24/96khz

[kmonkey]

Whyterabbit, FWIW, i am mostly referring to the huge "Lets test plugins" thread on Gearslutz, where they tested a lot of plugins and found the aliasing to be far reduced at the higher sampling rates.

That is what I mean by "sounds better"

Basically the character of the tone is the same, but you get less aliasing artifacts and usually a more open sound, that is why I use the words "sounds better"

So in other words instead of "scientists" we actually have "people on a forum", instead of "proved" we have "measured" and instead of "always sounds better" we have "had less aliasing".

ok. that kinda speaks for itself. the less aliasing thing would have been kind of obvious though, Nyquist and all that. proper ABX testing to determine whether the aliasing artefacts presents are perceptable and/or affect a subjective quality assessment of the sound would be the kind of direction you'd actually want to go in to start proving that kind of hypothesis, though. scientific proof has to be kinda rigorous that way...

I agree to everything you said above. People should realize that we live in 2012. Today top leading developers are clever and they are making top notch algorithms (and products) which are sounding superb even at 44khz due internal algortihm optimizations of various kind (oversampling, whatever..several good examples : u*he zebra, cytomic the glue, drop, fxpansion, harmless etc.).

Add to that that even old cursed access virus is working at 44khz yet it was enough for it to be some kind of cult instrument in music history. I never heard anyone complained about 44khz limit.

You guys should really ditch old soft instruments, invest in new and start to make some music (at 44khz)..

[hibidy]

As of today I am now in the land of 96khz. Already everything is sounding more open and airy, crisp analog round fat gooey goodness.

See: Placebo Response.

Lol! You guys are probably stuck at 44.1 and have no idea what you are missing.

Sorry... having skimmed this thread a bit, I decided to start at the beginning, then only got this (that ^^^ up there) far into it, before:

Seriously?!!!



A lot of very good (and accurate) commentary here, that clearly illustrates otherwise.

since it's two different things in that quote, which are you referring to?

[Uncle E]

But 88.2kHz (which is twice of 44.1kHz) is an odd number to 48kHz (twice as much would be 96kHz). So there is still a lot of math involved. Same with going from 48kHz to 44.1kHz. Thankfully there are some really good standalone SRC's that can handle that, but still.

Personally, I just double the target sample rate, whatever that may be. Less math, less to go wrong, and, most importantly, significantly less time sample rate converting.

[Shabdahbriah]

As of today I am now in the land of 96khz. Already everything is sounding more open and airy, crisp analog round fat gooey goodness.

See: Placebo Response.

Lol! >>> You guys are probably stuck at 44.1 and have no idea what you are missing. <<<

Sorry... having skimmed this thread a bit, I decided to start at the beginning, then only got this (that ^^^ up there) far into it, before:

Seriously?!!!



A lot of very good (and accurate) commentary here, that clearly illustrates otherwise.

since it's two different things in that quote, which are you referring to?

The bolded one.

but, point taken

[hibidy]

still not following.

I think I'll just bow out. It's obvious that regardless of facts, or even implied differences, the sonic police don't approve. AND instead of offering proof of how there ISN'T a difference, it just sounds like a pot shot because of some other inadequacy.

To the op, again, enjoy. That's what this should be about anyways.

[@midnight]

Whyterabbit, FWIW, i am mostly referring to the huge "Lets test plugins" thread on Gearslutz, where they tested a lot of plugins and found the aliasing to be far reduced at the higher sampling rates.

That is what I mean by "sounds better"

Basically the character of the tone is the same, but you get less aliasing artifacts and usually a more open sound, that is why I use the words "sounds better"

So in other words instead of "scientists" we actually have "people on a forum", instead of "proved" we have "measured" and instead of "always sounds better" we have "had less aliasing".

ok. that kinda speaks for itself. the less aliasing thing would have been kind of obvious though, Nyquist and all that. proper ABX testing to determine whether the aliasing artefacts presents are perceptable and/or affect a subjective quality assessment of the sound would be the kind of direction you'd actually want to go in to start proving that kind of hypothesis, though. scientific proof has to be kinda rigorous that way...

I agree to everything you said above. People should realize that we live in 2012. Today top leading developers are clever and they are making top notch algorithms (and products) which are sounding superb even at 44khz due internal algortihm optimizations of various kind (oversampling, whatever..several good examples : u*he zebra, cytomic the glue, drop, fxpansion, harmless etc.).

Add to that that even old cursed access virus is working at 44khz yet it was enough for it to be some kind of cult instrument in music history. I never heard anyone complained about 44khz limit.

