Sample Rate Discussion (off of AITB Germanos EQ Thread)

If you need to write about anything else please do it here..
User avatar
yr
Vip Member
Vip Member
Posts: 438
Joined: Mon Mar 29, 2010 9:04 am
Location: Amsterdam

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by yr » Thu Feb 17, 2011 1:52 pm

Why would you care if someone else can hear the difference or is interested in subtleties?- as long as you can (or can't) guess the "right" answer using a proper ABX test you have the information you need. No need to argue about what constitutes a meaningful difference (which is rather subjective).

Personally, the subtle and complex ways Nebula can effect sound is exactly the reason I love it. Some people say they can't hear the differences between the preamp presets or that the consoles (and even R2R) are way too subtle. So what- if you like the sound and can hear the effect you use it. I've had enough cases were I couldn't exactly hear what a preset was doing in an ABX test and decided not to use it. No big deal, and no need to declare the rest are imagining.

If you decide to make a poll, you should probably use a typical Nebula chain (multiple instances) for testing. The results might be more realistic and clearer (depending on the source material and presets you choose). I find the comparison as done by Alexb to be problematic for reasons I've already mentioned.
Reuven | post-production & sound-design | scenography |
website | nebula presets

Stolle
User Level VI
User Level VI
Posts: 68
Joined: Mon Mar 29, 2010 3:27 am

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by Stolle » Sat Feb 19, 2011 6:36 pm

Ok people. I finally found an hour off for some testing. Here is what I did. I fired up my analogue synth (DSI Mopho) and let it play a static sequence (On the run). I then recorded it both in 96 khz and 44.1 khz 24 bit. I then added a Nebula chain of AlexB VBC (line2, clean buss, vintage mixbuss).
I then exported the two tracks with Nebula and bypassed to 44.1 khz 16 bit with dither.

Here is the result:
44.1 bypassed Nebula - http://www.parentes.nu/MophoRun44bypass.wav
96 bypassed Nebula - http://www.parentes.nu/MophoRun96bypass.wav
44.1 with Nebula - http://www.parentes.nu/MophoRun44NEB.wav
96 with Nebula - http://www.parentes.nu/MophoRun96NEB.wav

The results to my ears? I can't tell any difference between 96 and 44.1 without Nebula. With Nebula I hear a slight increase in distortion in the 44.1 -version. The distortion is different somehow. I must admit though, I'm not sure I could tell which is which in a blind test. I did find out however that my computer can't handle many Nebula instances at 96 so I will stay at 44.1 khz. I haven't decided if the difference is big enough to consider upsampling just for Nebula rendering and such. What do you think? Can you hear the difference? Should I've done my test in another way? Would the difference be more apparent on other material?

All the best
/Kristoffer

Mplay
Member
Member
Posts: 294
Joined: Fri Aug 20, 2010 10:25 am
Location: Willemstad, Curaçao

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by Mplay » Sun Feb 20, 2011 1:18 am

I stumbled on Dan Lavry's paper on sampling. He states that the benefit of working at higher sample rates is the fact that artifacts caused by digital filtering occur at higher frequencies, further from audible frequencies. Working at 48kHz already has benefits over 44.1 (and 64kHz is the ideal sample rate, any higher has no benefits)

:dive for cover: :D

TranscendingMusic
Expert
Expert
Posts: 1253
Joined: Sat Mar 27, 2010 7:01 am
Location: USA

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by TranscendingMusic » Sun Feb 20, 2011 3:24 am

mplay just to correct that slightly...
The said optimal rate of 60Khz is actually deemed as the "most optimal" (but 44.1 being just as suffice as a standard for now) for audio capture. Lavry's main point about moving artifacts out of the audible range fall under oversampling. Further it has been stated that for the oversampling to have a true benefit needs to be some where around 300Khz. Where as using 96 for example you indeed get something different than 44.1 but that's just it; it's different not necessarily better - this also accounts for why in some blind tests listeners would prefer the 44.1 over the 96 WHILST comparing to the original because the 96 version or the processes therein change the top end so much. But this is all "on paper"; in technical theory.
Cubase Pro 10.0.60 x64
Ryzen 2700x | Win 10 x64

mathias
Expert
Expert
Posts: 2731
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by mathias » Sun Feb 20, 2011 4:14 am

here is another test:
i recorded a little improvisation with two tracks guitar and one bass-track.
i recorded in 48 khz as i normally do.
i preprocessed the takes a little bit, to get the the sound of the instruments a little more balanced.

now the test:
i put a globeconsole directout on every instrument and a globeconsole-masterprogram on the masterbus.
the guitars have each a reverb (non-nebula).
i bounced one version with the 48 khz takes and made a second identical mix with the takes upsampled to 96 khz and the nebulas loaded in 96 khz.

