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

[SamOS39]

8 channel mode on octamed is VERY unstable  :-P

ummm .. doesnt octamed support 16bit samples..? im sure ive imported 16bit samples intooctamed v6 ..?

[bloodline]

Karlos wrote:

bloodline wrote:

That's rather clever and almost certain to take a million years...

It's seemingly just a lookup table. I wasn't going to do it by manual inspection. I was going to graph the differences ;-)

Well get on and do it... if not let me know, and I'll build a look up table based on my idea...

[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

[platon42]

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?

AHI supports Mono, Stereo and 7.1 (aka output channels, no other number of channels (so if you really want to have 5.1 channel output, you would need to use the 7.1 mode).

Changing the number of supported output channels is a PITA, because there are dozens of optimized mixing routines for each sample type combination (see SelectAddRoutine in mixer.c) and addroutines#?.[c|s].

However, AFAIK it supports any number (<256?) of "voices" (again with the given width Mono, Stereo and 7.1) that may be mixed together to that output HW channels mentioned above.

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?

There is a mode where the driver may take over the mixing completely (makes it possible to do this in hardware or via DSP). I just looked it up, you can actually mix different sample types of a "voice" to the output channels.

AFAIR the Paula DMA drivers use this mode (no mixing is done, but limited to <4 "voices").

Martin Blom is still around these days and is very helpful

[riftcon]

platon42 wrote:

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?

There is a mode where the driver may take over the mixing completely (makes it possible to do this in hardware or via DSP). I just looked it up, you can actually mix different sample types of a "voice" to the output channels.

It sounds like you have done a lot more AHI programming than me From reading the programmer API docs I think this mode requires  you to allocate the HW channels from the driver and control them yourself with pitch, volume, pan etc? Which is fine I guess, but it doesn't look like AHI will use the HW channels automatically even if they're adequate for the job?

In this mode it also looks like you query the driver for its supported pan, volume etc resolutions and you have make sure you set them within the defined range. There's no mapping of values, which I like. It just exposes the hardware at a low level.

[ChrisH]

Can anyone here publish the original Paula response curve? I'd love to write a paula effect processor!

Actually... I probably have all the equipment needed to find it out myself... does anyone know a decent menthod?

I would imagine the best way to *generate* the Paula response curve would first require you to know what kind of D/A converter Paula used.  I *vaguely* recall it used a series of (9?) resistors to get the voltage drops required for each of the 8 bits, which is not very accurate due to the resistor tolerances, and therefore tends to produce a non-linear response curve.

Assuming my recollection is correct, then you need some method to find the relative values of each of those (9?) resistors.  That should be MUCH more accurate & less error prone, especially compared to trying to sample the output of Paula across her entire dynamic range.

I imagine that the 14-bit audio calibration tool used some clever method to accurately find out the relative resistor values, so that it could then compensation for them.

[bloodline]

ChrisH wrote:

Can anyone here publish the original Paula response curve? I'd love to write a paula effect processor!

Actually... I probably have all the equipment needed to find it out myself... does anyone know a decent menthod?

I would imagine the best way to *generate* the Paula response curve would first require you to know what kind of D/A converter Paula used.  I *vaguely* recall it used a series of (9?) resistors to get the voltage drops required for each of the 8 bits, which is not very accurate due to the resistor tolerances, and therefore tends to produce a non-linear response curve.

An R-2R Ladder DAC eh? I guess that would work for something as simple as 8 bits... Ok... are these integrated on the Paula IC or on the Board? I need to find a Paula pinout..

Assuming my recollection is correct, then you need some method to find the relative values of each of those (9?) resistors.  That should be MUCH more accurate & less error prone, especially compared to trying to sample the output of Paula across her entire dynamic range.

It would be easy to work out... but that's not going to give me real word performance.

I imagine that the 14-bit audio calibration tool used some clever method to accurately find out the relative resistor values, so that it could then compensation for them.

IIRC the calibration tool was rather subjective... which required you to decide when the response was linear... it was a long time ago...

[bloodline]

Looking at an A500 schematic it seems the Audio DACs are actually in Paula herself... pretty cool...

[Cheeeky]

That answer lies not in what can you make the hardware do, but what the blend of software and hardware can do.

basically for the pro/semi-pro it is PCI Cards and AHI drivers, USB Midi should definitily have drivers written but this is all out of the scope of your question. Yet in pure digital the line between what is pro and what is not is blurred...


My Personal preference:

sp/dif or optical out should be a serious consideration.

basic upgrade for paula standard should be:

8 channels

left right assignable

8/16/24 modes

forget the centring - anything like that is stereo and played at different volumes - I think people are getting confused by dolby surround - yes it has a centre channel, but this has an independent output and requires a compatible amplifier system.

