SuperPAULA - if you have experinece in amiga music please give feedback

[biggun]

Hi,

Thomas, and me did some brainstorming which sound capabilities make sense to include in the enhance PAULA chip of the NATAMI.

If you have experience in creating music then I would appreciate if you could participate in the discussion.

We want to include good enhancements into the new Paula chip. Of course the focus should be put on only putting important addition to the chip and not bloat it with unnecessary stuff.

As we know the original Paula had
- 4 channels (2 Left / 2 Right)
- each supporting 8 Bit samples
- and in addition to this 6 Bit Volume.
- The resulting audio output quality was 14 bit.
- All 4 channels used DMA to read their sample data.
- The DMA of two channels could be combined to read both
the sample and the volume values in for one channel.
- Doing this Paula could DMA create 2 channels with 14 bit.


The Natami New Paula:
- Audio output quality upgraded to 24 bit.
- Support for both 8-bit and 16-bit samples.
- combined with 6-bit or 8-bit volume.

I'm interested in your educated opinion, whether it makes really sense to increase the number of supported channels over 4.

A few opinions that we had:

a) The original Amiga could have a sound at left or right.
If you wanted to have a sound in the center both left and right then you needed to play it in two channels.
Having a center channel would be nice addition.

b) Octamed showed that its relative cheap to virtuelle increase the number of channels by mixing channels with the CPU. If you for example play a MP3 in a game on the natami then it will use 2 channels (1 left / 1 right) - This would leave you two free channels for effects. The CPU can mix many virtual channels into those in no time at all.
In less time then decoding a mp3 you can probably mix 1 dozend extra channels - So having extra HW audio channels might be important at all.


If you have experience in creating music it would be nice if you could add your opinion.

Cheers
Gunnar

[biggun]

Actually octamed 8-channel mode was ... very CPU hungry.

Yes, octamed was CPU hungry on a A500.
Of course the memory bandwidth of the old 68000 and in general form the CPU to the AMIGA Chipmem was a real limiting factor even on AGA.

But the Natami with 68060 will have no problem at all, to mix 16 or more channels.

You have to mind that the NATAMI 060 is not only running at 90 MHz and that the Fastmen is many times faster than the AMIGA fastmem was, and most important that the bus between CPU and Chipmem is many many many times faster than it was on classic AMIGAs.

In theory you could create a DMA channel on the NATAMI (like special blitter mode) that can mix sources. But as the CPU could mix for free it probably not worth doing.

Even a 48KHz sound with 16Bit samples is only 94 KB per second. If you have 16 channels this is 1.4 MB sample data to mix per second. That just about 1% of the memory bandwidth of the Natami 060, so really no problem.

[biggun]

Hi Riftcon,

Thanks for your input

* A panning register so channels could be moved from left to right.

Valuable point, thanks.
How much resolution of the RIGHT/LEFT gradition is really needed?
It it needed to have a fine graine of 256 values or
would a simpler right/left position of something like (0%, 25%, 50%, 75%, 100%) be enough too?

* 4 channels IS a bit limited.I would definitely prefer 8

Thanks, noted.

[biggun]

spihunter wrote:
These days a pro musician or even someone recording a demo would need more capabilities then a 24bit Super Paula could provide.

Oh, the Natami is NOT designed for pro musicians at all.
Sorry if my question sounded like this.

The question was to ask experienced music developers, as what in their opinion are the minimum requirements for good sound in a Amiga home computer.

We all agree that CD is good quality sound, and for perfect CD player quality all you need is 2 Channels with 16 bit samples each.

So 4 channels with 24bit quality is a very good start, I think.
I acknowledge that HW support for having channels in between left and right, and a few channels more will help creating AMIGA mods with zero CPU useage.

[biggun]

riftcon wrote:

Hmm, I think 5 bits of precision would be good enough. It would be nice to be able to pan a playing sound from left to right and back in not all too discrete steps. At least not too audible. I guess 4 bits MIGHT be enough, but slow panning might suffer.

It would also be great for positional sound in games.

EDIT: I just looked it up, the Gravis Ultrasound got away with 4 bits (unsigned, 0-15) and that worked just fine in trackers on the PC. So I think 4 bits would be fine.

