Best sound quality - run 96khz samplerate, do not oversample.

[tony tony chopper]

the standard rate on blu-ray is now 96k, care to explain that one?

Sure: it's called marketing.
The principle is that if something is commercialized, it's because there's someone to buy it. Of course there's always a guy to tell you "if it was useless, then they wouldn't be selling it".. and then buys it, justifying its existence.

I would too put my money on bigger numbers, if I had to invest. How many "64bit precision" do you see in plugin specs out there? Marketing..

i didn't say it was, i only said he'd be better off since he seems to upset that people aren't agreeing with those opinions.

people can't really disagree with facts

what's amazing is that you're the one insisting with voodoo crap here.

if there is still some misunderstanding - take a pulse waveform, apply a oversample and decimation filter several times in a row and compare the results to the original signal.

you should notice there is severe phase distortion.

newsflash: there are ways to compute an alias-free pulse without oversampling. In which case it's useless to process it at a higher samplerate.

[Fritze]

i didn't say it was, i only said he'd be better off since he seems to upset that people aren't agreeing with those opinions.

people can't really disagree with facts - i mean, apparently they can - but you shouldn't be so emotionally damaged by that.

Yeah, that's life. Fun, isn't it?

Wishlist: I would like to have a working oversampler thingy that would allow me to oversample one given plugin while inserted into my DAW. It should have the ability to switch between settings for online and offline (render) mode. That would solve me some of those problems.

[kmonkey]

i didn't say it was, i only said he'd be better off since he seems to upset that people aren't agreeing with those opinions.

Trying to generalize here? Which people? For the record i agree with his opinion not your. And i actually think that it is hilarious that you as supposedly "man of deep DSP programming knowledge", with several instruments behind don't know that there is a way to compute alias free pulse (which in that case i making oversampling waste of CPU actually). Even i learned that just by being KVR member for several years.

[Fritze]

And i actually think that it is hilarious that you as supposedly "man of deep DSP programming knowledge", with several instruments behind don't know that there is a way to compute alias free pulse (which in that case i making oversampling waste of CPU actually). Even i learned that just by being KVR member for several years.

Hm, I rather agree with tony tony chopper BUT aciddose didn't say that he doesn't know how to program an aliasing free pulse wave. And he didn't say that to get a pulsewave free of aliasing you have to oversample it. Or anything like that. He just gave a certain example, obviously not the best one.
While this is a heated debate, as always with this and such topics, it's cool to stay with the facts and not question peoples sometimes strangely presented DSP knowledge.

[aciddose]

what i said had nothing to do with producing a pulse.

read what i said again, this time actually paying attention.

the effect you should be concerned with is the phase distortion created by application of several filters in series.

actually, the people producing formats like blu-ray actually know what they're doing and have valid reasons for it, even if you don't understand them.

[t3toooo]

blu-ray is video.

[jupiter8]

blu-ray is video.

For silent movies ?

[t3toooo]

blu-ray is video.

For silent movies ?

yes forgot of course,put your 96 khz for amazing sound quality.

[mystran]

Regarding the "plugins should handle aliasing automatically" part:

I've been thinking about this from the other side lately, namely what to do with those people that insist on running their host at higher rates.. because that's bit of a problem for a plugins that do run clean at 44.1kHz (either because they oversample internal anyway, or don't generate any aliasing to begin with); suddenly you're wasting tons of CPU for nothing, just to suffer more issues with numerical accuracy.

In fact I've been thinking that maybe the right thing to do would be for a plugin to simply accept whatever rate, resample it to a "fixed" internal rate (realistically one can probably up/down sample by power of 2 until one ends up in the general range) for processing and then resample back for the host again.

This way you could all run your hosts at 192kHz and we could have our plugins at 48kHz and the waste of CPU would be limited to your hosts having to juggle around 4 times as much sample data.

[aciddose]

we have to of course get as close as possible to our desired rate, since the 2x/halfband will only get us within an octave or so in most cases.

that seems like the most obvious implementation though of course. it's unfortunate that plugins aren't already written that way, although i assure you some are.

another proposition is to interpolate into the destination rate if it's above your internal rate. of course the issue with that is whether anyone might notice and care.

[tony tony chopper]

actually, the people producing formats like blu-ray actually know what they're doing and have valid reasons for it, even if you don't understand them.

I'd like to hear what you believe to be those valid reasons.
Marketing is a valid reason btw (for the side that's selling it).

That said, the future is certainly not a solid storage medium. Well, it's not even the present anymore. The present & future are a myriad of digital formats as floating files and, quoting you, "it's a simple fact of life".

[aciddose]

why should i even reply to you? i would be repeating the reason for the third time.

[Aleksey Vaneev]

When going from 44.1k to 88.2k sample rate you not only get "1 octave of clarity", you get 2 octaves! Because aliased components travel back on the spectrum after reaching 88.2kHz boundary and they have to travel one more octave to reach the audible band. If you work at 192kHz instead of 44.1kHz you get almost 8 additional octaves for harmonics in comparison to 44.1kHz.

As a general rule this is nonsense- what usually matters is how fast the partials fall off as a whole. Most nonlinear processes introduce an unlimited amount of partials, not just a few selected ones. What's relevant here is how much attenuation you get by oversampling. For strong nonlinear processes such as distortion, this can be very little. Also note you do not gain "8 octaves of clarity" by synthesizing a naive sawtooth at 192 kHz. You gain almost nothing because the partials fall off at only 6 dB per octave. This is why synths using naively generated waveforms still sound bad at 192 kHz.

Richard

Well, it's not a nonsense, because I was only talking about an added headroom for harmonics ("octave" was a wrong term I've used in this case).

But generally only hardclipping can be considered as adding an infinite number of harmonics. Most other types of saturation fall-off quickly, and 8 harmonics beyond 20kHz is enough for most "tube-like" saturators and compressors.

When talking about synthesis you should never forget that practical tone range is limited to about 1kHz - higher tones are rarely used for music IMO. Of course, synthesis should be handled differently to saturation - the increased sample rate will only solve the most obvious harshness, but for ultimate clarity the synthesis itself should be designed with antialiasing built-in.

So, instead of looking at the problem technically I suggest you to look at it from 95% "variants of use" perspective.

[Fritze]

Regarding the "plugins should handle aliasing automatically" part:

I've been thinking about this from the other side lately, namely what to do with those people that insist on running their host at higher rates.. because that's bit of a problem for a plugins that do run clean at 44.1kHz (either because they oversample internal anyway, or don't generate any aliasing to begin with); suddenly you're wasting tons of CPU for nothing, just to suffer more issues with numerical accuracy.

In fact I've been thinking that maybe the right thing to do would be for a plugin to simply accept whatever rate, resample it to a "fixed" internal rate (realistically one can probably up/down sample by power of 2 until one ends up in the general range) for processing and then resample back for the host again.

This way you could all run your hosts at 192kHz and we could have our plugins at 48kHz and the waste of CPU would be limited to your hosts having to juggle around 4 times as much sample data.

Very cool idea!

[Aleksey Vaneev]

Regarding the "plugins should handle aliasing automatically" part:

I've been thinking about this from the other side lately, namely what to do with those people that insist on running their host at higher rates.. because that's bit of a problem for a plugins that do run clean at 44.1kHz (either because they oversample internal anyway, or don't generate any aliasing to begin with); suddenly you're wasting tons of CPU for nothing, just to suffer more issues with numerical accuracy.

High-quality resampling requires quite a lot of CPU resources. More so for "fractional factor" sample rate conversion (e.g. going from 96k to 44.1k).