Which Sampling Rate - 44.1 or 96?

[DarkFriend]

To date, I've been recording 24 bit, and 96,000 Hz, then obviously mixing to MP3.

A lot of studios I've been in contact with seem to be at 24 bit and 44.1 Hz. Obviously, the lower sampling rate generates smaller WAV files and is easier on the plugins.

If I move from 96,000 to 44,100, will there be a significant loss of quality in my recordings? I've always wanted to go as high as possible, but if the "big" studios are at 44,100, I guess I should be too?

Thanks for your input.

 

[XLR]

I could tell a big difference between 16 and 24 bit tracking. But when comparing 44.1 to higher sample rates, I didn't hear a difference. That was with RME and Lucid AD converters. Knowing how quirky gear can be, I'd imagine it would depend on exactly what you were using.

 

[DarkFriend]

I've been recording at 24/96 and my computer can't take it, really. Aside from songs that occupy 4 GB of space, after 3-4 tracks, I'm in crack and pop land even just one effect laid on top.

Any time I want to hear the mix, I have to mixdown and listen. I can't simply play and listen (which would save a LOT of time).

I've been considering 24/44.1 for these reasons..

 

[XLR]

Any time I want to hear the mix, I have to mixdown and listen. I can't simply play and listen (which would save a LOT of time)...

That would make working practically impossible.

I've only used 96 and 192 on projects where it's been specifically requested. I didn't have playback problems myself, I just didn't hear a difference, and I've got pretty good monitoring. I record mostly acoustic gtr, which is pretty sensitive to fidelity issues.

 

[bennychico11]

lots and lots and LOTS of info about this topic on the forum already. Just do a search

stick with 44.1 or 48 and 24bit. Above that is a waste, IMO.
also, why are you mixing to MP3? that actually will do WORSE to your quality than deciding to record at 44.1kHz

 

[Massive Master]

I record at the target rate 95% of the time. Even for classical. 44.1kHz for CD, 48kHz for video.

I *might* upsample for certain processes - maybe. But not usually.

If you can't get absolutely stellar results at 44.1kHz, recording at higher sample rates isn't going to help anything.

24-bit (over 16-bit) is a no-brainer though...

 

[masteringhouse]

If your computer and DAW can handle the higher sample rate it's better for some types of processing that you may be applying to the mixes, limiting for example. If you're computer chokes on higher rates, stick with lower values.

I've been getting mixes from one client that has switched from 44 to 88K and there was a definite improvement in imaging and a less grainy sound.

Of course that's with everything else being the same. Going from 44 to 88 isn't going to magically create a better mix.

 

[Farview]

You big loss of quality will be when you mix it to mp3. If you are mixing down to mp3, the whole 'quality' argument is out the window.

More than one high-end converter designer has said that if you hear a difference between 44.1k and 96k, your converters are broken.

 

[danny.guitar]

This has been discussed to death. Use whatever sample rate the medium you're mixing down to calls for, ex: CD? 44.1. Video? 48. And maybe for DVD you can take advantage of 192KHz, 24-bit audio but you will need a top-of-the-line computer to be able to mix at those qualities without annoying pops, clicks, skipping, lag, etc. And personally, I would just record at 44.1 and upsample to DVD quality.

 

[masteringhouse]

More than one high-end converter designer has said that if you hear a difference between 44.1k and 96k, your converters are broken.

And the same designers also say that during the processing of audio there can be an overall improvement when using higher sample rates. Recording is one thing (where the converter may be sampling at an initially higher rate), if mixing in the box with digital audio processing it's another.

I don't think Daniel Weiss upsamples and then downsamples lower sample rate audio internally in his limiters for no reason.

 

[SouthSIDE Glen]

And the same designers also say that during the processing of audio there can be an overall improvement when using higher sample rates. Recording is one thing (where the converter may be sampling at an initially higher rate), if mixing in the box with digital audio processing it's another.

If we assume this to be true, that raises this question in my mind:

Is there in such a case a difference in quality - all else such as actual converter quality being equal - as to whether one records at a higher sample rate or whether they record at a lower rate and then upsamples in mixing?

Personally I'd like to see 66.15k conversion, but since no one yet (that I know of) offers that, I personally think 44.1k is just fine and that 88.2 or above is at best wasteful, if not detremental (according to some Katzian theories). BUT...

If processing at higher rates does make a worthwhile difference (I'm not completly sold on that yet, but I'm listening ), my kneejerk thought there would be that it'd be better to bring the signal in recorded at that higer rate as that would be feeding the plugs a more accurate source signal instead of a 44.1 representation with just a bunch of zeros pasted on at up-sampling.

Does that make sense, or am I misunderstanding something there?

G.

 

[masteringhouse]

Glen-

I do agree that bringing in the material at the higher sampling rate is better than going through the conversion, SRCs can have a "sound" like everything else. The Barbabatch site has several examples of this.

I think the following link answers your question better than me trying to summarize:
http://www.dsprelated.com/showmessage/31688/1.php

But if were to summarize it would be that upsampling allows one to filter out the garbage better before going back to the final delivery rate.

 

[SouthSIDE Glen]

I think the following link answers your question better than me trying to summarize:
http://www.dsprelated.com/showmessage/31688/1.php

But if were to summarize it would be that upsampling allows one to filter out the garbage better before going back to the final delivery rate.

I just read through that thread twice...great thread; it actually answered/corrected a couple of non-related concepts I had in my head.

I'm still digesting all of that thread, trying to understand it properly. it sounds like your non-summary summary pretty much hits it right.