Yes, I think one of the first AMIGA games doing this
panning of sounds from LEFT to RIGHT and back was Archon.

They did this by playing the same sample on one channel on each side and alter the volume to get this effect.

Actually the HW needs to do about the same work for panning channels.
For each channel that can be panned, the HW will have to mix two channels - One to the right and one to the left.
In other words for the HW supporting one panning channel
needs as much HW mixer resources as two non panning channels.

[biggun]

Piru wrote:
The best solution would be to allow old and new audio to co-exist, this way you can have old apps using audio.device or direct hw register banging after allocating the channels, and new apps using the high definition audio thru AHI.

Interesting point, I'll note this one.
But do you think any user will run two music applications old/new at the same time?
That like playing AmiNetradio and the Game Hybris at the same time, isn't it?


A question to AHI:
Can you tell us what feature might be helpful to run AHI/SDL sounds with low CPU usage?

Cheers

[biggun]

warpdesign wrote:

I'm interested in your educated opinion, whether it makes really sense to increase the number of supported channels over 4.

Of course it does... And the more the better...

If you are experienced in creating module music,
please tell us how many channels are in your opinion are needed for a good mod. If you have no experience in creating music, then please do not respond.

If mixing is not noticable in today computers, it sure will be with a slow 060

What you say is non true at all!

Even with a stock 68000 you can mix 8 channels.
For a 68060 mixing 8 extra channels takes no time.

[biggun]

quote]
I totally agree with Piru, have a decent Paula emulation... and a nice modern audio chip with an AHI driver... Audio is a commodity now, technologically it's reached a plateau... you can't really do anything innovative (as is still possible to some degree in the GFX space). Great audio chips can be bought off the shelf for next to nothing.[/quote]

I know were you are coming from.
I think there is a very important point you are missing here. It might be my fault explaining it not good enough before.

The think is that:

-The Natami is fully Paula compatible already.
-The Natami can already switch to a new mode providing working 24bit Paula enhancements.
-And the Natami has working upsampling to 96kHz
-And the Natami has a very high quality DA converter.
The DA is of higher quality than the typical "next to nothing" sound card will have.

I fully agree with you that it makes no sense try to beat professional sound studio cards with the Natami.
If you need Studio feature the simplest solution is to buy such a card. And of course you can plug in any PCI soundcard into the Natami if you want.


But as the Natami sound device in on the board already
and as is able to produce good quality sound.
It makes good sense to think about how to get the most out of it without less effort.
Adding some extra features as panning, or 8 channels support
are only little firmware work for us now.
So if more channels is what people want, or if support for interleaved wave data is what would speed up AHI
then we can add this all "for free" now.


Please mind that adding an extra sound chip to the board is not a good idea. Putting another chip on the board would require that its DA-out gets mixed back with the Paula signal. This solution would consume a lot extra space on the board. As it would require extra wires from the FPG, we would not a more expensive FPGA to connect that extra chip.
And of course this would make the mainboard a lot more expensive as the board would need to be bigger. And this would require a new board layout.

The audio out of the Natami is quite good quality for a consumer device.
We have now the possibility to add some nice features to  the Natami audio for free.
We could as well add new features with a firmware update at a later time.

If you have input and sensible wishes now, fine
If not then this is okay for us too.


Cheers
Gunnar

[biggun]

Also it would be nice if the end spec was published, so, for example, MiniMig could implement that audio spec (on a hardware level at least, maybe the outputs wouldn't be the same quality) which would increase the market for that implementation, thus increasing the software that would support it.

This is an excellent point!
Thanks for bringing this up.

I'm a big fan of open specification and fully agree with you that this is an important part.

You are absolutely right, that if the new Amigas will
share a common Paula audio enhancement this would be a good thing.
If the new Generation of Protracker Mods will play both on  MiniMig and Natami this would be really cool.

It think it will be real nice if we as Amiga community will manage to come up with a good spec for New Paula all together.

Cheers

[biggun]

Karlos wrote:
IMHO, as a musician, I'd like to see the following things per channel for an enhanced Paula:

1) 8/16-bit sample depth support
2) 8-bit volume register
3) 8-bit pan register (-127 full left, 127 full right)
4) Optional interpolation for playback of sample data below the hardware mixing frequency.

