KVR Audio

V0RT3X · Post by **V0RT3X** » Fri Nov 27, 2015 1:23 am

I've been wondering if i should be working at 96khz for better fidelity.

Normally i just run at 48khz, 24bit on all my projects.

What do you use?

phreaque · Post by **phreaque** » Fri Nov 27, 2015 1:31 am

48khz @24bit because it is fair enough.

Googly Smythe · Post by **Googly Smythe** » Fri Nov 27, 2015 2:33 am

Quoting from a topic in the SOS forum by Hugh Robjohns, the SOS tech editor:

"If you want an audio bandwidth up to 20kHz, then you'll require a sample rate of slightly more than twice that, hence 44.1kHz. We (humans) can't hear more than 20kHz, but in some circumstances there can be small but worthwhile technical and practical advantages in working with a higher sample rate, hence 96kHz. There is little evidence to support higher sample rates... but that doesn't stop manufacturers offering them or people claiming magical qualities for them!"

Link to the thread:
http://www.soundonsound.com/forum/showf ... =1#1175015

chk071 · Post by **chk071** » Fri Nov 27, 2015 3:04 am

So, basically, saying that running synths at a higher sample rate than 44.1 kHz, because they "sound better" then is bogus?

sprnva · Post by **sprnva** » Fri Nov 27, 2015 3:16 am

I run 96kHz here. It's definitely noticeable over 44.1. Not only in added detail, but also how some plugins behave.

aciddose · Post by **aciddose** » Fri Nov 27, 2015 3:44 am

Read existing threads on the topic.

For example: http://www.kvraudio.com/forum/viewtopic ... 1&t=444362

It would be nice if someone had the time to write an article on this topic that could be simply linked as the first and only reply to any questions about aliasing, oversampling, sample rates and so forth.

Short version:

Anti-aliasing filters have an effect in the pass-band that can be moved out using a higher sample rate.
Using oversampling (a higher sample rate) is an incredibly inefficient method of anti-aliasing. It is the worst possible method apart from none at all.
For many reasons oversampling slightly can give you more accuracy. For example typically infinite frequency is mapped to nyquist in discrete processing. In these cases oversampling is not efficient but it is the only method by which you can reduce the distortion of the frequency response due to this warping.

You have to determine based upon a case-by-case basis whether working at a higher rate is worthwhile. It may be, it may not be. Quite likely it isn't worth it.

In most cases using plug-ins with accurate coefficient calculations and good anti-aliasing, rendering at a high rate such as 1mhz will produce no audible difference. In cases where you use naive plug-ins with a lot of aliasing it might be entirely worth it.

Don't use your ears. You can go ahead and use them first, but don't accept your own poor judgement (like anyone) of whether there is any difference. Ears are garbage and can't be trusted. Take a measurement that is objective and accurate so you can understand exactly what the difference is, if any. If you can't measure and identify a difference that's the same as if there isn't one.

Squidsneeze · Post by **Squidsneeze** » Fri Nov 27, 2015 4:38 am

Hi,

well, one can't say "higher sample rates are better". There are many situations where this simply isn't the case.

But there are some pro's I experienced while developing plugins. For example delay times are more accurate. I tested some delay plugins with 100% feedback in different sample rates. After a short period of time I noticed big timing differences. Hardware effects like algorithmic reverbs often process in much higher sample rates just because of this.
Higher sample rates tend to preserve better stereo image information.
Also, naively said: compressor and limiter are waveshaping effects and at higher sample rates you'll notice less distortion and/or aliasing.

But you should also think a step further. Once you have a hq 192kHz audio file and your target medium needs it to be 44.1 or 48kHz, you have to convert it down. Processing at higher sample rates is pretty useless, if you convert it with a lousy tool - and many daw just have converters with less good results.

Unlike aciddose I'd like to say: trust your ears. Nobody wants you to use higher sample rates (unless you work for some broadcasting companies or film industry). So it's up to you and the differences you can hear. If you "feel better" using 96kHz, do it. If not, don't do it

Best,
Danny

camsr · Post by **camsr** » Fri Nov 27, 2015 5:17 am

I've been using 96khz for a while. I wish my computer was faster too

keyman_sam · Post by **keyman_sam** » Fri Nov 27, 2015 5:59 am

What about working in 48 KHz and rendering in 96 KHz? Any advantages to that approach?

The one shortcoming is that all the synth stuff needs to be in MIDI only, otherwise audio gets resampled with no benefits. Thoughts on this approach?

