Paula ROCKS!

[Foebane]

I don't care what anyone else says, Amiga's Paula was THE best sound chip (bar none) in a home computer system when it was launched 35 years ago (and finalised in silicon a year before). It may be "lo-fi" (whatever the hell that means) but the audio versatility compared to everything else before it is unmatched, and it is the main reason I still listen to Amiga music to this very day: it can sound as low-key as a chip-tune with simple waveforms, or full-blown module music in 4 or 8 channels, or even pseudo CD quality with 14-bit audio in the best high-end demos!

Paula ROCKS, and nothing else at the time could beat it for the price. I dare anyone to try!

[gimbal]

You and sandruzzo should start a fanclub

http://eab.abime.net/showpost.php?p=...3&postcount=27

[Thomas Richter]

Originally Posted by Foebane

I don't care what anyone else says, Amiga's Paula was THE best sound chip (bar none) in a home computer system when it was launched 35 years ago (and finalised in silicon a year before).

Oh well. Paula sound is a very simplistic design: Four DMA channels followed by a D/A converter. Nothing really unusual at its time. The Mac had something similar, just a two-channel D/A converter. Unfortunately, one of its two channels was required to control the rotation speed of the built-in floppy (wierd stuff indeed) to enlarge its capacity a bit - a Woz design of sorts.


Of course, the principle worked with the Amiga because it had sufficient RAM to place the samples in. Earlier home computers did not, and thus had to use other synthesis methods for music. In terms of complexity, SID (of the C64) was a milestone and added a lot of new stuff (me, saying, actually coming from the Atari side). Paula was not, it is rather basic.

[8bitbubsy]

Originally Posted by Foebane

I don't care what anyone else says, [...]

Not sure why you would make a thread about this if you don't care what anyone else has to say about it...
But I agree though, Paula was amazing for its time, especially considering the way they managed to pull this off by using as little logic as possible (for cost reduction).

Originally Posted by gimbal

http://eab.abime.net/showpost.php?p=...3&postcount=27

[Foebane]

Originally Posted by Thomas Richter

Of course, the principle worked with the Amiga because it had sufficient RAM to place the samples in. Earlier home computers did not, and thus had to use other synthesis methods for music. In terms of complexity, SID (of the C64) was a milestone and added a lot of new stuff (me, saying, actually coming from the Atari side). Paula was not, it is rather basic.

Who cares if the hardware is simplistic and basic? It's whether it SOUNDS good or not, that's what matters. And Paula blows that Atari ST rubbish and your precious SID chip out of the water!

[ross]

Originally Posted by Thomas Richter

Oh well. Paula sound is a very simplistic design: Four DMA channels followed by a D/A converter. Nothing really unusual at its time.
...
In terms of complexity, SID (of the C64) was a milestone and added a lot of new stuff (me, saying, actually coming from the Atari side). Paula was not, it is rather basic.

I disagree here.
It is absolutely not 'rather basic'. The tons of cheap/internal PC audio cards are rather basic...

Paula's design may also be 'simplistic', but the combination of a versatile DMA/manual usage and setup, variable and fine period data fetches / frequency selection, internal analog out of phase mixing, and volume through pwm approach, make it sound unique (and permitted some fancy and probably unwanted feature like 56Khz and 14 bit play).
And this is why a card like Gravis Ultrasound has been so successful among PC owners and game/demo coders: it has implemented Paula's philosophy on PC.

So no, Paula is not as a simple 'Four DMA channels followed by a D/A converter', and every time I read it I feel compelled to answer

But I agree on SID, it's absolutely a milestone, much more than Paula.

[Foebane]

Originally Posted by ross

But I agree on SID, it's absolutely a milestone, much more than Paula.

Perhaps for the type of audio synth SID is going for, but it's so LIMITED compared to PCM and samples.

Have a look at the attached image, which is of Paula from 1984 - does that look simple? I think SID would be simpler, somehow.

[ross]

Originally Posted by Foebane

Perhaps for the type of audio synth SID is going for, but it's so LIMITED compared to PCM and samples.

Well, SID is from a different era and with different approach to audio.
And is not 'limited', features on SID are great considering where it was intended.

But I'm biased, I love SID

[chip]

I'm not able to tell nothing from a technical point of view

But are you guys really able to listen to C64 music ?

I have nothing against C64, SID chip and all regarding this wonderful machine

But i really soon get bored after the listening of 2 or 3 C64 sids

At the opposite i never stop to listen to Amiga modules (both Protracker and exotic ones)

[Thomas Richter]

Originally Posted by Foebane

Perhaps for the type of audio synth SID is going for, but it's so LIMITED compared to PCM and samples.