the original mixdowns are in 32bit-float.
i brought both mixes down to CD-Format then.
here the links:
mix 48 khz 32bit float:
http://dl.dropbox.com/u/9435398/Nebula- ... -48khz.wav

mix 96 khz 32 bit float:
http://dl.dropbox.com/u/9435398/Nebula- ... -96khz.wav

mix 48 khz in CD-Format:
http://dl.dropbox.com/u/9435398/Nebula- ... Format.wav

mix 96 khz in CD-Format:
http://dl.dropbox.com/u/9435398/Nebula- ... Format.wav

download with "rightclick, save target as...", and open in your daw. the player in your browser could degrade the sound.

in this case there are no long chaines involved, which will certainly bring other results.
maybe i extend my test, when i have time.

but it is still interesting enough.
as already stated here by others, i can not really find the 96 khz version better sounding, when brought down to 16 bit 44,1khz. there is another sound in the top end of the spectrum, that is for sure.
if i could stay with 96 khz, i would prefer it
over 48 khz.

have fun, (i hope dropbox will not stop too fast, when a lot of people download :-))

mathias
system 1: windows 8 32 bit - samplitude pro x3, waveform 9 (tracktion), reaper
system 2: mac osx yosemite - reaper(64bit), waveform 9 (64bit)

both systems on: macbook pro (late 2009), core 2 duo 3,06 ghz, 4 gb ram, graphic: nvidia geforce 9600M GT 512 MB

Martinez
Member
Member
Posts: 293
Joined: Mon Aug 09, 2010 2:03 pm

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by Martinez » Sun Feb 20, 2011 6:32 am

I found this interview with Rupert N**e in which He talks about sampling rates and His thoughts on the subject.

I thought You all might find it of some interest in relation to this particular thread.

http://www.poonshead.com/articles.html

Mplay
Member
Member
Posts: 294
Joined: Fri Aug 20, 2010 10:25 am
Location: Willemstad, Curaçao

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by Mplay » Sun Feb 20, 2011 11:42 am

TranscendingMusic wrote:mplay just to correct that slightly...
The said optimal rate of 60Khz is actually deemed as the "most optimal" (but 44.1 being just as suffice as a standard for now) for audio capture. Lavry's main point about moving artifacts out of the audible range fall under oversampling. Further it has been stated that for the oversampling to have a true benefit needs to be some where around 300Khz. Where as using 96 for example you indeed get something different than 44.1 but that's just it; it's different not necessarily better - this also accounts for why in some blind tests listeners would prefer the 44.1 over the 96 WHILST comparing to the original because the 96 version or the processes therein change the top end so much. But this is all "on paper"; in technical theory.
Thanks for the info. I have to say my personal reason to leave 44.1kHz has nothing to do with Nebula SRC. It was the fact that I wasn't able to capture my favorite softsynth ACE (by U-he, which uses a lot of oversampling) recording at 24/44.1. The rendered files just sounded different. Working at 32/48 solves this problem without taking too much resources.

I couldn't find any major problems using my 44.1kHz or 96kHz programs working at 32bit/48kHz.

Stolle
User Level VI
User Level VI
Posts: 68
Joined: Mon Mar 29, 2010 3:27 am

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by Stolle » Sun Feb 20, 2011 2:46 pm

mathias wrote:here is another test:
i recorded a little improvisation with two tracks guitar and one bass-track.
i recorded in 48 khz as i normally do.
i preprocessed the takes a little bit, to get the the sound of the instruments a little more balanced.

now the test:
i put a globeconsole directout on every instrument and a globeconsole-masterprogram on the masterbus.
the guitars have each a reverb (non-nebula).
i bounced one version with the 48 khz takes and made a second identical mix with the takes upsampled to 96 khz and the nebulas loaded in 96 khz.

the original mixdowns are in 32bit-float.
i brought both mixes down to CD-Format then.
here the links:
mix 48 khz 32bit float:
http://dl.dropbox.com/u/9435398/Nebula- ... -48khz.wav

mix 96 khz 32 bit float:
http://dl.dropbox.com/u/9435398/Nebula- ... -96khz.wav

mix 48 khz in CD-Format:
http://dl.dropbox.com/u/9435398/Nebula- ... Format.wav

mix 96 khz in CD-Format:
http://dl.dropbox.com/u/9435398/Nebula- ... Format.wav

download with "rightclick, save target as...", and open in your daw. the player in your browser could degrade the sound.

in this case there are no long chaines involved, which will certainly bring other results.
maybe i extend my test, when i have time.

but it is still interesting enough.
as already stated here by others, i can not really find the 96 khz version better sounding, when brought down to 16 bit 44,1khz. there is another sound in the top end of the spectrum, that is for sure.
if i could stay with 96 khz, i would prefer it
over 48 khz.

have fun, (i hope dropbox will not stop too fast, when a lot of people download :-))

mathias
I just listened to the examples in CD-quality and I must admit that I don't really hear a difference.

zaminx
User Level XI
User Level XI
Posts: 184
Joined: Sun Mar 28, 2010 6:57 pm

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by zaminx » Sun Feb 20, 2011 3:05 pm

System Drive/ Libary/Audio/Presets/AcusticaAudio

in there you will find it

futur2
Member
Member
Posts: 303
Joined: Sun Mar 28, 2010 8:16 pm

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by futur2 » Sun Feb 20, 2011 5:12 pm

