is there any correlation (good or bad) between the sample rate you are working at and the key you are using...?
example:
if you were in the key of G, could there be any noticeable (audible) difference between working at 48 kHz vs. 44.1 kHz?...aside from the "usual" differences when using different sample rates. (say, in the low end for example)
(this is about working with keys/tones related to the sample rate, not about the pros/cons working with the different sample rates in general...)
i have noticed some strange things.
have you?

There shouldn't be. The low pass filter in the convertors should prevent any ring modulation between the clock frequency and the music you're playing.

this is more about recording/working/processing in the DAW before it hits the D/A...
example...a virtual instrument running at 44.1k vs. 48k...
i have experienced that tones that are divisors of the sample rate can create a "problem" in some apps.
so if you have for example a sine bass and a note (or more) is a divisor of the sample rate, could it sound different if you changed the sample rate so it is no longer divisible?

When I saw the title of the thread I laughed out loud...

But then I saw we were talking virtual instruments.

Yes, there is certainly a large body of anecdotal reports of VI's behaving differently at different sample rates.

I'm not sure there's necessarily a relation with key (not, at least that I've seen discussed) but there are definitely reports from many that suggest they like a given VI's operation at a given sample rate. (Preferences being subjective, sometimes one guy will like operation of a given plug at one rate and someone else might like it at a different.)


Now, of course, we also know that for most humans, our expectations, hopes, doubts, and fears (must be election time here in the States) can color our perceptions -- and with something squishy and hard to define (like how a given synth is supposed to sound), that can make pinning down actual differences in operation that much trickier.

This is part of the reason why some people argue 48k might be better. it's just a far more divisible number. regardless of the key of the music.

That and it's comaptibility with video is why I'm tending to try to use it more.

interesting...
heh

well...we could be talking about the app. that hosts the VI as well...

I have a further, extremely subjective insight, which actually slips over the border into insanity-- I sometimes feel like selecting even numbers in the fields where you could select any integer beyond the decimal point, but leaving it at ".00" gets a cleaner result. Find a nice padded cell for me, okay?

what about processes that happen inside of a DAW, like truncation or clipping, or VST/AU effects processing, automation, etc...?
the artifacts are not the same if you have a freq. of the sample rate...

Good point.

The host -- and hosting protocol -- is always part of the equation as well.

You should pardon the expression, equation...

But if you know you're crazy, doesn't that make you smarter than all the people who think they're sane?

heh

well as long as it keeps you listening :D

actually this does have impact in certain things. like Logics time machine. do a time stretch and go 'oh dear!!' ... add a few cents pitch shift + or - as well, and suddenly you hit a divisible number.. and it's ok.

So then, I am not so crazy after all? At least on this one microscopic topic?

Got plenty of others, though...

no not crazy.. if you have a penchant for disastrous Time stretching, .00 should work just peachy.

If a softsynth or plugin implements shortcut/incorrect DSP algorithms, then ya this could potentially produce some weirdness around multiples of the sample frequency. It would probably be easily audible.

For normal recording and simple processes like gain or summing, sample rate ratios would make no difference.

everything makes a difference...

Hey, I make sure all of my tempos are divisible by three; we're all a little crazy.

Oh FFS now i've heard every digital myth that's out there.There simply cannot be anyone left. You do realize a second is a completely arbitrary measure of time? Try and measure samples per PI seconds and see if 48000 is still easier divisible than 44100. It is obviously complete nonsense.

not yet...there's more!


maybe i will try some just intonations and see what those do...

another example: let's compress / limit / or clip one of these tones @ 44.1k...
i can see (many) artifacts very clearly for 880Hz,
but when i look at 882Hz...where did they all go? all i see is harmonics...
switch the sample rate to 48k...and the same thing happens for these two tones?
is a tone related to the sample rate better in this case?
is there something going on with 882Hz @ 44.1k we can't see too well, but it is still there, because we can't just make things disappear...can we?

When you clip a tone, the sampling rate comes into play, as the sample rate influences the overtones generated by the clipped samples (as the "undefined" period of time that the clipped samples represent changes with teh sample rate) ... but that's beyond the VI's influence - you hit a system boundary.
Oversampling can help to move the artifacts out of the audible frequency range or at least this will change the color of the "clipped" sound. But that is not connected to the "sound equations" in the VI - those should always be the same.

What do you mean by "seeing" the artifacts? Are you talking about an analyzer?

Then these "artifacts" may as well be artifacts from the analyzer. If you "hear" differences, your DA-converters come into play, too...

In general there should be no difference in how the VI calculates the sound...the sample rate is just one of hundresds of coefficients in the VI's "sound equation".

what i don't get is why 44.1 kHz and not 44 kHz...?
seems only logical because "concert A" or whatever is 440 Hz and that gazzinta 44 kHz nicely but it's a mess with 44.1 kHz...
can anyone really notice the difference of 50 Hz > 20kHz anyways...?
i thinks this is just troublesome maths to have such a dumb number sample rate...

Because of the video gear used for early processing.
60Hz x 245 field lines (X3 to satisfy Nyquist) = 44100
50Hz x 294 field lines (X3) = 44100

Needed a common for both NTSC and PAL. 44100 was the highest usable for techy reasons. Odd that 48 is the standard for video production.

Another thing that does not compute for me is that video is actually 29.97 fps, not 30. Therefore, should not the actual usable sample rate have been 59.94 x 245 x 3 = 44055.9 (not sure how PAL would fit with that)? I guess we're past all that now, but still using 44.1 as delivery standard.

i always tune my VIs with a peterson autostrobe 490