Looking for the most accurate Paula audio emulation

[Bren McGuire]

mate what? do you realize the amiga does not output even at 44.1Khz?

Originally Posted by paul1981

My WinUAE isn't the latest version I admit, but it appears to only permit audio mixing at 48KHz maximum. If this was increased to 96KHz or even 192KHz, then the sound emulation would be much more accurate.

[no9]

To be precise, the Amiga does produce a signal far beyond 44.1 kHz
[ Show youtube player ]
but it is not significant by any means and is not desirable from a sound quality perspective.

[Daedalus]

Also, if the Amiga is running in a double-frequency display mode, e.g. Productivity or DblPAL, Paula can also output at double frequencies, so can easily play 44.1kHz samples under those conditions. So a truly accurate emulation would need to take that into account.

Of course, playing a mod that was only intended for the ~28kHz limit of standard screenmodes won't need that sort of output, but when emulating Paula there will always be an error because Paula's output is at the frequency of whatever sample is being played and isn't fixed. Whether this error is significant enough to worry about depends on the sample and the end user, but increasing the mixing frequency will reduce it.

[paul1981]

Originally Posted by Bren McGuire

mate what? do you realize the amiga does not output even at 44.1Khz?

Daedalus has explained it for me, thank you! In the meantime, we can all enjoy better Paula emulation by increasing the rate to 48KHz in WinUAE, or beyond if the option later becomes available. I'd like to see the next version of WinUAE support 96KHz.

[Bren McGuire]

that is not how audio works increasing your modern computer's output rate does not improve in any way the amiga audio!!!

Originally Posted by paul1981

Daedalus has explained it for me, thank you! In the meantime, we can all enjoy better Paula emulation by increasing the rate to 48KHz in WinUAE, or beyond if the option later becomes available. I'd like to see the next version of WinUAE support 96KHz.

[no9]

Originally Posted by paul1981

Daedalus has explained it for me, thank you!

Fantastic! But it doesn't work like that. Having the usual 44.1/48 kHz output set in an application tells you nothing about the internal audio processing that is taking place. This limit extends beyond what the human ear can perceive, so the accuracy of emulation is determined solely by how it is handled internally in the application. It may be the case that the output at higher sample will render the sound more faithful but not granted.

When discussing the accuracy of Paula emulation, it is important to consider not only the possible sample rates it can play, but also the distortion introduced by Paula, such as aliasing or quantization noise from 8-bit samples. And it is NOT limited to 28 kHz (sample rate, so effectively 14 kHz of what we should hear in ideal conditions). So you can't say that "MODs don't need 48 kHz sample rate" because they do if they are going to sound faithfully like on Paula. Look at the picture below. There are the tones starting from 50 Hz up to 320 Hz replayed on Protracker (Windows Clone one, but pretty faithful in this case). But there are also a thin parallel lines above that start from 4000 Hz and they are the result of aliasing distortion which you need to emulate if you asking for faithful Paula sound. They go beyond your hearing range. In this case it makes no difference if you use your Amiga in PAL or double PAL modes, because always you need full range audio spectrum to carry all that garbage that comes out from Paula. Btw. tracker musicians take this into account while picking right sounds and utilize that harshness of low frequency sounds which comes with Paula for aesthetic effect.



Btw. while your ear are perfectly covered with standard 44 or 48 kHz resolutions, your system or sound card might handle one of them better than the other. This happen to be a factor why they sound different and have nothing to do with the application where the sound comes from.

[Daedalus]

Originally Posted by Bren McGuire

that is not how audio works increasing your modern computer's output rate does not improve in any way the amiga audio!!!

You see, that's not what I said. You seem to be claiming that all you need is any sample rate higher than what the Amiga outputs, but that's not true, and is what I was trying to explain (and no9 explains better). If you're playing a sample at, say, 10kHz and outputting it at 11050Hz, you're going to get noticeable distortion that simply isn't there on the real hardware. It's impossible to match the audio output sample rate to that of an emulated Paula, because Paula can and does change its output sample rate all the time, and can play 4 different sample rates simultaneously. WinUAE no doubt does some great work to best approximate the output for our perception, but it's just that: an approximation, and one that must take into account the output sample rate.

Originally Posted by no9

So you can't say that "MODs don't need 48 kHz sample rate" because they do if they are going to sound faithfully like on Paula.

