Any use of going beyond 44100?

[stefancrs]

So, to sum up - there is not much point in going beyond 44.1 as no one who listens will generally be aware (or care) unless they are listening under geek conditions.

Unless the synths/fx really needs oversampling but does not supply it internally. Then cranking up the "global" sample rate will help. But it will also make every synth/fx that does not need oversampling eat a whole lot more cpu. So that sucks (I myself avoid synths that produces aliasing noise)

[kritikon]

I tried out a few things at 88.2/32 last night. Must say I didn't hear much difference in my most-used synths. One or two improved, but not by much.

But.........holy shit! The CPU useage was spectacular. The first one I tried out was Wusikstation (been meaning to install version 1.1 for a while now and this was a good time). Wusik is pretty good in terms of CPU. I played a few series of 3-note stabs, with a patch that has a reasonable release so it was no doubt getting up to 9 notes often...but taking into account that Wusik is well coded (4 notes for the price of one etc...) my CPU meter was up to 80-90% And that's with absolutely nothing else consuming CPU - no audio channels, no FX, no Eq, no other VSTi loaded. Normally I can have several Vsti loaded and running, and playing as many notes as I want.

Next up Crystal.....yup, you guessed it, a few 3-note chords and the system peaked and crashed.

I run an AMD 2.5GHz with 1G RAM by the way. Not spectacular, but plenty good enough for me at 44.1/16 to do almost anything without having to worry about rendering like I used to.


Quite honestly...who gives a f*** if quality marginally improves on an individual VSTi with higher rates? In the context of a mix it'd be completely missable anyway. At 88.2/32 my system is all but unusable, and it's a pretty reasonable setup.

So I bring my system to its knees so that I can get Vanguard to sound smooth? Aside from the fact that Vanguard only gets dusted off when I want to get gritty dirty sounds.
Until I have an AMD 7.5GHz with 10GB RAM, I'm not even going to consider running at those silly rates.
If the Beatles could make glorious music that stands up today, on a 4-track tape....I can live with digital 44.1/16 thanks. Won't visit this topic or subject again.

[Sascha Franck]

So I bring my system to its knees so that I can get Vanguard to sound smooth? Aside from the fact that Vanguard only gets dusted off when I want to get gritty dirty sounds.
Until I have an AMD 7.5GHz with 10GB RAM, I'm not even going to consider running at those silly rates.

Just seconded.

Well, one could bounce down files at higher quality when being done offline - but then, they wouldn't sound the same as in realtime, so that's pointless to me.

[ericj23]

ps can your 44.1 system get the amplitude of the sine wave at even 10khz correctly - not unless the wave is in phase it can't - lost information

yes it can, no information lost.

Look you here:



The upper graph is the continuous wave from the sequence 0,1,0,-1,etc. The lower graph is the wave from the sequence 0.71, 0.71, -0.71, -0.71, etc, which is the same as above, shifted by 45 degrees. (Where 0.71 is just short for the square root of 1/2, actually).

The point is that DAC conversion will just get everything right, provided is does a good interpolation and filtering. From what I can guess, CoolEdit considers 30 or so neighboring samples to calculate a point on the wave. Your high end DAC should likewise.

alas this example (both of them) are perfectly in phase - one peaks at the sample rate - one peaks at 0.5 times sample rate - try some other numbers and pay very close attention as to how the dacs filters affect the fine detail of the shape of the upper part of the waveform - then make it more difficult and mix in some bass freequencies so that the 15 hz sine does not oscillate about 0 amplitude

[stefancrs]

ps can your 44.1 system get the amplitude of the sine wave at even 10khz correctly - not unless the wave is in phase it can't - lost information

yes it can, no information lost.

Look you here:



The upper graph is the continuous wave from the sequence 0,1,0,-1,etc. The lower graph is the wave from the sequence 0.71, 0.71, -0.71, -0.71, etc, which is the same as above, shifted by 45 degrees. (Where 0.71 is just short for the square root of 1/2, actually).

The point is that DAC conversion will just get everything right, provided is does a good interpolation and filtering. From what I can guess, CoolEdit considers 30 or so neighboring samples to calculate a point on the wave. Your high end DAC should likewise.

alas this example (both of them) are perfectly in phase - one peaks at the sample rate - one peaks at 0.5 times sample rate - try some other numbers and pay very close attention as to how the dacs filters affect the fine detail of the shape of the upper part of the waveform - then make it more difficult and mix in some bass freequencies so that the 15 hz sine does not oscillate about 0 amplitude

Erm, both examples are of a sine with a amplitude of 0.8 (80%), and in both examples the sine peaks at that amplitude. No matter how you would move the phase of that sine it would still peak at 0.8 (the stored _samples_ does not peak there, but the stored signal does).

Any sine below nyquist can be stored with no information loss due to sample rate.

[dr. orange v2.0]

I arrange at 44,1k, then bounce all the MIDI and record in a new project at 96k and mix at 96k. Especially softsynths sound much more defined and better with a higher SR

[Amberience]

I compose and arrange at 44,1k. Its more important to have a higher bit depth if you can afford it, than it is more important to have a higher sampling rate. I always record at 24bit, never 16bit. You get so much more dynamic range and you can get great recordings.

[dr. orange v2.0]

there are many people saying that you don't notice any difference between 44,1k and higher SRs such as 96k, but the difference is obvious! Even if you don't have no trained ears, you hear, or rather feel the difference!

[R.A.W.]

and the most important thing is to resurrect threads that are almost 2 years old..

[stefancrs]

there are many people saying that you don't notice any difference between 44,1k and higher SRs such as 96k, but the difference is obvious! Even if you don't have no trained ears, you hear, or rather feel the difference!

Holy Thread Resurrection Batman!

The difference you hear is most likely due to reduced aliasing, which is the synths / effects fault, not the sample rates, so to speak. So in that scenario, a higher sample rate makes a difference. Otoh, oversampling just the needed parts or using alias free synths / effects would be just as good.

[Toxikator]

As has been said, bit depth is far more important than sample rate, but sample rate matters too.

When mixing down, you won't need to mix outside of 44.1kHZ and 16 bits. However, in the world of DSP, pitch changes (especially with plugins that do it rapidly) ARE relevant, so recording at 32/192, while super-intensive, does theoretically provide you more room to work with your samples.

Consider the attached image:


You can see that the interpolated waveform is identical. However, if you were to double the pitch of each and then half it again, the resolution of the first one would not be as good as the resolution of the second one.

This is irrelevant for straight audio, of course, but when processing samples it's absolutely relevant.