Advanced Paula Features

[minator]

Paula was mainly used to play samples and IIRC very occasionally wavetables.

However, it was capable of more than just sample playback.

It should also be capable of AM and FM synthesis.
There is an order you can set the channels to modulate each other with modulation being AM, FM or both.

The FM modulation is 14 bit but being read as a pair of bytes means it'll only run up to 14KHz.
The AM modulation is the same but it's only 6 bit out of the 16 bits read.

You can chain the 4 channels together so it should be able to do something like 4 operator FM synthesis (or AM or AM/FM).

Did any Amiga software ever do this?

[giantclam]

FMsynth? (it's on aminet)....there was (at least) one other, Synthia...

[TCD]

This thread might be an interesting read (and watch): https://eab.abime.net/showthread.php?t=114590

[Karlos]

Technically it would be Period Modulation, not Frequency Modulation. It's an important difference because the period value is inversely proportional. What people usually think of with FM, i.e. Yamaha OPL etc. is also a different mechanism again, Phase Modulation. This is a numerically (and computationally) better behaved approach.

The Period Modulation is not especially useful musically but you could presumably get some interesting clangs out of it

[demether]

there is good soft synths on the amiga, if you want the blip bloop blop feeling. Aegis, octamed, ahx, etc...

[khph_re]

Da Jormas demo Poing mucks about with the modulation:

https://www.pouet.net/prod.php?which=65872

[Paulee_Alex_Bow]

Originally Posted by TCD

This thread might be an interesting read (and watch): https://eab.abime.net/showthread.php?t=114590

Thanks for mentioning my video. The assembler code to turn on AM/FM is in there. Of late, I've noticed that TREG and Deluxe Sound might use this feature too!

[minator]

Originally Posted by khph_re

Da Jormas demo Poing mucks about with the modulation:

https://www.pouet.net/prod.php?which=65872

So someone used it, interesting. Thanks.

[minator]

Originally Posted by Paulee_Alex_Bow

Thanks for mentioning my video. The assembler code to turn on AM/FM is in there. Of late, I've noticed that TREG and Deluxe Sound might use this feature too!

Didn't realise you'd covered that in your video.
So, it does appear to have been used, but mostly as a ringmod like effect as far as I can see.

I'm wondering has it ever been used for synthesis, as in FM synthesis?

[Paulee_Alex_Bow]

Originally Posted by minator

Didn't realise you'd covered that in your video.
So, it does appear to have been used, but mostly as a ringmod like effect as far as I can see.

I'm wondering has it ever been used for synthesis, as in FM synthesis?

Hi

As all the various FM softsynths just seem to render samples, it doesn’t appear that this has ever really been used for synthesis. I find the “FM” very unpredictable, even with two lots of the same waveform, and I feel that very careful cycle accurate data would need to be used to get that working. Da Jormas struggled with this too.

The AM is far more usable on any pair of waveforms, of course.

Fun news though, I have requested that this gets added to OctaMED Soundstudio and turned on per step (I hope it can be done!) so hopefully this feature will inspire a whole new breed of Amiga music.

[8bitbubsy]

Originally Posted by Karlos

Technically it would be Period Modulation, not Frequency Modulation. It's an important difference because the period value is inversely proportional. What people usually think of with FM, i.e. Yamaha OPL etc. is also a different mechanism again, Phase Modulation. This is a numerically (and computationally) better behaved approach.

The Period Modulation is not especially useful musically but you could presumably get some interesting clangs out of it

AFAIK, 'frequency' doesn't always imply units of Hertz. A period is also a frequency unit, isn't it? It's just inverted and has a different curve. While the behavior/outcome is different, in the end it's just about semantics, unless frequency modulation has a strict standard specifying that the frequencies has to be in Hz. Just a thought...

[Gorf]

Originally Posted by 8bitbubsy

AFAIK, 'frequency' doesn't always imply units of Hertz. A period is also a frequency unit, isn't it? It's just inverted and has a different curve. While the behavior/outcome is different, in the end it's just about semantics, unless frequency modulation has a strict standard specifying that the frequencies has to be in Hz. Just a thought...

Well, since it is as you say inverted, "period" or "duration" or "wavelength" are NOT frequency.
Same way as distance is not speed. That's not just semantics but physics.

[Karlos]

Originally Posted by 8bitbubsy

AFAIK, 'frequency' doesn't always imply units of Hertz. A period is also a frequency unit, isn't it? It's just inverted and has a different curve. While the behavior/outcome is different, in the end it's just about semantics, unless frequency modulation has a strict standard specifying that the frequencies has to be in Hz. Just a thought...

Phase modulation, which is what digital FM synthesis tends to use, is something of a special case that isn't reproducible by simply modifying the frequency (or period) of an oscillator . The reason for this is that the output of a modulator is use to modify the input phase of the carrier by an arbitrarily large amount in either the positive or negative direction. You can add a negative phase offset larger than the positive forwards value you get from the time index, in some sense driving the oscillator *backwards*. The phase difference wraps at the duty cycle of the waveform, that is, driving a sinewave forward or back by 4pi is the same as 2pi, which is the same as zero.

When the input phase is increasing linearly, the effect on the carrier is as if it's running at a higher frequency. If the converse is true, it's as if it's running at a lower frequency. Thus phase modulation can achieve true FM, but it's not what makes it useful as a means of creating complex spectra.

[CCCP alert]

Using multipart loop points for carefully created samples via the CPU would allow for something like virtual analogue synthesis and give you something similar to ADSR.

Not much you can do about filter deficiencies without using professional equipment to do the 'Prodigy style' filter effects on the samples created first but it's the same multipoint sample loops per sample principle to get around it. 7mhz 68000 isn't going to be able to do real-time modular synth style filter calcs IMO

But the basic bus architecture and DAC quality is enough to do this IMO.

[minator]

So it's FM achieved by period modulation. So you'd need to invert the waveform.
And probably shape the modulator wave so it does linear FM.

From reading the Amiga Hardware Reference Manual, it says the modulator replaces the equivalent registers. It's a bit vague so I'm taking that to mean it's just writing into the period register directly.
This could get complicated because this means the centre frequency of the modulator will determine the pitch of the carrier. There's no way to change the pitch of the carrier independently of the modulation. As soon as you play a different note, I'm not sure what will happen but it'll probably go horribly out of tune.

To actually play different notes, you'll need different waveforms for the modulator with different centre points.

I'm beginning to understand why nobody did a synth around this, it'd be a complete pain!

[Karlos]

I don't know if the intention for it was to be musically useful but I did wonder if it could be used to encode data in a manner akin to how the zx spectrum did it. You'd have your carrier playing a fixed square wave output at one of two different period values, controlled by the modulator, which is DMA fetching the values from a buffer. That buffer would be filled by inserting the appropriate period value depending on the state of the next bit to be encoded.

[Karlos]

FWIW, integer arithmetic is not ideal for phase modulation. The OPL hardware does it using a form of floating point, IIRC. It uses, among other things, logarithmic encoding too, so that multiplication of values (which is very common) can be achieved by adding.