How did I not see this coming! 96khz sounds great but...

Well I’m 53 so 96khz is truly bllcks for me as I’m lucky to clearly hear above 16KHz :slight_smile:

So … before you potentially waste your money.

Get someone else to properly and scientifically ABX double blind test you and make sure you can pick out 96KHz over 44.1 KHZ 100% of the time otherwise your just guesssing and therefore wasting time money and rescources in general.

Personally I’ve yet to see and be present with ANYONE in a a room who can pick 96KHz out when double blind ABX tested over 24/44.1 so there you go :slight_smile:

Best advice ever and really do it double blind! If you don’t, just stop using 96k, that is an order! :wink:

As already suggested a new audio interface could help (better driver performance / buffer usage).
I highly recommend UAD Apollo interfaces as you gain the additional CPU power for the included plugins (Quad system for 96 kHz at minimum).
BTW this is where the UAD systems still shine (besides the pure quality of the plugin coding itself) → higher sampling rates. There are still home users who deny the UAD systems the right to exist in times of multicore processors. But everyone working with higher sampling rates knows how fast a CPU is eaten up by a bunch of good plugins.
A good multicore system, UAD power and a lot of bouncing usually does the trick.

And don’t listen to these people who clearly do not know what they are talking about. Higher sampling rates always sound better, always(!) as soon as you do the slightest processing! This is the same simple minded bunch of people who are telling you that every DAW sound the same because they null (of course they NEVER EVER null as soon as you only turn one single knob(!)).
Higher sampling rate → higher quality processing → better results (depending on the source material, highly audible even after downsampling to 44 kHz(!)) → but sadly not in Cubase (see here: Poor Sample Rate Conversion (SRC): How do you cope with it? - Cubase - Steinberg Forums)

Lol he asked for discouragment

Once you taste blood, you’re hooked. :smiley:

Foobar2000 has a great ABX plugin. Test your own setup, don’t believe, be sure.

themarqueeyears, the trouble is i started making music when i was 10 on the playstation with music 2000 and have used a massive variety of programs up until now (26yo where i started recording my own 24/96 drum samples… i noticed the difference immediately.
my way of thinking behind this is that you are effectively giving the frequency range more headroom so that harmonics can manifest with more accuracy. (recording in and playback with 96khz)
thing is, even though we cant hear these higher frequencies, they still exist, and still are vibrating our bodies ears and brains so it is actually adding qualities to the information. like oxygen which we cant see, it still enters our body.
but your right i have never blind tested so i am going to do this and i will post my findings!!

tedannemann, i follow you, but i ask, am i too attached to waves plugins? are UAD as good?
as for programs that sound different, they definitely do sound different!! Ableton vs Cubase? Cubase sounds MUCH better.

jazzyp, im interested. nexis, i dont like freezing stuff, i want to manipulate everything all the time! (jeez i dont ask for much huh)

Thanks
T

Please no more of this pseudo science, that is simply not how digital works, 44.1k is as accurate as 96k.
I don’t mind someone saying they find 96k better, there are a number of reasons this can be true.
Some plugins and converters do sound very different, and for that reason alone (that can be easily tested with a Null-test), it can absolutely make sense to record/work at 96k.
Sorry if I’m offending anyone, not my goal here.

Really? So what your saying is, if you are recording something, and have two microphones, one set at 44k and the other at 96k, there is no difference in the audio information???

I think you’re taking about two different things. You eluded to an increase in accuracy, but it’s not true that you gain accuracy by recording at a higher sample rate. First of all the actual sampling is done in the MHz range after which it is filtered and down-sampled to whatever rate you want to save at, and secondly you get no additional resolution in time of any given frequency with increased sample rates; you either acquire it correctly or not at all.

The other thing you guys are talking about, or you at least, is whether or not we can somehow make use of the frequencies we can’t hear. I think it’s an interesting proposition, but it’s one that I as far as I can recall have never seen verified scientifically.

I think it’s generally better to comment by using actual examples of why processing would yield different results, specifically, and how they are significant enough to worry about. The above is mostly unnecessarily hyperbolic and polemic in my opinion, and doesn’t really serve to educate the reader.

I think you’ll find a lot of engineers that prefer UAD plugins. I personally think UAD is pretty far ahead, and that’s despite using Waves plugins for work all the time.

As for programs sounding different, I’ll offer you the following way of thinking about it: Consider the ways a DAW affects the sound, and then compare those ‘processes’ between different DAWs. Summing for example should be about the same, with discrepancies so far down from 0dBFS they’re completely irrelevant (i.e. due to dithering). Same for gain change. Processing however can be different. So EQs can sound different, although a lot of times they probably sound about the same or close enough to not matter.

So, if the difference between for example Ableton’s and Cubase’s is because of the included plugins, then ask yourself if that matters if you end up buying and using a bunch of Waves/UAD plugins instead… It won’t matter much then, correct?