You guys should really ditch old soft instruments, invest in new and start to make some music (at 44khz)..

well the reason why I got down this path is one of the instruments I own (Fabfilter Twin) aliases quite badly at 44khz, you can hear it quite easily in the upper register when using sawtooth wave in combination with any of the filter mode's other than "clean"

I believe it is due to the filter code and Fabfilter should really take a look at their plugins and add oversampling to the ones that don't have it yet could definitely use it (Twin, One, Volcano, Pro-C, for example)

[Compyfox]

Sheesh, midnight, stop the quote pyramids - PLEASE!

Personally, I just double the target sample rate, whatever that may be. Less math, less to go wrong, and, most importantly, significantly less time sample rate converting.

So let's say you work (constantly or not) in 48kHz, downsampling to either format is a non-issue nowadays. But if yout target product needs to be 48kHz, you essentially record in 96kHz? (downsampling only while mastering of course)

Again, I try to see a reason behind it other than having a more finer resolution while recording (while totally ignoring bitrates, which also count to the equation). Let's not discuss internal clocking of devices or some other crazy stuff here - but plain 96kHz.



96kHz ultimately cuts ADAT slots in half unless you have a native 96kHz device connected through MADI, Firewire (800! and even then, limited channels) or have assentially a card(bus) for that. So this is a fact.

Another fact is the higher CPU usage if you apply plugins, either with OS "active" (or adaptive, which turns off at higher sampling rates) or not.

And... we have the problem with multi-track recordings and data transfer of the SATA pipe. Let's say you are on SATA 300 but have a multi-track recording with like 30 channels and a running time of 3:30miutes. This project should roughly eat up 1,7GB for just this one project (96/24 with 3,5min are like 58MB per channel, at 32bit it's already 77MB for 3,5min and 2,3GB for the whole project). And I'm not counting disk streaming for samples yet, freezes or stems.


Even with modern PC's and especially Mac's (which seem to have halted through the years) I definitely see massive problems coming up if the idea is to constantly go higher and higher in terms of sample rate. Mainly the disk streaming and the CPU usage. Granted, you can always go for a Sandy Bridge and SATA 600 (with drives that have a huge cache), but this is also a financial issue.

Again, if you work in movie postpro or (let's say) STEM mastering (with 4-5 stems maximum, which shouldn't be an issue CPU and HDD load wise), and you have the resources to go run that all in real time - then for all means, do it.

For regular music production, especially those that are pressed to it's limits (which will end eventually as soon as the EBU R-128 system will be ported over for music playback as well), I don't see a reason. And thankfully, here we have the cheating capability to OS.


I still hear so many great productions made in 44kHz (but at least 24bit!) only, maybe even recorded on a 10 year old 20bit digital mixing console (I'm looking at you, ROLAND!) that sound just as good as recordings from mid/end 90ies. If you know your limits however.



So I wouldn't say "96kHz is better at all costs" - it can and will be, if we have the capabilities to actually appreceate it (moderate dynamic range, proper playback devices, affordable authoring). But as long as there is no mass consumer solution, and also not a suitable loudness agreement, I see it as a waste of resources. Especially for a lot of genres in the electronic music realm.

Why don't we focus on actually recording in higher-than 32bit float while printing on HDD? Granted, most ADC/DAC's are still locked to 24bit on the input/output side, but if it's all about resolution while recording and editing, why not simply raise the bitrate instead?

A 48/32 (integer however) recording at 3,5min only eats 39MB per channel - much easier to handle for the HDD on multi-tracks. With 4x and 8x fixed/optional OS we still have a resolution of 192kHz (internally) to work with.

And while we're at it, why not finally change the frequency response form 20-20k to something like 10-22k for ADC/DAC?



Really, I don't get what this fuzz is all about if the definite limit is still the computer hardware. Maybe it's due to that, that certain engineers believe that "analog outboard gear/analog mixing is still superior"? Don't think so in terms of noise and harmonic distortion, but if we talk about "resolution", it's a different thing.


In the end, the music counts. Not at what gigantic sampling rate you recorded it other than "oh wow, now I can offer it on HD audio sites and release it on Blu-Ray Audio".


Personal opinion of course.

[kmonkey]

Sheesh, midnight, stop the quote pyramids - PLEASE!

Personally, I just double the target sample rate, whatever that may be. Less math, less to go wrong, and, most importantly, significantly less time sample rate converting.

So let's say you work (constantly or not) in 48kHz, downsampling to either format is a non-issue nowadays. But if yout target product needs to be 48kHz, you essentially record in 96kHz? (downsampling only while mastering of course)

Again, I try to see a reason behind it other than having a more finer resolution while recording (while totally ignoring bitrates, which also count to the equation). Let's not discuss internal clocking of devices or some other crazy stuff here - but plain 96kHz.



96kHz ultimately cuts ADAT slots in half unless you have a native 96kHz device connected through MADI, Firewire (800! and even then, limited channels) or have assentially a card(bus) for that. So this is a fact.

