Sample rate question

[Artcutech]

everything i've been doing has been sett at 44k, 512 buffer size and when exporting it's been 16 bit depth and triangular dither option.

Recently i was watching a tutorial and someone said you should always have bit depth either 24 or 32, and he had his sett to no dither, so then i started thinking about sample rate, his was at 44k but is it better to have it sett a little higher like 96k? and if so should i increase my buffer size maybe to 1024 instead of 512 if sample rate is sett to 96k?

If the sample rate and buffer size are changed will my old arrangements be effected in a bad way if i try to open and play them?

thanks and any advice helps

[Stromkraft]

everything i've been doing has been sett at 44k, 512 buffer size and when exporting it's been 16 bit depth and triangular dither option.

Recently i was watching a tutorial and someone said you should always have bit depth either 24 or 32, and he had his sett to no dither, so then i started thinking about sample rate, his was at 44k but is it better to have it sett a little higher like 96k? and if so should i increase my buffer size maybe to 1024 instead of 512 if sample rate is sett to 96k?

If the sample rate and buffer size are changed will my old arrangements be effected in a bad way if I try to open and play them?

24 bit is minimum most of the time. Compared to 16bit depth it's a game changer going from an amplitude resolution of around 65000 to more than 16 million discreet (integer) values. Once you have recorded at 16 bits, there's no point in converting to 24. The internal resolution of the Live mixer is already 32bit.

Running especially synths and to a certain extent effects in oversample mode can pay off quite well, but the advantage of running the whole project at a rate higher than the intended end sample rate is probably marginal at best if you got the computer power to be able to do that. You can use very low latencies which can be an upside, but again you need the computer power.

It could possibly be that the internal Live synths and some third party synths that can not be manually put into High Quality or oversampled mode sound better in an oversampled project as that may be the only way to make some run in that mode (yet again, some may not), but I have never verified that. I run all softsynths of mine, that can, in high quality or oversampled mode, at least when rendering.

As for exports, 16bit dithered is OK for when you intend to publish as is (or convert to another format for publishing), but if you are going to leave this for further work like if you want to master in another project or DAW or leave to another party, 24 bit undithered is probably the best choice, If leaving to a mastering engineer it can be a good idea to ask how much head room is expected.

I've run multiple sets in 96kHz last year due to recording the Roland TR-8. Recording to audio at a higher sample rate doesn't affect the quality of sound as far as I can hear when recording hardware synths or instruments. Still some audio interfaces might sound better that way.

[Artcutech]

thanks StormKraft you've always been a big help on here.

I guess i remember the days where you couldn't put anything on to cd that was over 44k or something like that.

So as long as i intend for my end sample rate to be at 96k, and bit depth 24 and my buffer at 512 this should mesh ok?

and the final wave file should have no problem playing on any system or device?

sorry about the novice type questions, thanks again

[Tarekith]

CD is still 16bit, 44.1kHz, so no matter what you're going to end up there if that's your final format.

Personally I write songs at 24bit/44.1kHz, there's very little benefit to working at higher sample rates most of the time these days I find. Probably not much harm in doing it either as long as you use another app to upsample to 96kHz, but with little to no audible benefit I'd rather have the CPU power in reserve instead.

[Artcutech]

nah i won't be working or planning on using cd as my final format but i do remember those days where everybody made sure to mention how you had to stick to that specific sample rate if you wanted to be able to put it on cd and it kind of stuck with me not to move off of it.

I just recorded some vocals at 96k and the quality seems better but maybe i'm just tripping, anyway i see everybody using 44k in their tutorials at 24bit depth and after you just said that's what you use Tarekith i think i'm going to stick there, but definitely try stuff back and forth in the future

[Tarekith]

Probably a good idea, experimenting with this stuff yourself every now and then is the best way to learn if it's really worth it for you or not. One thing to keep in mind is that Live uses the excellent SOX sample rate conversion when going from a higher to a lower sample rate. However when going from a lower to higher sample rate, it still uses an algorithm which is efficient, but honestly not the best sounding. So if you do decide to convert to a higher sample rate, you'd likely want to invest in a better SRC tool to get any benefit.

[Artcutech]

thanks for the advice

Ableton never ceases to amaze me, so much user friendly common sense has been put into it

[Stromkraft]

So as long as i intend for my end sample rate to be at 96k, and bit depth 24 and my buffer at 512 this should mesh ok?

and the final wave file should have no problem playing on any system or device?

You can release at 96kHz if you like and generally there's no harm recording at that sample rate provided you can afford the resource costs involved. I'd advise to have that as only an option though, because I think some outlets you probably want to be at won't necessarily be accommodating or be especially happy towards this sample rate. I'm not sure how large part of your intended audience would appreciate this either given the larger file sizes. If you're sure then why not though?