I'm not sure if that was for my benefit, but apologies if it was. I can see how my post was misunderstood. When I said playing a mod intended for 28kHz modes wouldn't need that output, I meant the 56kHz Paula modes I had mentioned just before that and nothing to do with a host machine's audio output.

[paul1981]

Originally Posted by Bren McGuire

that is not how audio works increasing your modern computer's output rate does not improve in any way the amiga audio!!!

I'm not quite sure what you mean by that. I advised to set WinUAE audio output to 48KHz which is a selectable option within WinUAE, I'm not talking about altering any other settings.

Tony himself can perhaps let us know if audio emulation is performed at those differing rates depending on which rate is actually selected. I already know the answer though because I used my ears and did the test first before posting on here...48KHz sounds better than 44.1KHz. Now, either that is the case OR, all the emulation is performed at 48KHz and down-sampled to whatever you select. Either way, the 48KHz setting is superior sounding in WinUAE. Play some decent mods and have a listen for yourself.

[paul1981]

Originally Posted by Bren McGuire

that is not how audio works increasing your modern computer's output rate does not improve in any way the amiga audio!!!

Well actually it would do due to oversampling, but that's by the by. When I did the test between 44.1KHz and 48KHz I had my audio card locked to 96KHz anyway in order to rule oversampling out.

@no9
I'm not even considering Paula distortion (which is a very good point which makes true emulation harder), I'm just thinking of it from a purely sampling point point of view. You have 4 channels, each with differing periods, potentially tiny phase differences of heaven knows what (very small portions of time, particularly important with sounds utilising higher Paula sampling rates).

[no9]

Originally Posted by Daedalus

If you're playing a sample at, say, 10kHz and outputting it at 11050Hz, you're going to get noticeable distortion that simply isn't there on the real hardware.

Please elaborate what distortion do you mean. Do you take that into account? https://en.wikipedia.org/wiki/Nyquis...mpling_theorem

Originally Posted by paul1981

@no9
I'm not even considering Paula distortion (which is a very good point which makes true emulation harder),

Which is already done for a long time. Emulators and other software are very good in mimicking Paula for a long time. There may be a matter of some minor details tuning like Paula's frequency response or... I don't know... if you need really hard^hard hardware experience then maybe there is need to add some interference in the audio signal coming from power lines, ground loops, broken RCA cables etc. to bring you the true hardware experience. But probably that's not what you are talking about and for me this is the part of hardware I really prefer to skip over in software.

I'm just thinking of it from a purely sampling point point of view. You have 4 channels, each with differing periods, potentially tiny phase differences of heaven knows what (very small portions of time, particularly important with sounds utilising higher Paula sampling rates).

I have shown in the picture above that Paula's low sample rates require the same output range as the high ones. From that point of view there is absolutely no difference between 44.1kHz, 48kHz and even 192 kHz. As a rule, because without a knowledge of how the signal is processed inside WinUAE I can't be sure if this influences an emulated sound or not.

[paul1981]

@no9
I know very well about the Nyquist rate, but if this theorem was truly true, then why does 96KHz audio sound better than 44.1KHz audio and 192KHz audio sound better than 96KHz audio? Why do these higher sample rates exist? Are the manufacturers who make this equipment ill-informed, or are they just stealing our money? That's a rhetorical question by the way.
Like I already said, 48KHz WinUAE emulated Paula does indeed sound better than the 44.1KHz setting.

[Bren McGuire]

mate you got the wrong person at all i was responding to the other guy who just said "output the emulator at 48Khz to improve Amiga audio quality" i am in complete agreement of what you and no9 said

Originally Posted by Daedalus

You see, that's not what I said. You seem to be claiming that all you need is any sample rate higher than what the Amiga outputs, but that's not true, and is what I was trying to explain

I meant this guy who doesn't seem to understand how audio works or has drank too much audiophile snake oil

Originally Posted by paul1981

we can all enjoy better Paula emulation by increasing the rate to 48KHz in WinUAE

[paul1981]

I'll take the audiophile snake oil every time my friend, as it means I get better sounding audio.
Much love,
the other guy
xxx

[demether]

Originally Posted by Bren McGuire

prove it or it's just audiophile snake oil

I don't need to prove that emulation is not the real thing. It's just the definition of the word : emulation.