But isn't that pretty much the same reasoning given for higher sample rates to begin with? IOW, the resons they give for upsampling in post are pretty much the same reasons given for using higher sample rates throughout; i.e. easier and better LPF control of the aliased frequencies.

If that is the case (and I'm still digesting it all, so I'm not positive I have that right), and it is also the case that it's better to feed a full sample from the beginning than it is to upsample, then it seems to me that upsampling in the plug is not only inferior to sampling high to begin with, but would then be also subject to the same Katzian problems supposedly inherant in ultra-high sample rates to begin with.

Which, boiled down, really brings just brings us back full circle to the original question of sample rates to begin with, with no real movement one way or the other in the debate.

Just thinking out loud here.

G.

 

[masteringhouse]

But isn't that pretty much the same reasoning given for higher sample rates to begin with? IOW, the resons they give for upsampling in post are pretty much the same reasons given for using higher sample rates throughout; i.e. easier and better LPF control of the aliased frequencies.

If that is the case (and I'm still digesting it all, so I'm not positive I have that right), and it is also the case that it's better to feed a full sample from the beginning than it is to upsample, then it seems to me that upsampling in the plug is not only inferior to sampling high to begin with, but would then be also subject to the same Katzian problems supposedly inherant in ultra-high sample rates to begin with. Pretty much the same argument as delivering mixes at 24 bit over 16 bit.

Which, boiled down, really brings just brings us back full circle to the original question of sample rates to begin with, with no real movement one way or the other in the debate.

Just thinking out loud here.

G.

Yep, I believe it would be better at least theoretically, if not audibly, to leave things at the higher sample rate throughout. There are some practical realities though that also creep in, e.g. can your computer keep up with these rates. In these cases the damage done by improper processing outweighs the advantages in upsampling. Another example may be a piece hardware/software that doesn't support a higher rate and the advantage in using it outweighs the benefit of upsampling. In a case were you do need to upsample given a source that is at the target rate, using upsampling helps remove quantization and alias nasties from the resultant audio. If you can deliver at the higher rate, you also avoid the processing needed to get to the higher rate, so it's an added benefit in addition to upsampling. Pretty much the same argument between delivering at 24 bit over 16 bit audio.

Can you fill me in on the Katzian theories? Everything I've read from Bob supports upsampling.

 

[SouthSIDE Glen]

Can you fill me in on the Katzian theories? Everything I've read from Bob supports upsampling.

I *think* it was Katz, maybe it was Lavery. The gist of the article was that once one got past a sweet spot of somewhere above 44.1k - but below 88.2 or 96k - that there was actually a diminishment in signal quality; that more bits actually started equaling more distortion on the far side of that bell curve.

This is the article that made me long for a 66.15k sample rate as being close to the "sweet spot".

I don't think I have that article bookmarked, let me see if I can find it again...

G.

 

[SouthSIDE Glen]

I've been searching for that Katz article, and i can't find it just yet. I originally got it from someone else's post on this board so someone here knows what I'm talking about. I think - but I'm not quite sure- that the conversation and/or the Katz article were discussing the relative differences between high-sample rate PCM and 1-bit DSD.

Anyway, I have found a couple of references from Dan Lavry where he talks about loss of accuracy increasing with PCM sampling rate:

http://www.gearslutz.com/board/high-end/118517-mix-96k-not-3.html#post1234224
http://lavryengineering.com/documents/Sampling_Theory.pdf

He seems to be referring mostly to physical design limitations. I'd swear I remember Katz's article taking it one further by explaining in terms of actual information theory and not just physical design limitation - but I can't remember the specifics.

Anyway, they both refer to a decrease in actual sample accuracy in one form or another as sample rate goes up (in realms outside of Nyquist; Nyquist in and of itself still holds it's own). As sample rate goes up, the number of accurate bits goes down, for a couple of different reasons, both practical and theoretical.

If anyone here remembers that thread or that article and can find that Katz article, and put the link here, I'd appreciate it. in the meantime, I'll keep looking when I get the time to search deeper.

G.

 

[masteringhouse]

Good posts, but let's not confuse DSP processing with making a good converter, they are two different animals. These posts seem too be directed toward the later.

 

[jaybriggs84]

I wonder if people used to argue about 15 ips vs. 30 ips 20 years ago like people argue about sample rates today

 

[SouthSIDE Glen]

Good posts, but let's not confuse DSP processing with making a good converter, they are two different animals. These posts seem too be directed toward the later.

That's why I really want to find that Katz article. I might be wrong, but I thought he put forward a line of reasoning that was mostly information theory and pure math, and not physical design limitations. Maybe I'm remembering incorrect, maybe not. I really want to find that article again to find out for sure, one way or another.

EDIT: Re the ips issue, I don't recall any such level of disagreement - at least not in my circle of communication back then. The physics were/are much simpler to understand and the results less dependant upon hardware design with tape speed than they are with sample rate. Also, back then there was no internet where anybody with a keyboard and modem could chime in regardless of personal backstory .

G.

 

[masteringhouse]

I wonder if people used to argue about 15 ips vs. 30 ips 20 years ago like people argue about sample rates today

15 ips has a bit more "bump" in the bottom and more noise than 30 ips. In general the feeling was/is 15 ips rocks more, 30 ips for purity. The arguments as I recall were not as numerous though. It's more obvious than some of the differences in digital. Of course when you have a 15 minute long drum solo like many of the songs back in the 70s, the choice was pretty obvious.