I want to add that recording to higher sample rates also allow for high fidelity slow downs and might — again, I have no tests I've made — be beneficial for certain other audio operations as well in some cases.

Personally, I'm sticking to 44.1 kHz for releases, but of course I'd consider higher if I was thinking about releasing in a format that preferred a higher sample rate for some reason.

Digital audio is different
It's important here I think to realize that digital audio is different to digital video in the sense that audio is recreated when made analogue. In video individual pixels, represented as light which is the analogue domain, are what we actually look at. When we are at a certain pixel density at a certain viewing distance adding more won't give a more clear picture as we're at "retina" resolution. At least any gains would be very subtle. If we zoom in or make the projection we look at much larger we may still notice individual pixels as just that. We can see the pixels.
(There have been some attempts to have infinite resolutions but that is of no interest here)

Pixels vs samples, i e actual vs recreated
It's easy to believe individual samples are comparable in concept to pixels, but they are not. We're not listening to the samples themselves when we listen audio from digital formats. The audio signals we hear are recreated from the individual samples when made into an analogue signal.

"Zoom" into image, "zoom" into audio
The zoom aspect above is also different. We zoom analogue audio by making it louder (ignoring the time aspect as that normally is fixed). While what we hear then may affect acoustical properties of our listening environment and thus the sound we hear the basic waveforms are the same no matter the volume. As long as audio is in the digital domain we can zoom into this in ways that are unique to digital, but this is only true within the digital domain, not the analogue.

Investigating samples
The only way to be able to compare to pixels in a meaningful way here is I think to take the resulting analogue audio signal and re-record this at a higher sample rate, say 192kHz. Once recorded you wouldn't be able to identify the original individual samples. Which would be somewhat different when comparing to an upsampled version. Granted the signal wouldn't contain much info above 22kHz and the 2 recordings would likely nullify each other more or less depending on the analogue path and aspects of the ADC, but that's not disproving my point here. Which is that digital audio samples are recreated into sound in the DAC before we can hear the analogue signal trough the air.

Investigating pixels
In comparison with digital video if you did the same type of test and you filmed a video projection in for example 720p resolution with a 4k camera you would be able to blow up and identify the original pixels. Because these were visible all along, unlike individual samples points.

OK, that was more lengthy than I hoped for, but when we're discussing sample rates then what samples really represent is worth keeping in mind I think.

[lapieuvre]

Since the CD era seems behind us, I always use 48Khz/24 bits.

I first export my audio files in 48Khz / 32 bits, then apply a mastering plugin (Ozone) to render my files in 48Khz/16bits

Then do mp3 in 48Khz / 256mbps.

44.1Khz is out of my vocabulary

[Tarekith]

If you're selling your music online through an aggregator, a lot of them will not accept 48kHz wav or aif files. So even though the CD era is behind us, there's still a lot of times you need to release music at 44.1kHz. If it's just MP3's for your own consumption or to share with friends, by all means go for it. Though I doubt there's any audible difference between 44.1 and 48 myself.

[ark]

Each bit in a sample represents 6 dB of dynamic range.

Therefore, 16-bit audio is 96 dB and 24-bit audio is 144 dB.

96 dB is CD quality. If your levels are adjusted appropriately, 96 dB is how much louder the loudest sounds you want to accommodate are than the quietest.

Right now I'm sitting in a quiet room in a suburban home. The loudest sound is the little clicks from my keyboard. While I'm not typing, the loudest sound is the quiet whirr of my computer's cooling fan. A little smartphone app tells me that that's about 45 dB. A really good concert hall might be as quiet as 30 dB if the audience is really quiet.

Add 30 dB to 96 dB and you get 126 dB. That's loud enough to cause permanent hearing damage almost instantly. A symphony orchestra going full tilt, heard from the 10th row, is about 105 dB, and that's LOUD.

So realistically, 96 dB of dynamic range is enough to accommodate any performance that you might want to listen to--IF you take care to adjust that dynamic range to match the listening conditions. This observation corresponds to the fact that if you're playing a CD, you can walk right up to the speakers during the blank space between tracks and not hear anything, and still have the loudest parts be uncomfortably loud.

But you do have to do something about gain staging. The advantage of using 24 bits is that you don't.

However, in Ableton Live, you don't have to worry about that so much, because all internal audio processing is in 32-bit floating-point, and all mixing is done in 64-bit floating-point. So for all practical purposes, you are not going to run into any dynamic-range limitations inside Live.

Therefore, all that really matters is that any samples you use are recorded with adequate dynamic range, and that the final output is properly staged so that it uses most of the dynamic range inherent in the signal format you have chosen.