[no9]

Originally Posted by paul1981

@no9
I know very well about the Nyquist rate, but if this theorem was truly true, then why does 96KHz audio sound better than 44.1KHz audio and 192KHz audio sound better than 96KHz audio? Why do these higher sample rates exist? Are the manufacturers who make this equipment ill-informed, or are they just stealing our money? That's a rhetorical question by the way.

Yeah, and it is very easy to Google the answers so I don't feel like it needs to be debated here, on Amiga forum. Enough to say: there are number of reasons.

Like I already said, 48KHz WinUAE emulated Paula does indeed sound better than the 44.1KHz setting.

I can believe that. But for me the concept of "betterness" while talking about emulation accuracy is dubious. The sound can feel like better while being far from the goal. Are you after high quality or high fidelity?

Anyway, I didn't say that WinUAE's output doesn't vary at different samplerates. It does. But is it significant? My point is that there are couple of the other factor that can come into play here: your system, your audio interface and how they treat such resolutions. And they may greatly! In WinUAE there are also interpolation settings which also may impact the sound greatly, but you seem to stick only to sample rates. I also don't claim that resolutions over 48 kHz won't improve the sound that comes from WinUAE. I'm just saying it is not sure, since we don't know internal sound processing in this application. And the whole hardware/software system in which it operates.

[paul1981]

Originally Posted by no9

Yeah, and it is very easy to Google the answers so I don't feel like it needs to be debated here, on Amiga forum.

I think whether you personally believe that higher sampling rates sound better is a good starting point if we're to be talking about higher sampling rates regarding Paula emulation. I say this because of the 4 channel variable sample rate of Paula audio, which takes it closer to analogue sound than traditional digital sound.

[no9]

Originally Posted by paul1981

I think whether you personally believe that higher sampling rates sound better is a good starting point if we're to be talking about higher sampling rates regarding Paula emulation.

There is no room for personal beliefs here. I already said that the crucial is the internal processing of audio in an application and anything above 44.1 kHz is secondary.

I say this because of the 4 channel variable sample rate of Paula audio, which takes it closer to analogue sound than traditional digital sound.

Let's entertain it a little bit. I'm all ears . How does it work, you think? Should anyone be excited because of this 'analogue' keyword? Does it make sound better or worse?

[paul1981]

Originally Posted by no9

There is no room for personal beliefs here. I already said that the crucial is the internal processing of audio in an application and anything above 44.1 kHz is secondary.

Are you avoiding the question? It was a simple one - do you or do you not believe that higher sample rates lead to better sounding audio? And let's not pretend either that no one knows what the word 'better' means. To clarify, I am not specifically talking about Paula here either, but real sound: that stuff that's not broken up at all into time slices at the Nyquist rate.

Originally Posted by no9

Let's entertain it a little bit. I'm all ears . How does it work, you think? Should anyone be excited because of this 'analogue' keyword? Does it make sound better or worse?

It was explained to you yesterday and today - time periods of Paula that fall in between the emulated sampling rate, the error increasing with higher Paula sampling rates, and what makes it worse again is that there are four channels, each capable of a different sampling rate. As for the excitement of analogue audio...yes, I get very excited everyday that sound can be picked up in a microphone and converted into an electrical signal. This enables me to hear recorded sound through a pair of speakers after amplification. Why wouldn't one get excited about it? Don't you listen to music?

[no9]

Originally Posted by paul1981

Are you avoiding the question?

Obviously! Since it was already answered ignoring that repeated question is the only viable strategy for this discussion.

It was explained to you yesterday and today - time periods of Paula that fall in between the emulated sampling rate, the error increasing with higher Paula sampling rates, and what makes it worse again is that there are four channels, each capable of a different sampling rate.

None of that make it anything close to analog. Fortunately.

Don't you listen to music?

Nope. Only sliced digital reproductions that can hardly be called music.

I rest my case. There is nothing to debate here. I wish you eternal fun at high sample rates!

[paul1981]

@no9
Let's face it, you just won't answer a simple question. I can guess why too - because you haven't tested this yourself. Just admit it, it's nothing to feel embarrassed about.

Also, I see that you are trying to poke fun at analogue audio. If you haven't got the right equipment to make judgements for yourself, then it doesn't do anyone any favours belittling it. This kind of behaviour by you and Bren rather, is the real snake oil here.