44.1 or 48kHz?

[lofty bridge]

Which one do you choose and why?

[narrow arrow]

48khz, I guess that a bit extra oversampling may help...

[proven violet]

Short version: 44.1 for album (CD audio standard) 48 for soundtracks (DVD movie audio standard).

These would be more of a guidelines, rather than rules.

Also, if you want to minimize aliasing (and get more of that "analog" feel) and you hardware allows it, you can consider having you project at 2x or 4x whichever you choose, and export to either 44.1 or 48 kHz at the very end. This way any effects, that add harmonics will have a bit more of headroom at the cost of processing power required. You can do it if you start having problems with aliasing, or just experiment and see if you hear the difference (if not - no reason to run the project at higher rates).

[little token]

48

[normal terrace]

48khz... little more cpu impact, lower latency.

[elder warren]

48khz atm

[tropic jackal]

48khz can be helpful for recording analog audio since the Nyquist frequency is further away from our hearing range, and your audio interface will therefore have a less steep (more natural) filter curve to deal with aliasing.

I've gone back and forth between 44 and 48 over the years. I worked at a lot of major recording studios still using 44khz, so basically don't stress out too much about it. @normal terrace makes great point about other trade offs!

[vivid musk]

essentially, it takes more thought and theoretical research than its worth for the pay-off of knowing which one is better for your particular circumstance. so just pick one.

[brisk oxide]

48kHz for about a few years. Prior to 2019-ish, I've recorded in 44.1 just fine. I switched to 48kHz to "intellectually" get better recordings, but honestly I tweak audio so much that I don't get to hear the difference.

[marsh citrus]

48 khz. A good explanation why 48 is almost "twice as good" as 44.1 khz is explained by FabFilter.
48 khz is the best compromise between CPU consumption and audio quality.
https://www.youtube.com/watch?v=-jCwIsT0X8M

[vivid musk]

Explained by Dan Worral*
I have learned a lot from that dude. Look up his own youtube channel.

[marsh citrus]

Thanks for the info!

[sly crown]

96

[plain flower]

48. I used to use 96 until I found some plugins that didn't work properly with such high bitrate

[winged pelican]

Not that much love for 96 and above?

[normal terrace]

I personally go to 192 only to play Pianoteq. The latency becomes so unnoticeable that it seems to play a mechanical piano. Gorgeous…

For the rest, actually I stick to 48. I can’t really hear the benefits of havng higher rates, probably ‘cause my age. So the added CPU impact doesn’t give me any good, but for sure it’s a subjective experience.

[hazy drift]

Mostly doesn't matter.

[vagrant ivy]

i always like hearing that. even though i don't believe it and still continue to work at 44.1k 🙂

[hazy drift]

Let me put it another way, there's probably other technical problems in your production that are much bigger than the difference between bit rates. There surely are people who have the compute power to run at 8x oversampling and master and then dither and who claim to hear stuff that not even owls or bats can hear.

[vagrant ivy]

the dither monsters are coming for you beware they only listen to classical

[marsh citrus]

I would love to work in 96k if it wasn't for the problem with the high CPU load....

[normal terrace]

As said, personally I don't hear any improvement going over 48 if not lower latency. Do you?

[marsh citrus]

I definitely do hear a difference with 96k e.g. with the amp simulation in Saturn 2 where a lot of aliasing happens or with some Softube compressors, where the dynamics sound even different e.g. Zener-Bender. This difference is still audible when you render it down to 44k.

[normal terrace]

What if you run at 48khz and use the plugin oversampling capability, if any?

[marsh citrus]

there is still a difference

[vagrant ivy]

What type of material do you hear the aliasing on?

[hazy drift]

Can you hear it if you blind test?

And can you see it if you null it?

[normal terrace]

That's interesting, actually.... But how can you null it with different sample rates? Just asking myself if that can work...

[marsh citrus]

Yes, you can hear it and see it in a null test. The difference is bigger when a 96k processed signal (rendered down to 44k) is nulled to 44k processed signal compared to a null of 96k (rendered down to 44k) and 48k processed signal (rendered down to 44k) in case of the Softube Zener-Bender EQ and compressor. It has something todo with the DSP model of the Zener-Bender which is calculated more accurate? with higher sampling rates. Somebody on Gearspace experienced the same.

[normal terrace]

And that plugin doesn't have an oversampling option, I guess.

[marsh citrus]

no

it has fixed internal oversampling

[sly crown]

Bazille and Zebra both sound better at 96khz. It is a distinct difference

[vagrant ivy]

Yo would u mind linking that thread. Curious about it since I have the zener

I’m still open ears for 96k… just my greedy cpu doesn’t want me to be there

[hazy drift]

Trying this now.

[normal terrace]

response???????

https://tenor.com/view/im-waiting-daffy-duck-impatient-gif-16985061

[hazy drift]

I can't hear it.