Err... apples and oranges. Do you know why Paula does not make any sense in an 8-bit system? Where do you put all the samples with only 64K available. Both the C64 and also the atari 8-bit can play sampled sounds. Interrupt-driven rather than DMA driven, but the possibility is there. It is just not a very practical approach simply because there is not sufficient memory available, so it wasn't used a lot.



Thus, instead, SID (or POKEY, for that matter) selected a different approach, namely parametric generation of music. It uses a lot less memory but requires a more complex chip design. Paula moved the sample generation from hardware to software, requiring a more capable base system - more memory, more CPU power.

Originally Posted by Foebane

Have a look at the attached image, which is of Paula from 1984 - does that look simple? I think SID would be simpler, somehow.

Err, what? I am sorry, but that is not how you measure "simplicity". PAULA is not only audio, but is rather a multi-channel UART, a programmable serial communications chip, along with a D/A converter. There is nothing special about such chips, they existed already at this particular time. Actually, PAULA is an upscaled version of Atari's "POKEY" chip, which was again an upscaled version of an UART chip.



So, yes, the overall design of audio generation is rather simple and there is not much novelty in the Audio department here.

[Thomas Richter]

Originally Posted by ross

I disagree here.
It is absolutely not 'rather basic'. The tons of cheap/internal PC audio cards are rather basic...

Your generic AC97 PC audio circuit is just a lookalike of Paula. Two DMA channels (instead of four), and two 16-bit A/D converters (instead of 4 8 bit D/A converters). Thus, exactly the same design.


More advanced PC cards have more facilities (such as filters, MIDI generation in hardware etc...).

Originally Posted by ross

Paula's design may also be 'simplistic', but the combination of a versatile DMA/manual usage and setup, variable and fine period data fetches / frequency selection, internal analog out of phase mixing, and volume through pwm approach, make it sound unique

So where is the "unique part"?

Originally Posted by ross

(and permitted some fancy and probably unwanted feature like 56Khz and 14 bit play).
And this is why a card like Gravis Ultrasound has been so successful among PC owners and game/demo coders: it has implemented Paula's philosophy on PC.

As explained, PAULA was neither the first nor the last design in this direction. Look at the Mac hardware - or the Atari POKEY design. The latter contains four 4-bit D/A converters instead of 4 8 bit D/A converters, and there is no generic DMA processor (aka Agnus) in the Atari to feed them, so the CPU has to do it. It was a very plausible design direction to use DMA channels as well. Not only feed the display data by DMA, but also the audio data.

Originally Posted by ross

So no, Paula is not as a simple 'Four DMA channels followed by a D/A converter', and every time I read it I feel compelled to answer

It is exactly that simple. There are no software-controlled analog filters as in SID, no waveform generators as in SID, no poly-counters as in POKEY. Of course, there was no need for all that as the CPU can manually compute the samples - or load them from disk - but I admire the additional hardware logic that had to put into the older chips to make them play music or generate effects. PAULA is a design that left most of that to software - given that CPU power was available to do the job the hardware had to do in the earlier chips.


Of course PAULA is more versatile than SID with this design, but at a price: The CPU has to jump in. It could, because it was more powerful, and there was more RAM to support it.

[8bitbubsy]

Originally Posted by ross

And this is why a card like Gravis Ultrasound has been so successful among PC owners and game/demo coders: it has implemented Paula's philosophy on PC.

I kinda disagree that GUS uses Paula's philosophy... GUS is doing resampling at a low'ish rate, while Paula has a fixed high rate using sample-and-hold. Because of the high Paula rate, you also get less errors (less aliasing). GUS' low mixing rate and linear interpolation makes it a dull performer, in my opinion. Still was a revolution to the PC demoscene at the time, though.

[commodorejohn]

Does Paula use S&H with a single DAC? I was under the impression it was four totally discrete channels from beginning to end.

[Foebane]

Originally Posted by ross

Well, SID is from a different era and with different approach to audio.
And is not 'limited', features on SID are great considering where it was intended.

But I'm biased, I love SID

Well, it's not limited in what I've heard it perform in demos, but then did the designers have in mind what the talented game and Demoscene musicians have done with it since?

Originally Posted by chip

I'm not able to tell nothing from a technical point of view

But are you guys really able to listen to C64 music ?

