorrow, just thinking about stuff. This job has been a huge treat and I want it to be as good as I can possibly make it. When I get the green light I'll post some mp3s.

Cheers :guinness: :sam:

Kev, great point... any of us who are doing this professionally are in the business of providing a service. I have yet to hear a client request a specific sample rate, but ya never know. Stranger things have certainly happened!

The topic of sample rates is completely subjective and particular to the AE in most respects. The topic of PT's horrible math affecting audio is NOT open for debate IMHO. It's damn near been proven that it hurts, and this is an area where Digi proceeds to skimp constantly. I believe it would take a complete program and hardware overhaul to correct this. An easy solution is, use PT for as little math and processing as possible! Use it for what it's good for... multitracking and editing.