Let's talk about a 32-bit Amiga Paula
category: general [glöplog]
Paula, the 16-bit DMA sound chip in all Amigas launched from 1985 onwards, was a revolutionary sound chip when most others could only produce simpler effects, and even though it is a relatively simple 4-channel 8-bit DAC with frequency and volume control, with modulation effects built in, it has established a high profile within the Demoscene, due to its wide variety of musical styles facilitated by the MOD format developed and refined by talented programmers since then.
Unfortunately, Commodore never saw fit to release a 32-bit version, to coincide with the AGA chipset.
If this had been developed, what do you think it would've consisted of? 16-bit PCM waveform audio, certainly, and how many channels? 8? 16? And would it have had DSP support or had one built in, as the Atari Falcon did? I guess we'll never know, but we can speculate, unless specs are available somewhere?
The existing Paula chip sounds awesome for 1985 when PCs could only manage simple speaker beeps, and the SID could only do synths easily but not samples, but it seems Paula has a simple trick up its sleeve that can approximate CD quality audio, namely the 14-bit mode. Each 14-bit channel ties up two regular 8-bit Paula channels to achieve the effect, but since Paula has 4 channels, a better-sounding stereo solution is presented that Amigas with HDDs can use to play back long single samples like CD tracks, which is still only 16384 possible speaker positions as opposed to a CD's 65536, but is still better than the mere 256 of basic 8-bit Paula!
So how would the Amiga have revolutionised professional computer music production if Commodore had released a 32-bit Paula-like chip alongside AGA? Will we ever know?
Unfortunately, Commodore never saw fit to release a 32-bit version, to coincide with the AGA chipset.
If this had been developed, what do you think it would've consisted of? 16-bit PCM waveform audio, certainly, and how many channels? 8? 16? And would it have had DSP support or had one built in, as the Atari Falcon did? I guess we'll never know, but we can speculate, unless specs are available somewhere?
The existing Paula chip sounds awesome for 1985 when PCs could only manage simple speaker beeps, and the SID could only do synths easily but not samples, but it seems Paula has a simple trick up its sleeve that can approximate CD quality audio, namely the 14-bit mode. Each 14-bit channel ties up two regular 8-bit Paula channels to achieve the effect, but since Paula has 4 channels, a better-sounding stereo solution is presented that Amigas with HDDs can use to play back long single samples like CD tracks, which is still only 16384 possible speaker positions as opposed to a CD's 65536, but is still better than the mere 256 of basic 8-bit Paula!
So how would the Amiga have revolutionised professional computer music production if Commodore had released a 32-bit Paula-like chip alongside AGA? Will we ever know?
Since they did in fact plan to make a true 32-bit Amiga successor at some point, we know pretty well what it would have contained:
Source: https://en.wikipedia.org/wiki/Advanced_Amiga_Architecture_chipset
Quote:
Updated version of Paula called 'Mary' with 8 voices that can be assigned either to left or right channel; each channel has 16-bit resolution with up to 100 kHz sample rate; additionally it does 8-bit audio sampling input.
Source: https://en.wikipedia.org/wiki/Advanced_Amiga_Architecture_chipset
Quote:
each channel has 16-bit resolution with up to 100 kHz sample rate; additionally it does 8-bit audio sampling input.
Sounds quite retarded, doesn't it? Why couldn't it do 16-bit sampling input?
In the meantime we had 14-bit sample playback and it was good for its time with MIDI.
Note that 14-bit sample playback was never intended by the Paula's designers; it's just a hack someone figured out would be possible at some point. (And given how nonlinear Paula's DACs are, I guess it's nontrivial to actually get 14 proper bits out of it? Not that I've ever checked the details.)
Sure, but it was way good enough for samples playing along MIDI instruments.
Quote:
So how would the Amiga have revolutionised professional computer music production if Commodore had released a 32-bit Paula-like chip alongside AGA?
It wouldn't. It would just look less embarassing next to the Gravis Ultrasound, which released around the same time, and also didn't revolutionise professional computer music production.
Quote:
Note that 14-bit sample playback was never intended by the Paula's designers; it's just a hack someone figured out would be possible at some point.
We Reveal Amiga’s 14-bit Audio Possibilities
Ahem. You were saying, Sesse...?
Glenn Keller, you're my idol!
Quote:
Ahem. You were saying, Sesse...?
Is that article supposed to refute Sesse's point? Leaving aside the credibility of Amitopia, the first sentences say:
Quote:
When Commodore launched Amiga back in 1985. They would never think of that Amiga could actually output 14-bit playback sound.
Isn't that's essentially what Sesse said?
Quote:
Is that article supposed to refute Sesse's point?
Yes, because the following sentences say:
Quote:
They promoted the Amiga as a machine capable of 8 bit sound. But no one at Commodore seemed to know how well designed Paula sound chip was made, or the designers didn’t get the notice that they deserve.