I have nothing against C64, SID chip and all regarding this wonderful machine

But i really soon get bored after the listening of 2 or 3 C64 sids

At the opposite i never stop to listen to Amiga modules (both Protracker and exotic ones)

I can listen to SID a lot more easily than say, the Atari ST tone generator chip (that latter chip is poor).

[Foebane]

Originally Posted by Thomas Richter

Err... apples and oranges. Do you know why Paula does not make any sense in an 8-bit system? Where do you put all the samples with only 64K available. Both the C64 and also the atari 8-bit can play sampled sounds. Interrupt-driven rather than DMA driven, but the possibility is there. It is just not a very practical approach simply because there is not sufficient memory available, so it wasn't used a lot.

Fair enough, but I was comparing 16-bit with 16-bit and upwards, I just didn't mention it. At least that's what I had in mind with the OP.

Thus, instead, SID (or POKEY, for that matter) selected a different approach, namely parametric generation of music. It uses a lot less memory but requires a more complex chip design. Paula moved the sample generation from hardware to software, requiring a more capable base system - more memory, more CPU power.

And that's why more power is better, IMHO.

Err, what? I am sorry, but that is not how you measure "simplicity". PAULA is not only audio, but is rather a multi-channel UART, a programmable serial communications chip, along with a D/A converter. There is nothing special about such chips, they existed already at this particular time. Actually, PAULA is an upscaled version of Atari's "POKEY" chip, which was again an upscaled version of an UART chip.

Really? I knew Jay Miner was behind both in general, but I didn't know the similarities were that close. But I forgot that the Jay Miner sound chips are also keyboard controllers and joystick/paddle controllers, too (in the case of POKEY). Of course, in the photo I provided, I'm sure the audio portion is on the right, or the areas copied four times.

[ZEUSDAZ]

It's the best!!!!!!
[ Show youtube player ]

[nikosidis]

Not all DACs sound the same and Paula does sound good. Always liked the sound it produces. Use your ears. All the tech. talk does not matter much.

[Paulee_Bow]

I love Paula, and the super-charismatic early sampling sound it offered.
Basically a poor-man’s Fairlight.

So versatile, that software allowed Paula to function as:

- A wavetable (in the PPG sense) synth
- An additive synth
- Sample Playback
- Hidden assembler functions allow for FM and AM modulation between channels
- Run simple simulations of subtractive synthesis (Like Aegis Sonix)

There is so, so much to explore with Paula.

PS, I have a dastardly plan to run the two channels through my Eurorack modular rig, whilst simultaneously MIDI controlling that very rig from the Amiga. Insanity awaits!

[8bitbubsy]

Originally Posted by commodorejohn

Does Paula use S&H with a single DAC? I was under the impression it was four totally discrete channels from beginning to end.

Paula has four DACs internally, one per channel.

[Thomas Richter]

Originally Posted by Foebane

And that's why more power is better, IMHO.

Certainly, though keep in mind how much power was available in the 8-bit area. I'm just attempting to judge the quality of the work, and hence the height of the invention. For SID, that was quite a substantial move forward - it was an integrated music synthesizer made by a musician for musicians. It's wasn't very good at special effects - POKEY could do that better.


Of course the PAULA design could do more, but it had a more powerful environment available. The pure height of innovation, i.e. the step from POKEY to PAULA, was rather small. Just collect the sample data from the DMA controller (aka Agnus) and widen the D/A converter from 4 to 8 bit.


All the "poly counter stuff" POKEY had was removed as it was unneeded. What remained was "channel linking", this is something POKEY already could do, though in a different way.

Originally Posted by Foebane

Really? I knew Jay Miner was behind both in general, but I didn't know the similarities were that close. But I forgot that the Jay Miner sound chips are also keyboard controllers and joystick/paddle controllers, too (in the case of POKEY). Of course, in the photo I provided, I'm sure the audio portion is on the right, or the areas copied four times.

They are pretty close, really. The whole chipset design of the Amiga looks like an "Atari on steroids". The DMA controller, which is Agnus in the Amiga, is called ANTIC in the Ataris. It feeds the display generator, GTIA on Atari, Denise on Amiga, and then we have the UART and sound generator, POKEY on the Atari, PAULA on the Amiga. One PIA on the Atari, two CIAs on the Atari.


POKEY had 4 channels that could be linked in various ways, could generate square wave sound (aka "DMA disabled mode" on PAULA), could feed this square generator into "Poly counters" for effects generation (lacking in PAULA) or could run in "VOLONLY" mode, aka "4 bit D/A converter mode", which enabled software-generated speech, for example. PAULA had this as well, with the addition that the DMA controller can feed the D/A input.