Soooo, how's 10.2 treating everyone?

[Greg Malcangi]

Originally Posted by gives

For years we all knew that RTAS did not sound as good as TDM and now that AAX is out, Avid as now said, since the AAX announcement, that RTAS is not and TDM was better.. Of course AAX is better than both and that is what they are saying now at Avid.

I think you are making assumptions based on a misunderstanding. TDM, RTAS and AAX are just plugin formats. What makes one plugin sound subjectively better than another is not the format but the programming of the plugin. I know RTAS versions of a plugin which sound better than the TDM version and vice versa, it's the programming not the format. It's like saying ProTools runs better on OSX than on Windows and therefore OSX is better than Windows. There are applications which run better on Windows than on OSX though, because it's more to do with how the application is programmed for the OS rather than the OS itself.

As time goes on and better algorithms are developed/improved then it's likely that AAX plugins will eventually sound better than RTAS and TDM plugins as updating the algorithms of RTAS/TDM plugins will decrease as they become more obsolete. But AAX is not inherently going to improve the sound quality of plugins, only better algorithms can do that.

G

[bacchus40]

i gotta agree with you Greg,.. i never bought into the whole TDM is better than
RTAS generalization.. and thats all its ever been..

what i know for sure is that TDM plug ins have always been temperamental, in some instances completely unusable in line with RTAS... therefore ... in real world practice.. at least in my own experience and within those studios that i had the fortune to relate with.. most mixers usually ended up running RTAS accross the board... and only cherry picking those few TDM plug ins which they new for a fact {subjectively really} sounded better than RTAS....

really.. it was more of a pain in the arse, and not a necessity to run TDM... at least with those i've had a chance to talk to, regarding this topic

the puddin' is in the algorithms.!... and WAVES RTAS always done well by me.!! so much so... i tend to tread lightly going forward.. i cannot afford a 2 year lag on AAX support...

[TimothyJohn]

Originally Posted by Pete Gates

Hi Tim,
Don't get your point there - the first link backs up what I said! The second one has nothing to do with it; it's a comparison of relative quality of different sample rate conversion software not a proof that any particular ratio of sample rate conversion is better or easier than any other.

Pete

Hi Pete!

And your point? That the computer is not solving 48 / 44.1? My point is that yes it is. But yes your observation is correct--it's not done like a calculator:

"Technical explanation: (48 to 44.1)
Insert 159 zeros between every input sample. This raises the data rate to 7.056 MHz, the least common multiple of 44.1 and 48 kHz. Since this operation is equivalent to reconstructing with Dirac delta functions, it also creates images of frequency f at 44.1−f, 44.1+f, 88.2−f, 88.2+f, ...
Remove the images with a digital filter, leaving a signal containing only 0–20 kHz information, but still sampled at a rate of 7.056 MHz.
Discard 146 of every 147 output samples. It does not hurt to do so since the signal now has no significant content above 24 kHz.
(In practice, of course, there is no reason to compute the values of the samples that will be discarded, and for the samples you still need to compute, you can take advantage of the fact that most of the inputs are 0. This is called polyphase decomposition[1], and drastically reduces the computation effort, without affecting the conversion quality.

This process requires a digital filter (almost always an FIR filter since these can be designed to have no phase distortion) that is flat to 20 kHz, and down at least x dB at 24 kHz. How big does x need to be? A first impression might be about 100 dB, since the maximum signal size is roughly ±32767, and the input quantization ±1/2, so the input had a signal to broadband noise ratio of 98 dB at most. However, the noise in the stopband (20 kHz to 3.5 MHz) is all folded into the passband by the decimation in the third step, so another 22 dB (that's a ratio of 160:1 expressed in dB) of stopband rejection is required to account for the noise folding. Thus 120 dB rejection yields a broadband noise roughly equal to the original quantizing noise.