Another fact is the higher CPU usage if you apply plugins, either with OS "active" (or adaptive, which turns off at higher sampling rates) or not.

And... we have the problem with multi-track recordings and data transfer of the SATA pipe. Let's say you are on SATA 300 but have a multi-track recording with like 30 channels and a running time of 3:30miutes. This project should roughly eat up 1,7GB for just this one project (96/24 with 3,5min are like 58MB per channel, at 32bit it's already 77MB for 3,5min and 2,3GB for the whole project). And I'm not counting disk streaming for samples yet, freezes or stems.


Even with modern PC's and especially Mac's (which seem to have halted through the years) I definitely see massive problems coming up if the idea is to constantly go higher and higher in terms of sample rate. Mainly the disk streaming and the CPU usage. Granted, you can always go for a Sandy Bridge and SATA 600 (with drives that have a huge cache), but this is also a financial issue.

Again, if you work in movie postpro or (let's say) STEM mastering (with 4-5 stems maximum, which shouldn't be an issue CPU and HDD load wise), and you have the resources to go run that all in real time - then for all means, do it.

For regular music production, especially those that are pressed to it's limits (which will end eventually as soon as the EBU R-128 system will be ported over for music playback as well), I don't see a reason. And thankfully, here we have the cheating capability to OS.


I still hear so many great productions made in 44kHz (but at least 24bit!) only, maybe even recorded on a 10 year old 20bit digital mixing console (I'm looking at you, ROLAND!) that sound just as good as recordings from mid/end 90ies. If you know your limits however.



So I wouldn't say "96kHz is better at all costs" - it can and will be, if we have the capabilities to actually appreceate it (moderate dynamic range, proper playback devices, affordable authoring). But as long as there is no mass consumer solution, and also not a suitable loudness agreement, I see it as a waste of resources. Especially for a lot of genres in the electronic music realm.

Why don't we focus on actually recording in higher-than 32bit float while printing on HDD? Granted, most ADC/DAC's are still locked to 24bit on the input/output side, but if it's all about resolution while recording and editing, why not simply raise the bitrate instead?

A 48/32 (integer however) recording at 3,5min only eats 39MB per channel - much easier to handle for the HDD on multi-tracks. With 4x and 8x fixed/optional OS we still have a resolution of 192kHz (internally) to work with.

And while we're at it, why not finally change the frequency response form 20-20k to something like 10-22k for ADC/DAC?



Really, I don't get what this fuzz is all about if the definite limit is still the computer hardware. Maybe it's due to that, that certain engineers believe that "analog outboard gear/analog mixing is still superior"? Don't think so in terms of noise and harmonic distortion, but if we talk about "resolution", it's a different thing.


In the end, the music counts. Not at what gigantic sampling rate you recorded it other than "oh wow, now I can offer it on HD audio sites and release it on Blu-Ray Audio".


Personal opinion of course.

Dude...sometimes i really like your passion...seriously i am not kidding..

[lfm]

Jitter doesn't cause pops and crackles. Whatever gave you that idea ? At most it will give the sound a kind of "haze" but it is inaudible in all modern soundcards. Jitter is a nonissue.

If you call it "haze" and I call it pops or crackles - big deal - call in the language police.
- Well, here you are, thank you.

My goodness you really are trying hard to misinterpret what I write.

In short it creates artifacts - and that was my intended point - and should be taken into account.

Well great but jitter doesn't cause pops. At least if you have a soundcard that is less than a 100 years old.

See above.

Well the thing that you obviously missed is that 48 and 96 runs from the same clock. A crystal that is running at like 128x48 kHz or somewhere around there. It is clocked from the same clock. To the crystal there's only 2 sample rates: multiples of 44.1 and multiples of 48.

It still have to sync to the lower clock of incoming samples. You wouldn't feed the DAC fifo 128 values of each incoming sample(if 128x48).

But since it's a serial interface in many situations(ADAT/SPDIF/USB) it would be bits to be identified and interpreted to bytes or words at a higher rate obviously than sample rate.

The idea of a PLL is to lock to the rate of incoming samples(whatever format a clock is imbedded kind of).

In the analog world this varies a frequency of local clock until the phase coincides with incoming rate - lock and and decide when it's safe to identify a full sample. In digital world the technique would be a bit different - not being free floating analog clock. So you combine incoming clock with local in a way that creates the edges when to identify sample and put it in the fifo to eventually be handled by the DAC.

Coming from another system this clock will most probably differ a little bit from whatever local clock is on D/A converter.

The PLL circuit is to handle this - identifying what is a sample in time.

My remark - short version - is that this is likely to fail more often at double or quadruple frequency compared to more ordinary 44.1/48k. Especially on cheaper interfaces - selecting cheapest circuits available.