Commodore obviously didn't know exactly what they had on their hands, which is not surprising, seeing as how they wanted the world's first multimedia home computer to be another IBM clone.
...what the hell are you talking about
Poor marketing of the Amiga on Commodore's part, that's what. Everyone knows that.
I don't know about how accurate that article on Amitopia is, and I've never heard of the site before now, but why would they be misleading with that article?
And what was Sesse's point, anyway?
I've read elsewhere that the "hack" is a false urban legend, and it would be a useful feature to have 14-bit as an option on the Paula circuitry, even if it only worked under certain conditions and needed extra CPU and memory to work, not to mention the chipsets. And Glenn Keller added modulation abilities to Paula as well, which are rarely used. And Sesse admits he never checked the details, anyway.
I don't mean to cause problems, I'm just defending Paula here.
I don't know about how accurate that article on Amitopia is, and I've never heard of the site before now, but why would they be misleading with that article?
And what was Sesse's point, anyway?
Quote:
Note that 14-bit sample playback was never intended by the Paula's designers; it's just a hack someone figured out would be possible at some point. (And given how nonlinear Paula's DACs are, I guess it's nontrivial to actually get 14 proper bits out of it? Not that I've ever checked the details.)
I've read elsewhere that the "hack" is a false urban legend, and it would be a useful feature to have 14-bit as an option on the Paula circuitry, even if it only worked under certain conditions and needed extra CPU and memory to work, not to mention the chipsets. And Glenn Keller added modulation abilities to Paula as well, which are rarely used. And Sesse admits he never checked the details, anyway.
I don't mean to cause problems, I'm just defending Paula here.
Quote:
why would they be misleading with that article?
Zealotry? Incompetence? I don't know, but let's take the example you quoted:
Quote:
But no one at Commodore seemed to know how well designed Paula sound chip was made, or the designers didn’t get the notice that they deserve.
Really? Not even Glenn Keller? It sounds like the author doesn't know that he worked at Commodore.
Quote:
I'm just defending Paula here.
Why are you defending it from Sesse's relatively accurate statement? And is it based on emotions or facts? If the latter, perhaps you can provide a reference where Glenn Keller, or anyone else on the original Amiga team, confirms that the ability to be tricked into outputting a noisy approximation of 14-bit audio was something they designed with intent, and not a happy accident that they didn't even know about?
Quote:
Why are you defending it from Sesse's relatively accurate statement? And is it based on emotions or facts? If the latter, perhaps you can provide a reference where Glenn Keller, or anyone else on the original Amiga team, confirms that the ability to be tricked into outputting a noisy approximation of 14-bit audio was something they designed with intent, and not a happy accident that they didn't even know about?
Points noted, absence. And much appreciated too.
I've obviously dug a deep hole for myself with assumptions of Paula audio and what I think it can do, versus what it actually CAN do, so I will just leave the floor to those who know Paula most to explain everything. I, after all, in the early 1990s, WAS once extremely zealous about the Amiga and thought it was the best computer ever for years, and I don't know the intricacies of the Amiga hardware to the level that sceners here do (I don't even code in assembly language), so I will leave all future discussions on this thread to those that DO know the intricacies of Amiga audio hardware.
Please, sceners, feel free to explain to me and others the intricacies of how to make audio on Amiga Paula. I will listen.
Quote:
It wouldn't. It would just look less embarassing next to the Gravis Ultrasound, which released around the same time, and also didn't revolutionise professional computer music production.
Exactly. What revolutionized professional computer music production was a steady increase in CPU and storage capabilities (of which the former enabled ever more sophisticated software), not more hardware mixing channels.
Quote:
Quote:So how would the Amiga have revolutionised professional computer music production if Commodore had released a 32-bit Paula-like chip alongside AGA?
It wouldn't. It would just look less embarassing next to the Gravis Ultrasound, which released around the same time, and also didn't revolutionise professional computer music production.
Oops, I forgot about the GUS.
Quote:
Note that 14-bit sample playback was never intended by the Paula's designers; it's just a hack someone figured out would be possible at some point. (And given how nonlinear Paula's DACs are, I guess it's nontrivial to actually get 14 proper bits out of it? Not that I've ever checked the details.)
@Sesse: My humblest apologies, you were right all along.
How does Paula's 14-bit audio work?
I came across this thread on English Amiga Board that covers the subject, and the technicalities are quite complex for a (relative) layman like me to understand, and even then the people who know more than me in the thread are not entirely sure themselves, but I didn't need to be all arrogant and presumptuous over my response to your post, as you too know more than I do. Sorry.
In fact, I've seen (and been in) a couple of Paula discussions and it really is quite the complex sound chip, but beyond a certain amateur point, I get lost in all the technical jargon about audio engineering and science they use.
All I can say is that, however 14-bit audio works on Paula, it sure sounds great for a 1985 device.