Stolle wrote: I just listened to the examples in CD-quality and I must admit that I don't really hear a difference.
did the same and listened to the CD-quality ones on my AKG K240 because one of my focal twins just broke down :(
it's hard to to discern a difference. if one sounds better on the headphones i tend to say it's the 48khz.

mathias
Expert
Expert
Posts: 2731
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by mathias » Sun Feb 20, 2011 10:40 pm

i extended my test from above:

the same mixes, but now with two additional nebulaprograms on every track and two on the masterbus, to hear if subsequent programs make a hearable difference (it is eq's, tapes, tubestages, consoles).
so now we have all in all three nebulas on tracks downsampled to 48 khz in one mix and in original 96 khz in the other mix. they go into two nebulas on the masterbuss.

here you go:

48 khz 32 bit float mix:
http://dl.dropbox.com/u/9435398/Nebula- ... -chain.wav

48 khz 32 bit float down to CD-Quality:
http://dl.dropbox.com/u/9435398/Nebula- ... uality.wav

96 khz 32 bit float mix:
http://dl.dropbox.com/u/9435398/Nebula- ... -chain.wav

96 khz 32 bit float down to CD-Quality:
http://dl.dropbox.com/u/9435398/Nebula- ... uality.wav


as i said before, if i could stay with 96 khz, i would prefer that one. it has a little more energy and definition in the highfrequency range and i like that. 48 khz sounds a little weak in comparison.
in cd-quality i don't hear much of a difference.

another thing is, i think i would mix differently when i worked entirely in 96 khz, so this is not reflected in this comparison.

have fun hearing,
mathias
system 1: windows 8 32 bit - samplitude pro x3, waveform 9 (tracktion), reaper
system 2: mac osx yosemite - reaper(64bit), waveform 9 (64bit)

both systems on: macbook pro (late 2009), core 2 duo 3,06 ghz, 4 gb ram, graphic: nvidia geforce 9600M GT 512 MB

User avatar
fuseburn
User Level X
User Level X
Posts: 133
Joined: Mon Apr 05, 2010 6:19 pm

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by fuseburn » Mon Feb 21, 2011 10:47 am

mathias, thanks for this test !
I compared both versions (first and additional chains) and I feel confirmed in my findings. 96k is the winner for me. Reverb is glued to the guitars, but still with room to breathe - most depth, if you will. Not much difference between 96k-44.1k, 48k-44.1k and 44.1k versions. If you WANT to hear it, you can find a few remnants of that airy feel in the 96k-44.1k version, but it's basically gone.
I've heard downsampling algorithms which can preserve a significant bit more of that feel @ 44.1k - what did you use for the final downsampling ?

So for me that means, working @96k with a target samplerate of 44.1k only makes sense if you've got the best SR conversion algorithm for downsampling available. If you never cared for downsampling quality - which I could perfectly understand - there's no need for you to work @96k.

iquinn
User Level 0
User Level 0
Posts: 3
Joined: Fri Apr 23, 2010 2:29 pm

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by iquinn » Mon Feb 21, 2011 5:48 pm

Is it possible to re-sample 96kHz libraries down to 44kHz with a better 3rd party SRC and create a new library?

mathias
Expert
Expert
Posts: 2731
Joined: Wed Mar 31, 2010 12:25 am
Location: South-West Germany

Re: Sample Rate Discussion (off of AITB Germanos EQ Thread)

Post by mathias » Mon Feb 21, 2011 7:39 pm

fuseburn wrote:mathias, thanks for this test !
I compared both versions (first and additional chains) and I feel confirmed in my findings. 96k is the winner for me. Reverb is glued to the guitars, but still with room to breathe - most depth, if you will. Not much difference between 96k-44.1k, 48k-44.1k and 44.1k versions. If you WANT to hear it, you can find a few remnants of that airy feel in the 96k-44.1k version, but it's basically gone.
I've heard downsampling algorithms which can preserve a significant bit more of that feel @ 44.1k - what did you use for the final downsampling ?

So for me that means, working @96k with a target samplerate of 44.1k only makes sense if you've got the best SR conversion algorithm for downsampling available. If you never cared for downsampling quality - which I could perfectly understand - there's no need for you to work @96k.
i have used the internal samplerate-converter of samplitude in the highest quality. it is quite good, but there are for sure better ones.
at the moment i will stay at my workflow, we will see what the future brings in storage media, that will be used on a broad basis.
i would only go up at the moment, if the product would be a hires-download for example.
maybe i change my mind, when i have a more potent machine, that can handle 96 khz in realtime.
to be prepared for future releases :)

mathias
system 1: windows 8 32 bit - samplitude pro x3, waveform 9 (tracktion), reaper
system 2: mac osx yosemite - reaper(64bit), waveform 9 (64bit)

both systems on: macbook pro (late 2009), core 2 duo 3,06 ghz, 4 gb ram, graphic: nvidia geforce 9600M GT 512 MB

Post Reply