There is no requirement that the resampling in the ratio 160:147 all be done in one step. Using the same example, we could re-sample the original at a ratio of 10:7, then 8:7, then 2:3 (or do these in any order that does not reduce the sample rate below the initial or final rates, or use any other factorization of the ratios). There may be various technical reasons for using a single step or multi-step process — typically the single step process involves less total computation but requires more coefficient storage." Wikipedia

Or perhaps your point is that a newer processor can do this more effectively,"without affecting the conversion quality".

The second link shows the result of various daws trying to do exactly that (96 to 44.1). So, it's not the processor, but the program (method)? What is interesting to me is how different the results are, depending on the src of the daw. I recall a site that posted audio results, and there were always sonic oddities in the upper harmonics. I think your point is that the processor isn't crunching numbers, it's slicing, dicing, adding 0's, compressing, etc. but in digital land, everything is numbers, 0's and 1's. So, we're talking in circles, I suppose. But, when converting 48 to 44.1, we are still asking programmers to design a program that uses processors to solve a problem involving an equation whose result is an irrational number. We then expect that the audio result be identical to the source.

Yes, you can ask a computer to solve a problem that involves an equation that yields an irrational number (pi?). It doesn't matter how fast or powerful the computer, or how ingenious the method, the processor still works harder. In fact, such equations are used to test the computational power of a processor all the time. (But who cares how hard the processor works!!! We just want it to sound great, and have limitless track count, plugins and VI's, with no latency!) But when the ratio of conversion in this case yields sonic results, there will be sonic oddities that did not exist in the source.

48 / 44.1 = 1.08843537(computed to 8th dec.)

1.08843537 x 44.1 = 47.9999997

On the second link, compare Pro Tools to Goldwave 5.18. Pro Tools' result is much better. If I have the time, I'll try to find the site that posted audio of various daws src's. (using sine waves, which theoretically have no partials).

Converting 88.2 to 44.1, or 192 to 48 requires less processing power.
88.2 / 44.1 =2, 192 / 48 = 4. No statistical problems. The results? The higher the rate, the more upper harmonics that are theoretically out of the range of human hearing. Converting down just eliminates an octave or so of said harmonics.

Tim
Pro Tools 10.2, 44.1k 24 bit.

[Greg Malcangi]

Tim:- Oh dear, you link to a page on wikipedia explaining exactly what Pete said and then quote from the wiki page how the conversion process upsamples to a rate with a common multiple to avoid the possibility of any irrational numbers.

You then go off at a complete tangent talking about how programmers can't deal with irrational numbers!? What's that got to do with sample rate conversion, anything Pete said or anything you quoted in the wiki page? You've already yourself quoted the reasons why the ratio you've arrived at is completely inapplicable.

Pete already explained to you that you can't just divide by two and the wiki page you quoted also explains how it's done, so I'm confused. Are you arguing against yourself or against Pete/Wiki? I can only assume you have not understood either Pete or Wiki or you are just deliberately trolling. Which is it?

[Ghost In The Attic]

Originally Posted by chrisdee

Its the excact same sounds, but to me they sound different at the different sample rates.

Please listen and tell me if you can hear it to.

http://dillner.net/test/

Yes, very clearly different.

[bacchus40]

can you hear the difference??? hugh?

question should be.. do you have a heartbeat!? lol..

yes.. of course... very different indeed!

Originally Posted by Greg Malcangi

When it comes to raw recorded audio files, I agree. However, I find that some VIs and some plugin processors seem to perform much better sonically at 96k than at 44.1 or 48k. This is likely due to differences in the plugins' internal algorithms rather than the assumed improved "resolution" of 96k.

G

fwiw... i agree.. its got very little to do with the actual resolution of 96k.. i believe its more to do with internal algorithms as well... folks I know locally would choose their session sample rates based on what plug-ins they knew they would be needing..
as they were aware of the drastic changes in behavior between different sample rates... hence.. unpredictability factor... depending on the project.. they had an idea of what they could get away with.. specially with virtual instruments..

[TimothyJohn]

Originally Posted by Greg Malcangi

Tim:- Oh dear, you link to a page on wikipedia explaining exactly what Pete said and then quote from the wiki page how the conversion process upsamples to a rate with a common multiple to avoid the possibility of any irrational numbers.