Of course I'm over 50 and my high end hearing ain't what it used to be.

[vagrant ivy]

yeah man... i think a lot of these comparisons make a difference around the nyquist. which is well many kilo away from what i can actually pick up

[sly crown]

It's not just on the high end... (I'm over 50 too) It does depend on the patch in Bazille. For example, audio rate modulations 96khz makes a difference and in some cases a dramatic difference. If you just listen to a saw wave, then they will be effectively the same

[hazy drift]

I'll do a bit more experimenting this weekend

[vagrant ivy]

when running in 96 does bazzile double it in high quality mode? i wonder

[hazy drift]

It uses more CPU for sure

[sly crown]

I don't know what the HQ mode does in Bazille.

[vagrant ivy]

Still don’t know here either lol

[sly crown]

Exactly... hehehe

[vagrant ivy]

Now if only bitwig had oversampling plugin chains

[sly crown]

I'd welcome that... then I could keep my projects at 48khz

[vagrant ivy]

Yeah! Would be nice for those certain plugins that require higher sample rates.

[vagrant ivy]

The zener is so cpu intensive at 96. I need to figure out how to work it in. I love how it pulls everything together and adds that bit of punch and shine to a mix

[vagrant ivy]

something i found with the model 82... rendering at different sample rates. I did a test with automation. and at the high modulation levels here the sound pretty much goes silent at 44.1k at this point.

[ocean jetty]

48 because I work with video

44.1 saves you some cycles and gets rid of any SRC issues with 44.1 samples, but to me the alignment issues with video or resampling always aren't worth it. Get the bonus of less EQ cramping and better AA filters due to slightly higher nyquist further out of audible band. Most plugins I use oversample or do some type of better interpolation or integration technique for reducing aliasing in processing anyway so 96 seems pointless to me outside of creative sampling for repitching later or lower latency monitoring when tracking.

[plain flower]

I use 48 just because i do. It's a habit. I don't believe it's the best idea or recommend it or anything

Just a kind of anality i think. 48 is a better number

[thin torrent]

for music releases 44.1 and for video or game soundtracks 48

[sly crown]

Some of my favorite plugins sound better at 96... so that is why I use 96.

[ocean jetty]

Yeah I wish Bitwig had a resampler container you could drop on plugins

Reaper can oversample plugins, but I feel like a container you drop them in with an upsample input and downsample output would be nifty

I'm curious is that an aliasing problem or a blocksize issue you're hitting with those?

[sly crown]

Aliasing mostly... with Bazille, I am always doing lots of audio rate modulations

[ocean jetty]

Ah, yeah, that always needs some extra samples. I'm surprised Bazille doesn't have an internal option for it

I haven't actually explored a lot of Uhe stuff yet myself

[vagrant ivy]

Send feature request!

[sly crown]

Bazille is my favorite synth, software or hardware. That is why I am willing to run my projects at 96.

[cerulean sparrow]

Always 48khz. CDs and 44.1 quality are a thing of the past.

[trail marten]

Interesting. I love Pianoteq, but never thought that sample rate could change something that far. Does it mean that you can achieve lower latency than with 48 "de facto" ? How to say that : with the same buffer size, you get lower latency in 192 than in 48 ? The extra cpu load is "real time cpu", like the one that gives xruns, as in " lower latency = xrun", or global cpu ? (If you now Reaper : cpu vs rt-cpu)

[normal terrace]

Same buffer length and double sampling frequency means half the latency. It's simple arithmetic. If your hardware can afford it, the studio version of Pianoteq allows 192khz. Don't take my words for granted, try.

[trail marten]

I have the standard version and it doesn't allow 192, too bad. By the way : do you know a way to edit the instruments that are shown in the device windows (in the device panel, not in Pianoteqs UI). I can't figure how did Bitwig decide to list such or such instrument.

[normal terrace]

Mmmm... Looks like it's the first preset for each instrument...

Oh, I don't see anything from guitars... and nothing from karsten, kremsegg, CP80, predecessors, etc...

I currently have no clue on how to update/change that list and how it gets filled up... Can't be of any help on that... sorry

[wary yarrow]

Whenever I see those lists it's only for VST2s, I think the fxp preset format is related, but I can't remember how I learned that. You might look to see if there .fxp or fxb files in your content location paths, just a guess, or maybe the vst2 directly communicates these to the host?

[normal terrace]

@wary yarrow Mmmm... I have it for VST3 too, like Discovery Pro and the others by discoDSP. And in the filesystem I can find an ".fxb" that contains a "bank"... when I select a bank from the interface, that list gets populated with all the entry from that bank...