How may outputs are you thinking of putting on the board (could we have 6 mono outputs - assignable?) **

What I would really like to see is a paula mode + toccata emulation mode for 6 channel output with existing software, but this requires writing a toccata emulation library - a stable one!

** extra upgrade mode would definaltly be 32 channel mode set up as above but there needs to be an AHI driver created for this to work with existing software (I still want a working toccata library also :-) ).

So, without a toccata library or AHI driver the thing is pretty much useless to use with existing software - if these can't be written then stick with 4 channels but with the upgrades that you have already specified and see what happens with 3rd parties... We can plug in PCI cards.

on a side note - I have considered DSP stuff (my Flipper is great), but once again a serious lack of software in this department will hold this back and may be a wasted feature addition - Personally I would serious consider the ability to add Cell Clusters (oh yes!) to the system.

[polardark]

Cheeeky wrote:
sp/dif or optical out should be a serious consideration.

S/PDIF might reduce the sound quality since the most common implementations are not capable of 192 kHz sample rate. One option might be hardware resampling although this would most likely increase the hardware complexity and/or quality.

If professional musical use is a goal then ADAT Lightpipe is a more sensible direction to go in than S/PDIF since it supports many channels and/or sample rates up to 192 kHz. Then again, considering the limitations compared to modern electronic music equipment on the market, there are no reasons to use the SuperPAULA for professional audio unless it ends up having something very unique to offer.

Either way, personally i see few reasons to justify implementing S/PDIF.

[bloodline]

Karlos wrote:

bloodline wrote:

That's rather clever and almost certain to take a million years...

It's seemingly just a lookup table. I wasn't going to do it by manual inspection. I was going to graph the differences ;-)

The CyberSound preset calibration file is a table of 256 1byte values (I assume they are signed, since Paula is...)... I wrote a quick program to examine it... I'm not sure exactly how these numbers are used to transform the audio?

I'm going to build a few graphs until I get something meaningful...

[bloodline]

Ok... I think I've figured it out...

Each byte in the the calibration file seems to correspond to the difference between what one would expect  the 8bit->16bit value to be (using simple LSL scaling) and what the Amiga actually produces... This is the only way I can make sense of the file... anyone have any other ideas?

[minator]

One thing nobody has mentioned yet is the difference how the Amiga played audio and how modern systems do it.  Modern systems interpolate samples to get different notes then play this at a fixed rate (44, 48, 96KHz etc).

The Amiga didn't do this, it played the samples at different rates.

This is an important thing to consider because the low sample rates Amiga apps used generated a lot of quantisation noise and this was of course part of the Amiga's sound.

If you try and interpolate samples up to a higher fixed rate you'll lose this sound.  If you are upsampling (as seems to be the case) the number of sample you add per sample will be different for each rate, this means you'll be adding quantisation noise but in a different way.  Many of these frequencies will above human hearing range but not all.

So, in short if you're not careful you could end up with 8 bit Amiga audio sounding very strange.  I guess it depends on how accurate you want it but if you want high accuracy you'll need 2 different systems.

--

BTW why are you using a 68K processor?  Wouldn't it be better to use a PPC, that way you can emulate the 68K and you'd have a faster processor.  It also gives you at least the possibility of running the likes of MorphOS* (assuming you could talk them into supporting it).

*I'm not saying OS4 because it's in legal limbo and will probably remain there.

[Cheeeky]

As the sample rate can be any value I understand that Digital-Out would be complicated as it won't sync - but would it not be possible for the hardware to enable the digital out when a syncronizable play rate is chosen? such as :
8,000
9,600
11,025
12,000
16,000
22,050
24,000
32,000
44,100
48,000
88,200
176,400
192,000

I've never heard of ADAT Lightpipe and don't have any equipment that would support it... thing is, it doesn't make the music any better ;-)

[tiffers]

@anyone who knows

The talk about panning register, and using 2 channels with volume to create your own panning solution is an interesting one.

Is it always true that a voice/channel is hardwired to output to just one speaker? Classic Amiga hardware was this way. Is new hardware this way? Is the FPGA board used for NatAmi this way? Imagine if you had 16 voices, but what speaker they came out on (literally, which RCA connector on the back of your Amiga) depended on this panning register. Thus you would only need 1 voice to be generated or played, but this output is sent to potentially both channels, but with various resistance on each channel to effect a reduction in volume, to effect panning effects.

ie
100% left 0% right (register = 1000)
75% left 25% right (register = 0100)
50% left 50% right (register = 0000)
25% left 75% right (register = 0010)
0% left 100% right (register = 0001)

Only with more resolution than that. (say the 7 or 8 bits suggested in previous posts)

tiffers