pouët.net

Amiga - native Paula sound

category: general [glöplog]
I tried an experiment to make mixed stereo MP3s of some of my favourite Amiga demos, and what I feared has come true: the file size has blossomed by 1.5 times the original size! I have several dozen Amiga demos and mods in mono on there already and it already consumes a lot of space, which is limited to me, and so, just as I was starting to like mixed stereo sound, I'll have to scupper that plan.

The thing is, the mono MP3s are actually stereo sided, but identical sides, so the LAME encoder doesn't acknowledge the difference by data at all - I think that's what Joint Stereo means.
added on the 2013-09-28 22:11:20 by Foebane72 Foebane72
All my life I was having 2 stereo speakers not too far away from each other in front of me and Amiga stereo separation was not a problem at all. But on headphones it is PITA indeed.
added on the 2013-09-28 23:11:00 by rutra80 rutra80
@Foebane72: mp3's Joint Stereo is 100% stereo, it just stores data in different format than left/right (it's middle/side instead) but it still gives you 100% reproduction of stereo image.
added on the 2013-09-28 23:17:44 by rutra80 rutra80
When you turn interpolation on, you no longer hear it as it would sound on real Amiga.
added on the 2013-09-28 23:21:49 by rutra80 rutra80
Rutra80: Yes, I figured it was something like that.
added on the 2013-09-28 23:51:34 by Foebane72 Foebane72
Something I've just discovered today is:

Why do the Amiga demos by the group Black Lotus, namely "Ocean Machine" and "Starstruck", have mono soundtracks on the Amiga when professional 16-bit stereo MP3s of the tracks exist? The videos have Amiga graphics running, but with the MP3 tracks instead of the native platform's sound - is it cos it sounds better? The native sound is not even in stereo! Did they lack the disc space for it?

And how do those two demos sound on the Falcon platform? As good as the MP3 soundtracks? I don't have access to a Falcon myself, emulated or not, so I don't know. But I'm hoping someone does.
added on the 2013-09-29 18:44:09 by Foebane72 Foebane72
Quote:

Why do the Amiga demos by the group Black Lotus, namely "Ocean Machine" and "Starstruck", have mono soundtracks on the Amiga when professional 16-bit stereo MP3s of the tracks exist?

Maybe because 16bit stereo mp3's eat up quite a bit of CPU-time? Hence adpcm-versions are used instead.

Quote:
The native sound is not even in stereo! Did they lack the disc space for it?

Stereo adpcm soundtracks eat up twice the amount of memory (and discspace) compared to mono, so that could very well be the reason to use mono.

Quote:
And how do those two demos sound on the Falcon platform? As good as the MP3 soundtracks? I don't have access to a Falcon myself, emulated or not, so I don't know. But I'm hoping someone does.

The Falcon-versions use an mp2-version of the soundtrack, so they do sound better than their Amiga-counterparts.
added on the 2013-09-29 18:48:45 by britelite britelite
Quote:

Maybe because 16bit stereo mp3's eat up quite a bit of CPU-time? Hence adpcm-versions are used instead.


OK, fair enough. The Amiga wasn't known for its CPU processing prowess.

Quote:
Stereo adpcm soundtracks eat up twice the amount of memory (and discspace) compared to mono, so that could very well be the reason to use mono.


So you're telling me the whole thing is in memory at one time? I thought Amiga AGA demos with HDD access could load each part in as it was running, like PC demos do?

And is Amiga HDD diskspace really at that much of a premium?

Quote:
The Falcon-versions use an mp2-version of the soundtrack, so they do sound better than their Amiga-counterparts.


But not as good as the MP3?
added on the 2013-09-29 19:16:00 by Foebane72 Foebane72
Quote:

So you're telling me the whole thing is in memory at one time? I thought Amiga AGA demos with HDD access could load each part in as it was running, like PC demos do?

It could be a memory issue, I don't know for sure. Usually it's better to load up everything to RAM during startup to avoid possible timing-problems caused by a slow harddrive.
Quote:
But not as good as the MP3?

Probably not quite as good, but way better than the adpcm-version :)
added on the 2013-09-29 19:40:47 by britelite britelite
Thank you, Britelite :)

Just goes to show that Amiga sound just couldn't cut it back in the Noughties, or even before.

WHAT were Commodore THINKING, when they neglected to design a Super Paula to go with the AGA chipset? (facepalm) Amiga music got left behind in the dust, and lesser platforms like the Atari Falcon took over, so that ultimately they had the better audio, both native and MIDI :(
added on the 2013-09-29 20:54:23 by Foebane72 Foebane72
You can't really compare interpolation to Amiga directly.
Interpolation is done when you are doing software-mixing, because you are outputting a single waveform at a fixed frequency, and all samples have to be resampled to the pitch of the note they're playing.
Interpolation is a smoother way to perform this resampling, causing less artifacts.

