96Khz. Why?

[Stromkraft]

Is there really any point to working at 96Khz? I try it, and I think it sounds better, but it just might be more like a placebo effect.

I certainly agree that running some the output of some synths in 88.2kHz or higher sounds better. This has been demonstrated convincingly to be beneficial in other discussions I've read, which is one reason why I tried it (think it was on Gearslutz). The reason behind this probably have more to do with filter types and technical design than anything else. Monark from Native Instruments explicitly states it's made to be run in 88.2kHz mode or higher.

Oversampling is also used in many plug-ins. The fact is that it can be better, in at least some cases, from a resource standpoint if your signal chain is using a higher rate to begin with. But this is far from certain and will be affected heavily of the design of the DAW and the devices and plug-ins that you're using. I've heard from other producers that think their DAW works better with this.

Personally, I've so far opted to run all synths that I can in 88.2kHz output mode, but keep 44.1kHz in my project, as I haven't done any evaluations on any possible benefits of upscaling my projects. Any comparisons would have to be made in a final 44.1kHz product.

I maintain my belief (haven't tested) that a well-made 44.1kHz music product in a fairly high quality system will be on par or better than any 88.2kHz or higher sample rate sound-wise. It could be of course that some DACs actually work better with higher sample rates, but I'm not so sure about this.

At any rate, practice beats theory. Theories not proven by actual experiments are just theories.

[Angstrom]

I just remembered this T-Shirt I designed for the forum, way back in the distant past

[miekwave]

96k/24 has most benefit on highly dynamic recorded material, where if doing EDM, 22k is fine
96k/24 produces less artifacts going into any dynamic or frequency based DSP
96k/24 provides more fine resolution time stretching
96k/24 provides a vastly different character that 44k on many 'guitar amp' plug ins

[theophilus]

That's mostly what I was referring to when I said "modern DSP is so damn good". Usually when there might be an issue that affects audio quality, oversampling is employed anyway.

andy//cytomic said once that most of the cheaper oversampling filters aren't linear phase... meaning they introduce a frequency-dependent phase shift that can't be compensated for with PDC. so if you use a lot of parallel processes, you may get fewer phase issues with running at higher frequency vs. oversampling.

also, oversampling itself is a process that uses CPU (albeit a small percentage). so, in some cases, if nearly everything is oversampled anyways, you may actually get better CPU by running at a higher rate and not oversampling (though this is unusual, i admit) - after all, the oversampled plugin is still processing at the higher frequency anyways, you're just adding two extra stages to step the frequency up/down at the end that you may not need. and if the plugin uses a low-quality oversampling algorithm, it may sound better too (but there are even open-source high-quality oversampling algorithms, afaik, so this shouldn't be as much of an issue nowadays).

last, while 44.1KHz is fine for _reproduction_, we don't just record and play back audio... we run synthesizers etc. that are generating it, effects that process it, etc. many (maybe most?) of these processes are relatively immune to the frequency, but many others, especially when you get into component-modeled stuff, use feedback at some level. and since digital is not continuous, the digital algorithms are basically approximating what happens in the analog world - a higher frequency should lead to a better approximation, as i have more samples to look at. Also, for things like transients, while it's true that the end waveform will be exactly the same whether at 44.1KHz or 192Khz (if it's <20k at least), a processor looking at the waveform gets more detail at 192k than it does at 44.1k, and so even with the same algorithm may make slightly more accurate calculation, be able to respond in fewer samples, etc.

there is a thread on gearslutz also called 'the great vst analyzer thread' or something. on there, you can see that no, aliasing is not really a thing of the past...
https://www.gearslutz.com/board/music-c ... ad-85.html

[ZaBong69]

Hi,

is somebody here using 44.1k first during production and then 96khz only for rendering? I keep asking myself whether this isn't the best of both worlds - keeping processor load low most of the time and going for the highest possible quality just for the final step. I do not use that many samples, so switching between sample rates has few if any disadvantages it seems...

What do you people think?

K

[Angstrom]

Well, there might be a few issues with Upsampling.

Live has a fancy newish algo for downsampling, so if you load a high rate sample (eg 96khz) and play it back at 44.1khz the SOX algo does a great job at downsampling (high to low)

Unfortunately upsampling is less amazing. If you record a sample at 44.1khz and upsample to 96 then you are asking for a lot of interpolation.

Sure your effects might sound a little bit better with the shelving and aliasing raised way out of your hearing range ... But mostly that stuff is handled pretty well in each device, and if you heard a saturator throwing aliased harmonics down at you (sounding digital) you'd have resolved it by ear pre-render.