Then again, wasn't Paula's circuitry updated between chipsets in minor ways? I know the low-pass filter was improved, but as I understand it, that's outside of the chip.
Quote:
it really is quite the complex sound chip
I think you exaggerate the complexity a bit. Since DMA is performed by Agnus, Paula doesn't have much to do at all. It just outputs the received data at a specified rate and volume, optionally with modulation. Something like Yamaha's OPL2 (also from 1985, and used on e.g. Adlib and Sound Blaster cards) is much more complex, although the sound it generates may not be to everyone's taste. :)
While mixing samples in hardware was an impressive feature for an 80s home computer, it had to be done in the simplest way possible to keep costs reasonable.
GUS sounded really awful after adding couple of tracks more ... sound went completely muffled. Dang I've talked about this so many times but still cannot remember why it was doing this, ran out of mixing juice I guess?
Anyway. RME is where it is ;D
Anyway. RME is where it is ;D
The GF1 chip on the GUS had flexible resource allocation, so while it only had enough power to mix 14 channels at 44.1 kHz, you could indeed get more channels by reducing the mixing rate. I believe this property comes from the Ensoniq ES-5506 chip that GF1 was based on. I don't know if it's confirmed that Bob Yannes of C64 SID chip fame designed ES-5506 (and thus GF1), but since he co-founded Ensoniq it isn't unlikely.
What's the story here, GUS wasn't good enough for scene tunas, but Amiga was. Or maybe it's that with so and so many bits and sinc interpolation, your favorite Dr Outatune hits sound less bad.
The way I remember it, Amiga music was fantastically good and inspirational, and GUS was completely awesome, taking the possibilities much further and no Amiga guy could deny that, it was a night and day difference. VGA chunky pixels, Pentium power, GUS sound, there was no competition, the game turned around, after that Amiga demos imitated PC demos and not the other way around.
The way I remember it, Amiga music was fantastically good and inspirational, and GUS was completely awesome, taking the possibilities much further and no Amiga guy could deny that, it was a night and day difference. VGA chunky pixels, Pentium power, GUS sound, there was no competition, the game turned around, after that Amiga demos imitated PC demos and not the other way around.
Quote:
Amiga demos imitated PC demos and not the other way around.
And I think that was a great period of the Amiga Demoscene. The groups were using accelerators and exploring AGA and c2p routines with HAM8 and generally experimenting and trying different things, with the new standard of HDD storage offering even more scope for their works. And so many prods were created, too! Such variety! The PC may have become the king, but the Amiga could still hold its own against MS-DOS prods, especially those who continued with the Amiga.
I really did not like GUS, I can say that aloud. I had AWE64Gold :´D and it did not degrade audio with more channels - cannot remember if it used CPU for that or.. who cares, better sound (for a while, before pro audiocard)
GUS was only really relevant for
- 386/486 demos where proper software mixing would take too much CPU from the actual demo.
- 64k MS-DOS intros because the replay code was much smaller.
I don't know when PC trackers started to feature proper software mixers but I still remember early trackers (1992/1993) sounded terrible on Sound Blasters of that time (which were still 8bit). Here the GUS was really a revelation.
- 386/486 demos where proper software mixing would take too much CPU from the actual demo.
- 64k MS-DOS intros because the replay code was much smaller.
I don't know when PC trackers started to feature proper software mixers but I still remember early trackers (1992/1993) sounded terrible on Sound Blasters of that time (which were still 8bit). Here the GUS was really a revelation.
and the EMU8000 used in AWE32/64 beats the shit out of classic GUS in terms of features - more sample memory, resonant filters, hardware reverb/chorus, no sample rate drop with more channels, etc. even if the synth supports 16bit forward looped samples only. too bad that it was noisy in early revisions and Creative hid the low-level docs for a (relatively) long time, so it never caught much attention on the scene (althought it had received native support in Impulse Tracker and some players).
besides, Creative had the brilliant idea to ditch SIMMs in favor to proprietary memory modules in AWE64 (except for the Gold version which had 4MB on board instead of teeny 512kb :)
The Interwave chip (used in GUS PnP etc.) did fixed most of the GF1 flaws, but IMO it was too late to come, and still no proper SB support for non-demoscene stuff - not to mention the whole PC world moving to Windows at that time, and Gravis couldn't keep with proper Win9x drivers ¯\_(ツ)_/¯
besides, Creative had the brilliant idea to ditch SIMMs in favor to proprietary memory modules in AWE64 (except for the Gold version which had 4MB on board instead of teeny 512kb :)
The Interwave chip (used in GUS PnP etc.) did fixed most of the GF1 flaws, but IMO it was too late to come, and still no proper SB support for non-demoscene stuff - not to mention the whole PC world moving to Windows at that time, and Gravis couldn't keep with proper Win9x drivers ¯\_(ツ)_/¯