An Amiga however has 4 separate channels, each with their own A/D converter, and so each note is played at the proper frequency directly, rather than resampling it first (the DMA speed is adjusted to pitch).

So strictly speaking, neither interpolated nor non-interpolated software sounds exactly like an Amiga.
added on the 2013-09-29 21:22:02 by Scali Scali
(that should be D/A converter obviously)
added on the 2013-09-29 21:22:39 by Scali Scali
Quote:
WHAT were Commodore THINKING, when they neglected to design a Super Paula to go with the AGA chipset? (facepalm) Amiga music got left behind in the dust, and lesser platforms like the Atari Falcon took over, so that ultimately they had the better audio, both native and MIDI :(


yep, exactly my thoughts. It should have been sth like ES5503 or this (SNES).

ppl. still composed some great music on the Amiga, despite the restrictions of the old 1984 Paula design, but I was really jealous of the Atari Falcon sound capabilities back then (but did not have the $$$ to afford one).
added on the 2013-09-30 01:27:51 by xyz xyz
Quote:
ppl. still composed some great music on the Amiga, despite the restrictions of the old 1984 Paula design, but I was really jealous of the Atari Falcon sound capabilities back then (but did not have the $$$ to afford one).


That was down to more memory in the A1200 and A4000 leading to cleaner samples, but there was no other improvement.

I too was really jealous of the Falcon, and the implications of what the DSP meant and so forth, but luckily the Falcon died as hard as the Jaguar did, cos no-one wanted either.

So we got off there, I think.
added on the 2013-09-30 01:37:42 by Foebane72 Foebane72
BB Image
added on the 2013-09-30 13:57:31 by Deltafire Deltafire
Quote:
So strictly speaking, neither interpolated nor non-interpolated software sounds exactly like an Amiga.


There's a lot of confusion about this, because there are so many kinds of interpolation. Popular ones like linear, cubic, and sinc don't sound right. Minimum-phase bandlimited step (with or without the A500 and/or LED filter) sounds right.
added on the 2013-09-30 17:26:39 by absence absence
the Amiga had, like other professional grade studio samplers during that era, per channel variable sample rate.

These things are not easy to emulate since today's audio hardware usually implements one fixed-rate audio stream.

(..and I do not know of any consumer grade multichannel audio cards that let you set the sample rate per channel..)

The Amiga also had, like the document reed linked to describes, a builtin filter to smooth the waveforms.

You could enable the LED to turn off an *additional* filter which resulted in quite a muffled sound so it was rarely used.
added on the 2013-09-30 21:55:30 by xyz xyz
-off +on
added on the 2013-09-30 21:56:26 by xyz xyz
and yeah, if you're looking for actual Amiga sound emulation / replay code, personally I prefer this:

$ ./uade123.exe --headphones2 /cygdrive/h/mods/protracker/stu/mod.toi-tei

:)
added on the 2013-09-30 23:35:07 by xyz xyz
xyz, variable sample rate is not entirely true I think. All the channel rates are AFAIK subdivisions of the base clock rate, so you could instead say that it has a very high sample rate :) (3.5 Mhz?). Maybe it's mostly a matter of how you define variable sample rate.
added on the 2013-10-01 11:42:11 by linde linde
Yes, it is 'variable' in the sense that you can specify the replay rate per channel.
This is done by setting a 'period' value.
This is a countdown value. Everytime the counter reaches 0, a new sample is retrieved via DMA and output.

So yes, your sample rate is limited by the granularity of the period counter. The timer interval is 279.365 nanoseconds however, so it has very fine granularity.
added on the 2013-10-01 13:22:29 by Scali Scali
(yes, boils down to roughly 3.5 MHz)
added on the 2013-10-01 13:26:04 by Scali Scali
Quote:
@Foebane72: mp3's Joint Stereo is 100% stereo, it just stores data in different format than left/right (it's middle/side instead) but it still gives you 100% reproduction of stereo image.


Whoawhoa, hold your horses right there.

The point about the mid/side encoding is that the encoder can then allocate less bits to the side channel than to the mid one. That's mostly ok for usual stereo content, as many reproduction/radio systems use mid/side with disadvantages for the side channel and thus stereo music is usually mastered in a way that most of the stuff is going on in the middle of the field, but it's very likely to degrade quality on an amiga where L and R are actually two completely separate entities.

So if you want to encode MODs with 100% stereo separaion, always use full stereo.
added on the 2013-10-01 13:50:44 by kb_ kb_

login