So all you are really doing is upsampling by a non integer value.
If you rendered at 88.2 it would at least be a simple 2x, and live just guesses every alternate point in the sample and keeps all your original sample points

But upsampling to 96 is not the same. That means asking Live to guess at in-between samples and then ALSO to throw away your originals. Do you get what I mean?

So tl;dr NO. I wouldnt do that

[Tarekith]

I wouldn't do it with Live. It's sample rate conversion when upsampling still uses their older SRC which is... so so. It's only when downsample that Live uses the MUCH better SOX algorithm. I have no idea why.

So personally I would just keep things at 44.1kHz for the render too, or work natively at 96kHz if you really need to go that high.

[SLEEarts]

I use 96 khz pretty regularly for several reasons:

1. I think it sounds ever so slightly better, especially in projects where a lot of processing happens. HD space is cheap, and I don't want to compromise sound quality at all.
2. There's already a small market looking for these high resolution files. If this market were to grow, I'd like to be able to sell my catalog to them.
3. Take aliasing out of the equation, more or less.
4. Future-proofing is always a good idea.

[Division Monarchy]

After reading more replies, I think I'm going to go 96k on my next album. Reason being, it's not EDM and I'm going for a more organic sound with dynamics. Plus I'm getting into the Monark synth and was not aware that it was meant for high sample rates.

Plus, nobody listens to my music anyway so I don't have to cringe at the thought of doing something in hi-fidelity only to have it butchered by compressed internet audio.

[ZaBong69]

I wouldn't do it with Live. It's sample rate conversion when upsampling still uses their older SRC which is... so so. It's only when downsample that Live uses the MUCH better SOX algorithm. I have no idea why.

Thank you for this insight... I am currently using 88.2 all the time, but 90% of my sounds come from synths, not samples. As a bedroom dance producer, the occasional "Hey", "Yo" and "Move it" snippet drenched in reverb is enough anyway to keep things going, so the very minor quality loss of upsampling (if any in this case) can be ignored.

If I was using a lot of actual percussion samples or (shudder) real vocalists, things would be different...

K

[jlgrimes]

After reading more replies, I think I'm going to go 96k on my next album. Reason being, it's not EDM and I'm going for a more organic sound with dynamics. Plus I'm getting into the Monark synth and was not aware that it was meant for high sample rates.

Plus, nobody listens to my music anyway so I don't have to cringe at the thought of doing something in hi-fidelity only to have it butchered by compressed internet audio.

Concerning Monark,

By default it automatically oversamples to 88.2 khz (or 96 khz if you use 48 khz). So in a nutshell you don't have to go to 96Khz to get the benefits of Monark.


Higher sample rates is more for plug-ins without oversampling, wheras many plug-ins oversample by default, while others allow you to select how much oversampling you want, and while alot of others don't use oversampling.


Also keep in mind that while 96k or 88.2 khz often does make the plug-ins sound cleaner, some plug-ins were intended to sound their best at 44.1 khz even if using no oversampling. I worked at 88.2 at one point and while alot of plug-ins did sound better especially stock DAW plug-ins (I was on Sonar at the time), other plug-ins lost their oomph factor.

[jaynyc]

do you guys *not* think that Live audio effects and time stretching (warping) sound better at 96 vs. 44.1?

At least for creation 96 sounds noticeably better to me.... then use an offline tool for 96-->44.1 conversion for sharing/distribution.

[Tarekith]

I have oversampling turned on by default for the Live devices that benefit from it, so not much difference for me there. I haven't experimented with how different warp modes sound at various SR's, that's definitely something to experiment with if people are curious about this stuff.

[timday]

You need it if you're using Roland Aira gear because it only runs at 96Khz.

Thanks for that, Roland.

[Vathek]

it's like taking a picture with a higher resolution... pixels are smaller and thus you can zoom in deeper while keeping lots of detail.

This is actually plain false and a lot of the confusion about this topic stems from exactly this misleading analogy. 96kHz does NOT lead to a finer resolution of audio than 44.1kHz or whatever, it only captures a larger range (half of which is outside of the range of the human ear). For recording/playback of audio it makes not much sense to use higher sample rates at all, Dan Lavry who has been mentioned here before, explains that pretty well.

For digital music creation or heavy manipulation it might make sense, depending on the circumstances. One beneficial point worth mentioning about higher sample rates is that they reduce the latency of the system.

The topic is really rather irrelevant in the grand scheme of things and many people do this or that for the wrong reasons because it's supposed to yield "finer resolution". Like guys sampling their records at a bit depth of 24bit, when vinyl only has a dynamic range of about 70dB max.