This is no pseudo science as the higher resolution nearly always gives you better results. You will be amazed how better time stretching and pitch shifting sounds if the algorithms can do their work in a higher sampling rate. Let alone nearly all virtual instruments and plugins do profit from higher sampling rates. Especially the big digital devil called aliasing can only be beaten with oversampling / higher sample rates. I totally agree that the unprocessed signal can’t be differentiated between 44.1 kHz (actually everything over 17k depending on your age) and 96 kHz+. But as I already said this is the same stupid thinking as with the DAW null tests with unprocessed signals. I never ever heard an unprocessed signal on a professional recording - even classical or jazz live recordings are heavily processed all the time… That’s why higher sampling rates are absolutely needed in audio processing because of the cleaner more transient rich sound with less artifacts.

Please listen to these three audio examples. All made in Cubase to demonstrate how much you gain with higher sampling rates.
To show how much a higher sample rate matter, I even downsampled / converted the result to nearly the lowest MP3 quality (128 kbps):

1 Yeah Science B_____!.mp3

  1. Guitar loop from Cubase’s stock library + AmpSimulator. 192 kHz → 44 kHz → 192 kHz → 44 kHz …
    (Can someone lift the blanket from the guitar amp, please?)

2 Yeah Science B_____! 2.mp3
2. A brand new virtual synth (without oversampling) just playing a few square wave notes. 44 kHz → 192 kHz → 44 kHz → 192 kHz …
(Aliasing like hell!)

3 Yeah Science B_____! 3.mp3
3. Sine sweep with Cubase’s Tube Compressor on it. 44 kHz → 192 kHz → 44 kHz → 192 kHz …
(Radio tuning effect NOT intended! :wink:

Be aware that the preview of zippyhsare is downgrading the signal once again. To hear the real 128kbps MP3 you have to download the file (especially with example three).

If you don’t hear any differences, I highly recommend looking for another profession :wink:.

What you hear in all these examples is mainly aliasing in an exaggerated demonstration. However the concept of a more cleaner / transient richer / more accurate processed / (some even say more analog) sound also takes place with 96 kHz vs 44 kHz while only doing slight processing. It’s not hard to imagine how different a recording sounds if we are speaking 48+ channels all burying a bunch of artifacts in their signals vs no / less artifacts.

Higher sampling rates definitely sound better because of less artifacts / more precise signals while processing (meaning, fx, pitching, shifting, saturation, compressing and so on).

Tonight is poker-night, so I’ll listen tomorrow, not that I doubt there’s a difference.

But you get at least an A- for the time being for the snark above. :confused:

MattiasNYC

Thing is i am not trying to say that i am scientifically correct here. to be quite honest i dont have a clue about the science behind audio.

All i know is what i have experienced. that is that after years of using drum samples at 44k, switching to 96k made a MASSIVE difference in fidelity. as far as ableton and summing and nulls and all this rubbish, it doesnt matter to me.

what matters is that when i am using the program i want the audio to be pristine, and ableton seems to sacrifice realtime audio fidelity for whatever reasons (before using plugins). rendering the audio? it will probably sound indistinguishable from any other daw, i dont know, but i need the audio to be clean when i am working with it, and ableton does not offer this.

thanks to tedannemann for showing exactly what im on about with the audio samples. and yes i totally feel that 96k gives a more analogue feel, it might be phsycoacoustic? i dont know it just sounds better.

I think of it the same way i think of photography. You shoot in RAW format so that when it comes to manipulating the image there is more information in the file to work with, giving way for more manipulation without too much damage.
Thing is shooting a RAW picture, it looks better than shooting a jpeg anyway before you manipulate it, and i feel the same about higher res audio.

another way i look at it, although our ears are limited to 20k, things which create sound in reality are not limited to any frequency range. therefore, surely one would want to capture sound from reality with the least amount of limitations, and so if it is feasble to record with the highest resolution, why not? you are capturing a closer representation of reality?

tedannemann:
Thanks for posting those examples. I am however very curious how you achieved these, because that level of aliasing is absurd…
Aliasing can occur in 2 different ways I think.

  1. When going analog to digital, and the analog signal isn’t properly low-pass filtered. → this is not what you’re doing because you’re using signals that are digital to begin with. Besides, any converter that was made in the last 30 years (random number) should have a filter to prevent this level of aliasing.
  2. When downsampling high-samplerate material to a lower sampling rate. → I’m guessing this is what you’re doing, but I wonder how exactly because I’d expect a downsampling algorithm to do better than that…

Anyway, from my point of view going higher than 44.1kHz should only ever be considered when everything in your signal chain can handle the higher samplerate and higher frequencies. Including your speaker! If your speaker can’t produce the 40kHz tones they’ll actually cause aliasing too (It’s not true aliasing as it’s a mechanical thing, but I don’t know the right term for it and the effect is similar to aliasing :wink:).

