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

[Karlos]

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 :-)

I use OctamedSS for pretty much all my composition, using it to drive MIDI kit as well as sample playback. Mix mode works well but these days I tend to render stuff to AIFF then do post-processing and mixing on the rendered output.

[Karlos]

All of my musical equipment has at least 7-bit resolution (often a lot higher) for volume and pan since that's the basic minimum standard required for General MIDI conformitiy.

FWIW, I agree with Piru that you should separate the traditional Paula model from the new one, providing the former as a wrapped interface to the latter.

When designing the 'Super' features, consider implementing features from the perspective of an API such as AHI as well as thinking about current and future module formats.

I seem to recall, AHI uses 16:16 fixed format for volume and pan, which is pretty high resolution indeed. Suddenly hardware 8-bit panning resolution doesn't sound that excessive any more. Ideally, SuperPaula would be able to accelerate a decent chunk of AHI functionality.


Going back to a purely musical view, if you could add tuneable low pass filters to each channel, I'd be very happy indeed :-)

[Karlos]

bloodline wrote:

Karlos wrote:
All of my musical equipment has at least 7-bit resolution (often a lot higher) for volume and pan since that's the basic minimum standard required for General MIDI conformitiy.

Just checked all my Audio software and hardware, and everything is 7 bit... -63 to 63... I guess that's the legacy of MIDI...

[/quote]

You'll probably find there are ways to send 14-bit data as MSB/LSB pairs. The GM requirement is that you can send a 7-bit value which depending on the parameter is treat as signed or not. On some of my kit, these 7-bit parameters are regarded as 'course' controls and have separate RPN messages to set the 7-bit fraction LSB 'fine' control.

Some controllers are capable of directly interpreting a 14-bit successive byte-pair value. IIRC, pitch bend is one of them.

Going back to a purely musical view, if you could add tuneable low pass filters to each channel, I'd be very happy indeed :-)

12pole resonant (with Adjustable Q) low pass filters please :-) Some distortion effects might be easy to implement too...

While we're at it why not add a filter and amp envelope?

Well, provided you have registers for the basic controllers I don't think there's much harm in letting the CPU do envelope control :-)

[Karlos]

AeroMan wrote:
What about if instead of just one volume register for each channel, use two registes? One for left and other for right.

That isn't really any different to having an equal resolution volume/pan pair. The advantage of volume and pan is that it's a friendlier interface.

As a musician you usually want to adjust a given part's volume or stereo location. Having separate left and right volume controls implies that you always need to modify both in order to adjust either property.

You might as well make the hardware support it rather than force the software layer to translate pan/volume into a pair of left/right volume values.

[Karlos]

bloodline wrote:
What I do need is a latency under 6ms!

Hence the reason my hardware synth hasn't been replaced by some jumped up bit of software ;-)

[Karlos]

Going back to point 4 (I think it was) on my suggestion list, the ability to turn off interpolation all together should definately be an option.

As a musician, one of the things I use paula for is her dirty, heavily-aliased sound when playing back samples. Additionally, the non-linear sample value->amplitude is rather nice.

I suggest, in addition to the ability to turn off interpolation, perhaps that 8-bit sample playback should have an optional 8->16-bit lookup table that emulates the response curve of the original Paula.

[Karlos]

Remember the cybersound 14-bit driver calibration tool?

One could probably create something similar to estimate the response curve.

[Karlos]

The problem I have with that approach is that a very low frequency output of that nature is not likely to make it through the analogue amplification stage.

The calibration file produced by cybersounds 14-bit calibration tool probably already contains the information you need, albeit in the inverse.

[Karlos]

I was thinking the onboard amplifiers on the motherboard would basically filter out your very low frequencies as they look too much like DC when coming up against a capacitor.

You'd have the same problem on your sampler input too, I'd expect.

[Karlos]

bloodline wrote:

I'm loathed to use the 14bit calibration, since I know nothing about how that works!!

Me neither. I planned to simply compare a fresh uncalibrated file with the calibrated and see if I could intuit the relationship from the differences ;-)

[Karlos]

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 ;-)