How accurate is WinUAE OCS/ECS audio emulation?

[appiah4]

Has anyone testes WinUAE audio against real hardware? Is it accurate enough to use for judging Amiga game and demo music by? For comparispn I dont consider DosBox OPL3 emulation to be accurate at all..

[Toni Wilen]

Yes. Paula is pure digital (Output is analog but it is output from internal D/A converter) and relatively simple. Emulation is cycle accurate but because there are many bad audio players (CPU delay loop "dma wait"), fast cpu modes don't emulate it fully and play the sound correctly even if real "too fast" CPU would have caused lost sample or repeat of old sample.

[appiah4]

Thanks a lot for the info, much appreciated.

[chip]

I guess WinUAE audio emulation it's nearly perfect

I made this try

I created MP3s from audio recording by WinUAE and compared these recordings to the MP3s of the SOAMC project

This project is a collection of MP3s recorded using the real hardware

So, in other words, i compared the emulated audio against the real audio and, surprise, no difference at all

[AmigaRob]

I have a question about this. I saw an argument on here when I was still only a lurker, about Paula's outputs. I've never been sure if Paula actually has four unique D/As that are mixed in the analogue domain, or if they are mixed digitally. My assumption would have been that each pair of channels was mixed in the analogue domain. With Paula outputting at multiple frequency rates per channel, I would have thought it tricky to exactly emulate the output with any fixed rate PC output as I think you have to interpolate on the fly in the digital down-mix to match the output to the soundcard's rate.

I am not sure if that makes any sense or if I am just talking rubbish, although I can say from a sound engineering perspective, Paula sounds very different going out raw from the Amiga to what I get out of WinUAE. I am not sure most would spot the difference, but as someone who meals occasionally depend on their ears, I can definitely hear a difference, and a difference that might have an impact when using an Amiga in a production chain. WinUAE is an absolute marvel and a masterpiece, but my gut feeling is that it is beyond the limits of the concept of emulation to recreate exactly what the Amiga produces natively.

That being said, for ease of use, I would still generally use WinUAE for rendering out projects where I would use the Amiga as an instrument. What I would most likely do is render it out one audio track at a time, and then mix in a separate PC DAW. It solves a lot of sync issues, and is just more practical. I strongly suspect there would be a lot of drift and possibly flutter trying to do the same thing with a real Amiga.

[chip]

Are you really able to spot the difference between the sound output of the real Amiga and WinUAE audio emulation ?

Personally speaking i don't own a real machine, but, as i already posted, the MP3s i listened for comparison was created with the real hardware

You can have info about this project here
http://www.paula8364.com/
just to know the details about how recordings was done

And i made this simple try:
just chosen one module, set up WinUAE with EaglePlayer and recorded a .WAV;
then i converted the WAV to MP3 and listened this resulting one comparing it to the one on SOAMC

I repeated this process for different tunes

Result:
my ear doesn't hear difference at all

As simple as that, so, for me, WinUAE audio emulation is perfect

[Toni Wilen]

There was discussion in some other thread (didn't find it), one reason is volume behavior. (It is counter that counts from 0 to 64 in chipset clock rate, when counter is below selected volume: DA is enabled, when it is higher than selected volume: DA output is disabled. Sort of pulse width modulation)

Another possibility is filter which is of course analog and may not be perfect in emulation.

[chip]

Filter was disabled, both for SOAMC recordings and mine recordings

[meynaf]

Originally Posted by chip

Filter was disabled, both for SOAMC recordings and mine recordings

There are two filters on Amiga, and one of them you can't disable.

I can sometimes hear the difference between emulation and real Amiga too, but not for most tunes. I could also hear the diff between A600 and A1200 in some cases.
That said : pretty much unimportant as long as it sounds good.

[Toni Wilen]

AFAIK A1200 only have "led" filter, A500/A600 (and others) have 2 filters, "led" filter and another filter that can't be disabled (without hardware modifications)

EDIT: All Amigas still have some filtering but A1200 filters less than A500/A600.

[bernd roesch]

its possible with a convolution algorithm to be really exact with and without audio filter, if this is need. on a real amiga play a 1 byte impule with 256 value and record the result with a PC soundcard. the result can then put in convolution. mostly it is used for reverbs but you can simulate EQ or speakers too with that.

if winuae is able to load convolution files and process the output its also possible to add ambient or church reverb or the impulse of a 1968 marshall speaker, so amiga sound fuller

but you can use equalizer APO for win 7 and above. this can load such impulses. here is a page how to create a more advanced way. but use only 1 byte click impulse is ok, when you have no microphone record

http://www.grebz.fr/simulator_impulses_creation_eng.php