I hope you had a good hand. However, nobody should be offended here because the artifacts are that audible in the lower spectrum that even my grandma can here the aliasing. :slight_smile:

Without commenting on whether 96K is an improvement over 44K, I would like to point out that your CPU is actually quite powerful, and may not be the source of your problem at all. Have you done diagnostics on the system?

Just to clarify this again. In a blind test nobody hears a difference between a 44 kHz and a 96 kHz recording done with the same quality converter. And this is what people usually blare about in forums. No difference. Which is true for the pure same quality a/b comparison.

BUT the difference is quickly audible as soon as DSP (digital signal processing) is taking place → and this is what I am talking about. DSP is usually ALWAYS taking place as soon as you start mixing with FX! So that in the end there is quickly an audible difference between the 44 kHz and 96 kHz processed recordings.

Besides DSP, if you are insist on hearing an improvement when recording on a higher sampling rate two things can be true:
a) you really have bat ears
b) your converter sounds better at a higher sampling rate which often can be the case with poorly designed convertes

Listening to 96 kHz makes no sense but working / processing in 96 Khz makes absolutely sense and that’s what we guys are doing, right?

Be careful with the Ableton Live bashing because Live already offers double precision mixing since version 7 (namely 64-bit summing engine). Cubase still doesn’t offer this high resolution in terms of summing points. Since version 9 they also updated their sample rate conversion for exporting to a mastering level standard which let Cubase (and Nuendo!) look like an amateur software in this regard (no update for the SRC for over 10 years now).

The sad truth is, if you are record in 96 Khz in Live 9 and do some mixing (without stretching and processing, meaning not plugins involved), then saving in 44 kHz, you not only get other measurements but also highly better audible results than in Cubase! (I am not exaggerating here).

If no sample rate conversion or mixing takes place - the files more or less null (as long as you change Cubase’s pan law to Live’s equivalent).

Because of less artifacts, namely „aliasing“ one can indeed have the impression of a more analogue sound.
By the way, this is one of the reasons the Clavia hardware synthesizers gained so much of an status in terms of analogue sound (from a VA). They already used 96 kHz for internally processing in times where others where debating the signification of 24-Bit vs 16-Bit on internet forums!

Technically speaking this isn’t a good comparison. However, in the sense of manipulating in a higher quality and getting better results at the output stage - this is somehow true for higher sampling rates.

As I already said for pure listening pleasure no 96 kHz is needed. But as soon as DSP is taking place I highly recommend even working at 192 kHz (wich I believe is going to be the new standard sometime in the future when the CPUs are easily capable of working in such rates).

Ha, ha. See :mrgreen: ? This is what happens all the time „below“ your music if you are using up- or downsampling or any kind of analog modelled filter or fx without oversampling (practically all the time if your are using Cubase’s (analogue-ish) stock plugins).
That’s why 96 kHz processed(!) recordings clearly can bee differentiated from a lower processed recording - even when artifacts are not that obvious like in these examples of course.

True, but also when up-sampling(!) (anybody who’s using 44 kHz library’s for film stuff in 48 kHz for example) and most important if you are using any sort of overtone generating / saturating plugins without internal oversampling. This for example could be the Slate Virtual Mix rack (unfortunately no oversampling therefore heavy aliasing in 44 kHz) and a whole bunch of other well known plugin standards of the last decade!

And Cubase’s sampling rate conversion is by far the worst out there (audibly) in a professional DAW. I must admit I am currently on a SRC witch hunt, see here: https://www.steinberg.net/forums/viewtopic.php?f=226&t=101447

Below you find a simple Cubase project with the TestGenerator (sine sweep) + the Tube Compressor. This is a real world scenario - every electronic musician is using noises, whooshes and risers which exactly generate the same unbelievable bad aliasing artifacts (not to speak what happens to high cymbals, crashes and hihats). And the Tube Compressor is in its default state(!) (no drive exaggeration used here)! If you turn the knobs you can generate a hell of a lot more aliasing if you wish!
However just listen what the Tube Compressor is doing to your signal in 44 kHz and then switch to 96 kHz and if your audio interface allows, better to 192 kHz and listen again.
I must admit that this example is somehow extreme because of Cubase’s own TesGenerator already not generates a clean sine wave (and yet again we are sadly in the amateur ball park here again!) which exaggerates the aliasing compared to a „real sine wave“. Anyway, in a real world mix you are using different kind of plugins, processing or even some analog synth recordings which can generate more or less the same amount of aliasing with the simple Tube Comp in Cubase.

Cubase sine sweep project. Please switch sampling rates for your listening pleasure :wink:
http://www16.zippyshare.com/v/l8BDuQJf/file.html