Also in 96 KHz you get lower latency for the same # of samples. Unfortunately my computer can't even do 60 tracks in Live without stuttering, so this is either a pipe dream or I use Reaper (which I don't want to atm)

Tarekith · Post by **Tarekith** » Fri Nov 27, 2015 7:39 am

I do everything at 44.1kHz. While there's been times I find 96kHz sounds different (ie, aliasing softsynths back in the day), it rarely sounds any better to me. I personally like Dan Lavry's thoughts about how anything over 60kHz is a total waste.

lfm · Post by **lfm** » Fri Nov 27, 2015 11:11 am

I use 48k in projects but target certain synths with 96k through Metaplugin with oversampling activated. You benefit getting less aliasing, as just about all synths/samplers do resample for every note pitch.

In Sonar you can also use this on per synth basis, without any special wrapping plugins. Now it works by just marking the plugin to upsample, and works both for realtime and rendering.

Just targeting specific plugins/chains is way to go, I think. Best of two worlds. Depending of complexity of sound generated the improvement is more obvious.

I did a test last spring with Dimension strings.

So oversampling = all improvements in Metaplugin.

You can check it out here:
DP alone 48k
DP+MP 2x48k and back

Second part one octave up - there it is really obvious. The nasty stuff is down 2-3dB or something.

Then imagine having dozens of synths running in a mix, and adding some clarity and less harsch - what that does to the full picture.

Ichad.c · Post by **Ichad.c** » Fri Nov 27, 2015 10:20 pm

Been using 24bit 48kHz for nearly ever, no problems. That extra couple of kHz comes handy when you use outboard, because no 'brickwall-filter' actually exists(except for marketing-people), so at 44.1kHz you might loose a bit of highs depending on how many trips you make through the A/D D/A converters.

What acciddose said is spot-on.

Jace-BeOS · Post by **Jace-BeOS** » Sat Nov 28, 2015 12:33 am

I think higher sampling rates are interesting for sampling sounds in the real world at high resolution just to slow them down for special effects and sound design (you can possibly reveal parts of the sound you couldn't hear as those frequencies are brought into human hearing range). Other than that, I've no use for higher rates.

I simply haven't felt like using the extra drive space for audio at resolutions higher than 44.1KHz/16-bit. I think Logic defaults to 24-bit when it records audio (?), but I don't mess with the settings, because, aside from the storage concerns, I find that using anything outside the defaults exposes behaviors untested by developers (in all software; life on Windows sort of taught me not to customize anything anymore, and Apple is showing a similar trend these days, so I avoid customizing much of anything if I can stand defaults).

When I used Sonar, I explicitly used 44.1KHz/16-bit after a few projects running at 48KHz behaved unreliably. Sonar and audio device behaviors were annoying... "mismatched sample rate" messages when importing/exporting or dragging and dropping wave files into the project (would've been nice if Sonar would've just converted source to target transparently!), projects suddenly at the wrong pitch because the audio device didn't switch to 44.1KHz, digital connections and clock rate mismatches, etc. The only reason I even tried it was that my Alesis QS8 defaults to 48KHz and I used the ADAT output (until discovering it could just easily be switched to 44.1KHz without loss of quality that I could tell).

When I've tested things like Reaktor at higher sample rates, I've noticed many ensembles break (because again few people test all options) and the CPU cost is way more than the sonic improvement is worth to my ears.

The only thing I regularly used outside the "CD quality" realm was Sonar's 64-bit mixing engine, because it has noticeably greater headroom. But that doesn't impact audio file size/storage.

Lots of my favorite music was recorded at 16-bit/44.1KHz (because that's all they had at the time), and I don't have the skill to make my stuff sound as good as that stuff, regardless of having higher sample quality potential. What's the point?

zerocrossing · Post by **zerocrossing** » Sat Nov 28, 2015 1:07 am

I quickly checked out how things sounded when I got a new audio interface. It's got on board DSP as well so I thought, "what the hell? I may as well use the extra horse power."

What many say is correct. If the plug in is doing what it's doing correctly, you shouldn't have to run things at 96khz and, in fact, running them at that rate doesn't give you any audible advantage. Some things did sound a bit better, but overall the tax on the system (I like to run things in real time) just isn't worth it. Some plug ins, like Reaktor, let you run things at higher sample rates than the project itself is running at and things like audio rate modulated sounds do sound a bit better, but again, you pay a price.

V0RT3X · Post by **V0RT3X** » Sat Nov 28, 2015 1:21 am

Some good replies in here, thanks.

Cheers everyone

What sample rate do you work with?