But other have fxb and don't show that... and other... I don't know... Looks a mess to me... lol

[wary yarrow]

Hmm very weird, makes it even more confusing that vst3 .vstpreset files just appear in the browser. I guess this is some sort of backwards compatible mess that steinberg created

[cerulean willow]

48khz bit better against anti alising and lower latency… down grade to 44.1 possible everytime.

[cerulean ledge]

should we mix at 88k or greater than Bounce the MASTER at 44k ?

and how can we boost our computer power to be able to mix easelly at 192khz ? lets say you have 20k budget for a computer

is having 2 or 3 computer connected sending to 1 mixer can make this happen ?

that could cost less than 6k

3 times 7950x

does each track in bitwig use 1 core ?

[cerulean ledge]

My real question here is that - IF you find a way to make your music at 192KHZ , mix - master and all , will you be able to do/hear things that other cannot do on 48KHZ ?

like , Using Serum at 48KHZ with 4 X oversampling compare to using it at 192KHZ with now oversampling , will you be able to get a difference in sound of the REVBERB , DISTORTION ? and will that be audible on your bounce MASTER at 44.1KHZ - 24bit wav file ?

[cerulean ledge]

Just did a blind test and With all the pluggins I have ( sonible , soundtoys, Plugging alliance and more ) And I could not find the difference between Me bouncing the mix to 44.1KHZ 24bit from a 48KHZ or a 96KHZ source.

I guess ill mix in 48KHZ for the moment.

[grand obsidian]

regarding what you said earlier, i think upsampling is exactly that btw. so even if you use 44khz, then using upsampling, it will process it as if it was 48khz and then down again to what project khz is set to at

though it usually does it in power of 2, so i guess 88khz if project is 44khz

[cerulean ledge]

From what I learn - to mix at 48KHZ is better than 44HKZ , just because 44 is too tight for the system and the Requirement needed for the 2X loose potential in that area just worth it.

[rich geyser]

Considering that most samples around the world are 44.1 when working with 48 some tricky upsampling must be done. Is Bitwig any good at that? I think it's safer to stick with samplerate of your samples if you use samples at all.

[cerulean ledge]

what samplerate splice use / how to know what sample rate our sample is ?

[safe lichen]

I see a lot of people favoring 48kHz over 44.1kHz simply because of the higher Nyquist frequency, so less aliasing. Which is true, but why is nobody talking about the aliasing and noise you introduce when downsampling from 48 to 44.1? Because that is not easy to do and is quite destructive.

[harsh plank]

One thing I'm not sure was mentioned yet is that working at 96khz or above is not automatically higher quality, depending on the use case. When you process anything in a nonlinear way (which happens in many types of plugins), it has the potential to introduce IMD, where distortion adds unmusical harmonics at a lower common denominator of two higher frequencies. Working at 96khz means you are allowing IMD to continue to build up in your processing chain at frequencies above Nyquist, which can eventually creep down into the audible range. If you work at 44.1 or 48khz instead, this source of IMD will never affect you, because it is never generated to begin with. This is why it can be better to work at 44.1/48 and let individual plugins oversample internally when they need to (which most good plugins do if that's important, like for compression or heavy distortion).

I think Dan even covered this aspect in that video

[half crane]

96khz because I'm running everything into a clipper

[hazy drift]

I work at 44.1 because I want MAX dsp capability on my pc

[vagrant ivy]

i love all these reasonings. keep em up friends ❤️

[ocean jetty]

Proper resamplers can deal with non-integer quite well these days so it's sub-audible

and since this should only happen once (at playback on someone who's running at 44.1, or when transferring to CD if doing so) this is inconsequential

44.1 only really exists for CD at this point, no other distribution method uses it (Movies, games, etc.)

[safe lichen]

Not only CD, Spotify is 44.1 exclusively still. Which is (I think) the number 1 target for modern producers. But my point was, why even bother with 48 if it is going to be converted to 44.1 anyways?

I get it for anything that is sync'ed, like video and multimedia (gaming f.e.). It makes sense to choose a samplerate that is easily contained in the frame data or stream, thus a multiple of the framerate, which is often 24 or a multiple (which 48kHz is, and 44.1 is not), or also often 60fps (both 44.1kHz and 48kHz are a multiple).

[harsh plank]

Yeah, I thought most streaming platforms were 44.1khz in general. Youtube is 44.1, too

I've generally been making 44.1khz masters with a separate 48khz/16bit for soundcloud only

[ocean jetty]

I dunno I only buy stuff off bandcamp or stream FLAC from services that support it. Not too concerned about sinc resampling

although to be fair Bitwig does some weird minimum phase resampling of 44.1 that isn't great

Personally, Spotify is where people hear stuff on their phones and is generally not great quality (There are a lot of bad sounding album releases where the CD is significantly better than what hit the DSPs, google play music had the same issue where some of it sounded like a bad mp3 compared to the disc) and if I'm putting out content for consumption it's more than likely going to be video synced