The DAC's I've used since beginning 90's have all used Crystal chips(now bought by Cirrus Logic), both for PLL and DAC itself. I found them very reliable and good sounding chips. They were the best in the 90's, if they still are I don't know. Good enough for me anyway.

Check out what chips are used in the interface you select( for running 96k or whatever) but more importantly when running 96k. You are running more close to what the interface can handle - and that usually means more failure.

This experience is what I wanted to share - sometimes in too many words.

If using wordclock this clock (in DACs and A/Ds)would though be the same on sender and receiver and are less likely to produce any misinterpretations on any sample.

So check if vendor has support for wordclock - that will tell how serious they are about their craft.

If having a studio with all equipment up to be satisfactory it would be a shame to be let down by cheapo interface.

When I have my little studio with all $2000 mikes and preamps I will be sure to check out 96k as well for that final extra finish of it all.

So not to get "haze" in that final touch I will be thorough what I choose for vendor interface.

[whyterabbyt]

well the reason why I got down this path is one of the instruments I own (Fabfilter Twin) aliases quite badly at 44khz, you can hear it quite easily in the upper register when using sawtooth wave in combination with any of the filter mode's other than "clean"

I believe it is due to the filter code and Fabfilter should really take a look at their plugins and add oversampling to the ones that don't have it yet could definitely use it (Twin, One, Volcano, Pro-C, for example)

I suspect that you'll eventually get what you're after. With more horsepower available, and the rising prevalence of zero-delay filters et.c., I suspect many of the developers who do have the intent of keeping their existing products going will fold oversampling etc into them in their next major versions.

In the meantime, there's always ArkeCode's Oversampler for folk with similar concerns over individual elements of their existing kit. Might be worth poking Arke with a stick to see if its worth updating. (eg the x64 support looks to be slightly non-x64)

[hakey]

stop the quote pyramids - PLEASE!

+1 (and not just this thread)

[lfm]

Jitter may cause a sample to be misinterpreted and cause pops and crackle.

I'm not giving in on that.

Jitter is minute variations (at the nanosecond or picosecond scale) of the audio card clock rate, which means samples are not played back or recorded perfectly on time and the lowest the jitter the better the clarity/definition of the sound. It doesn't generate pop or crackles. Pop/crackles are due to buffer underruns when, at low latency settings, the cpu gets monopolized by an hardware interrupt and the sound card buffers are emptied before the cpu is freed and can fill them up again.

I used the wording pops and crackles to let everybody know what kind of problem we are talking about.

Why does jitter matter?
You get artifacts in audio.

Call it whatever is suitable.

How much is created in the audible range is hard to say.

And about specs I learned when manufacturing memory cartridges(like for drummachines and synthesizers) - that different manufacturers create chips with quite different specs - for what is to be the same chip.

In this case I äm thinking of how much current it's using from backup battery(CMOS memory). The first years I got one brand that was fine - they have lasted 15 years without change of battery. But then suddenly I got these crappy chips that were 100 times higher on standby current.

So reading specs is good - but also checking manufacturers quality and appliance to specs is better.

So what is picoseconds in specs might in the practical case be microseconds at some manufacturers.

But most probably - choosing an interface with wordclock support - and you are fine. Then vendor are really serious about what they do. Thinking RME and Lynx among others. There is a reason they are more expensive.

[SWAN808]

this forum is crazy

someone made quite a good test a few pages back - which most seemed to ignore - but instead continue to argue from a cerebral point of view - in stead of real world. It seems like people are more interested in being right, or proving someone wrong which is a trend I notice a lot around here-endless polemics that end up being locked.

The best way to decide this issue is to listen to some well made tests, try it on your own system and use an ABX program if you feel the need to and its not very clear. All this 'scientific test' stuff is absurd. If you need scientific proof to decide whether something sounds good it must be a painful process to make music.

[whyterabbyt]

It's obvious that regardless of facts, or even implied differences, the sonic police don't approve. AND instead of offering proof of how there ISN'T a difference, it just sounds like a pot shot because of some other inadequacy.

in other words, exeunt astride straw man...

[whyterabbyt]

It seems like people are more interested in being right, or proving someone wrong which is a trend I notice a lot around here-endless polemics that end up being locked.

What about merely wanting people who insist they're right to prove it, so that we all can be?

If you need scientific proof to decide whether something sounds good it must be a painful process to make music.

Probably. Not that anyone in the thread actually did say they need that, of course. Despite the self-aggrandising muttered asides of the self-proclaimed elite, most of the saner folk round here acknowledge that actual music isnt actually magically correlated with the gear, be it expensive branded cables, fasters DACs, OTB summing boxes, or the 'right' vintage synth doodah. Its all well and good for an individual to find their own 'answer', (as per the OP, really) but when it becomes dogma, as it most certainly does, then I think there's merit in seeing common sense prevail.
After all, in real life, mileages vary, and there's always More Than One Way To Do It.