You then go off at a complete tangent talking about how programmers can't deal with irrational numbers!? What's that got to do with sample rate conversion, anything Pete said or anything you quoted in the wiki page? You've already yourself quoted the reasons why the ratio you've arrived at is completely inapplicable.

Pete already explained to you that you can't just divide by two and the wiki page you quoted also explains how it's done, so I'm confused. Are you arguing against yourself or against Pete/Wiki? I can only assume you have not understood either Pete or Wiki or you are just deliberately trolling. Which is it?

Greg,
So how do you upsample to a rate with a common multiple? "Insert 159 zeros between every input sample". (etc.) I never said that programmers can't deal with irrational #'s. I said the task is more complicated, and that the resulting audio will be different than the original (as compared to src's involving ratios that don't result in irrational numbers). My point about people that design src programs is that they have come up with different solutions to the problem. The second link shows graphical representations of numerous results (albeit 96 to 44.1).

The Wiki quote talks about 48 to 44.1, not 88.2 to 44.1, or 96 to 48.

Sorry I haven't found the link with the audio samples--very illuminating.

Never been accused of trolling before --my intent is to be positive and search for clearer understanding (sure wish I could find that link--very odd sounds in the upper harmonics. The purpose of the site was to give audio examples of the very thing the second site illustrates graphically). If you're not recording for video, and are bouncing ultimately to 44.1, why use 48k, 88.2k, 96k, or 192k in the first place? I assume it's for higher fidelity--at least that's what most say, either in the direct audio recording, or in the performance of plugins and VI's.

Maybe I'm missing your point? That you believe the solution presented that upsamples by "insert(ing) 159 zeros between every input sample" etc yields higher fidelity? That it sounds better? Some say yes, some say no, some say no difference. (Maybe that's your point!! that there's no difference!). Or is your point that the process doesn't simply involve dividing the way a calculator does (48 / 44.1, as opposed to 88.2 / 44.1)? If that's your point, then well made!! The original problem stands though--no matter what process (in this case, upsampling to a different ratio) is used for the solution.

I'll continue to work at 44.1 for audio, and use 48 for video (which I rarely do). The solution presented in the Wiki site convinces me not to record at 48 and bounce down to 44.1. Anyway, it seems to me, if you want higher fidelity, go to 88.2 if you're bouncing down to 44.1, and either 96 or 192 if you're bouncing down to 48.

To the OP, sorry to go so far off topic--Again I have had great success with 10.2!!


Tim

[falkenstam]

Well its working really nice now FINALLY! Fig out all by my self...lean ..

[Emcha_audio]

Originally Posted by chrisdee

Sorry for asking in the first place and turning this thread off topic.
But since now half of it is about samplerate i dare post some comparison clips I made from an session of mine.

Its of the same Xpand2 SineBass, Strike Kit and Xpand2 Harp at 44, 96 and 192 khz. Each is bounced to disk from three different session. 44 khz sample is bounced from the 44 session, 96 from 96 session and 192 from 192 session. Its the excact same sounds, but to me they sound different at the different sample rates.

Please listen and tell me if you can hear it to.

http://dillner.net/test/

In your example there do seem to have difference, in the bass clarity and presence between the 44 and 96k, 96k having the bass more define crispy and clear, and also the reverb seems to have more definition the higher up you go. This might be due to the plugin only though. A spectrum analyzer also revealed this to me, as there is definitely more mid and highs coming in the higher res you go.

[chrisdee]

Originally Posted by Greg Malcangi

I don't know what you are trying to demonstrate with these samples.
G

Im only trying to demonstrate that different sessions at different sample rates sound different.
Atleast on my Pro Tools Hd system.

I find it strange and concerning that there is an attack sound on the harp sound
that is only present in the 192 khz sample. Also listen to the crash in the strike sample.

Since this are stock xpand2 and strike patches hopefully it can be replicated on another Pro Tools system if anybody is interested.