If you can do that per-channel, then I'd like to see at least 16 of them :-)

1) 8/16-bit sample depth support
I agree, and this is included already.

2) 8-bit volume register
I agree, its also included.

3) 8-bit pan register (-127 full left, 127 full right)
I agree, that panning is useful.
Its certainly good to be able to put some voices in the center. But I feel that 8 bit panning volume is overkill and not needed for most cases. I would prefer have simpler HW  panning support for the majority of cases.
I think the question is how often to you need panning.
I think every song can make use of having a drum "in the middle" or "having the bass in the 1-part LEFT / 2 parts RIGHT position".
For the rare cases where you need 256 positions of panning "for the helicopter if circling around you effect" you can always use two channels and do the panning the normal ways over the volume.

4) Optional interpolation for playback of sample data below the hardware mixing frequency.
I fully agree. And the new Paula ie "Pamela" chip does HW oversampling of all samples automaticly.

I think it to make the HW support those feature that most songs will need.
If a rare case does needs something special, then its okay to use the software for this. Even the not upgraded Natami, is at least over 100 times faster than a classic A500.
Its okay to use the CPU from time to time.
But it would be nice if we could catch the majority of cases with HW support. E.g if 95% of the AHI music does play without CPU usage and if the majority of SDL games will play without CPU usage then this is great.

What is your opinion as musician:
How many HW channels are needed to play 90% of the mods?

Cheers
Gunnar

[biggun]

I am not a musician, but ...

Only a musician should make statements like "xyz is needed for this and that in music".

We have discussed this before: On AMIGA it was always possible to make these panning effect using two channels (one left one right) and use the 6-bit volume to create panning effect.

It was mentioned before that AMIGA games used to use these effects. 6-bit volume was good enough for this effects on classic Amigas. Natami has 8-bit volume, so really enough bits for high quality panning.

We stated before that certain PC cards that provide panning use 4-bit resolution for this.

I think a panning for new channels is a useful addition to have sounds coming from center or middle-right center etc.
For "placing the instruments" on the stage 2-4 bits is more than enough. For high quality effects you can always allocate two channels getting 8-bit resolution.

My 2c

[biggun]

AmiDelf wrote:

How about adding an DSP? Like the one used on DelfinaDSP? Other than that, I wonder what sort of audio output do you have? Will it be minijack or phono?

Take Care

A good example of a System using a DSP was the ATARI Falcon.

At the time of the Falcon, adding a DSP made sense.
The Falcon main CPU was quite slow, the 030 of the Falcon was just slightly bit faster than a stock Amiga 1200.

The situation is now different.
The NATAMI 060 is at least 10 times faster than the Falcon. The Natami can do those effects with its 68k, which at the time of the Falcon were only possible because of the DSP.


There is not that much value of adding a DSP just for Audio effects to the AMIGA.

But having a general purpose DSP with like ALTIVEC features  could make sense. Such a general DSP could be used to various tasks ranging from MP3 audio decoding, to DVD playback, or acting as Texturemap shader.
We are evaluating these options.

Cheers
Gunnar

[biggun]

in the SID you get away with this because the sample rate is very high (around 1 MHz?).

The sample rate of the ADA is 192 KHz
The goal for the internal clock rate of Pamela is double of the CPU clock, thats 180 MHz in the case of the NATAMI60.

[biggun]

@all

Many thanks for your good contributions!

I feel that this brainstorming was very good and really productive. I think the feedback from the musicians was very valuable.


At riftcon:

riftcon wrote:
So I've had a look at the various AHI documents, the device interface in particular and some device source code.

As I suspected the AHI mixing routines output interleaved stereo sound in signed 16 (or 32) bit. So if the audio device doesn't support interleaved samples, the CPU will have to do some work. The Paula driver splits the stream itself.

Thanks for looking this up.
This is very valuable information.

I was looking at the AHI source briefly.
I'm not sure if I understood how/if AHI supports more HW channels.
Can you or anyone help me to understand this?

The question is:
If the HW for example supports 8 independent Channels, each with 8 or 16 bit samples, independent frequency and independent volume.
Will AHI make 100% use of this. I.E. will a 8 voice audio piece run fully on HW or will AHI still do some software mixing or upsampling of the channels?

Many thanks in advance