fm module

[Niowiad]

Cool, thanks for testing.
I don't know if FM8 has upsampling? Not sure, so I reckon to test against it with MSF in no upsampling mode. Also FM8 might disable the OPs or feedback system if not used, so it might be best to actually set values in the matrix to know for sure they are contributing to the CPU usage.

I'm not aware of upsampling options in FM8, neither in Blue2. Sytrus can.
My last post had Sytrus without upsampling and FM8/Blue2 in default (probably 1x).
And the post above was about MSF at 4x but also 1x.
Then my testing was quite an approximation of the behaviour. I had all 6 operators active and modulating others a little bit.
Being quite precise would be very hard, also because the values for modulation depth in matrix slots are different for each synth (eg. a sytrus matrix slot at 100% pushes the modulation on the carrier much further than an FM8 slot at 100%).
It was just just a brief and simple look at average values on possibly similar enough conditions.

[DarkStar]

Sometimes I feel like going the extra mile (such as the dedicated optimization for nonfeedback situation) is pointless. Makes me think that it's better to create mediocre stuff... no features, no advanced stuff, people wouldn't complaint ...

Oh yes they would

[gentleclockdivider]

Fm8 can upsample , from the menu choose high quality .
I'ts easy to notice when it's upsampled .
Just self feedback an operator , it will sooner collapse into noise at 44Khz then at 88Khz

[MeldaProduction]

Sometimes I feel like going the extra mile (such as the dedicated optimization for nonfeedback situation) is pointless. Makes me think that it's better to create mediocre stuff... no features, no advanced stuff, people wouldn't complaint ...

Oh yes they would

Yeah, I know . So the result is - someone will complaint anyway, so keeping the distance is the only cure!

[Niowiad]

Fm8 can upsample , from the menu choose high quality .
I'ts easy to notice when it's upsampled .
Just self feedback an operator , it will sooner collapse into noise at 44Khz then at 88Khz

I had never seen that option (the entire menu to be honest) so I have been using it at 2x the whole time.
I just wish I had been using it at 1x the whole time so this would have been a great surprise.

Back on topic, as I was saying, I'm getting about 2% cpu load with 6 operators going at it and matrix feeback enabled (at 2x oversampling for the FM module) in REAPER at low buffer and a very shitty CPU.
Do you have another daw to test it on maybe? (I'm asking because I see many users are using all kinds of daws at the same time).
Polyphony might be another factor we haven't actually put into the equation.

[werzel]

Niowiad, Would the default ratio settings help to explain the modulation differences in Sytrus and FM8?
Sytrus' default setting is at 2x, where FM8 is at 1x. (If my memory is correct!)

[Niowiad]

I get what you're saying, but I feel like this is a different issue.
Besides, I was testing them all with operators at ratio=1.
It's true that Sytrus opens at 2.000x, and since this used to annoy me I also saved its state at 1.000x so it sounds like all the other synths when I launch it.

The difference I was talking about, I guess it just comes down to the developer's decision and taste on how to set the scale which goes from minimum to maximum in the matrix slot. That value is basically "volume" of the modulator going into the carrier, which alters the carrier's frequency spectrum.

Given the same modulation % value in the matrices, some synths seem to push more (or less) volume into the carrier than others. FM8 at 100% kind of sounds like Sytrus at 75% to me, comparing the sound of the carrier by ear.
At least that was always the way I explained it to myself

Blue2 can push it even more than Sytrus, not that it's related to quality, just to make another example.

[mevla]

Sometimes I feel like going the extra mile (such as the dedicated optimization for nonfeedback situation) is pointless. Makes me think that it's better to create mediocre stuff... no features, no advanced stuff, people wouldn't complaint ...

Or not making it available to purchase while still in beta

[exponent1]

Hi,

Is this (mis)alignment intentional or a bug?

Mis-alignment is just bad user interface design

The structure ends up being very reminiscent of the Blue2 interface with the operators lined above the matrix. I won't lie saying it currently doesn't looks odd, but in my opinion the weirder thing is not having labels for rows and columns, which initially put me off but got used quicked than I thought. Labels would make it easier to read and bypass any alignment issue (blue2 as an example)

With that said, the aligment thing is a completely different story from what you stated earlier:

-Some modules like 'fm' could really benefit from a matrix view

Because the FM module indeed has the matrix view, and IMHO much more functional than other matrices as far as modulation goes.

If it's a bug, how do we report it?

[Niowiad]

I don't think it's a bug, neither intentional. That's probably how the melda gui automatically rearranges stuff in the designated space. The operator section has labels on the left which "squeeze" the operator columns and aligns them differently from the matrix, which does not have the labels.

In my opinion, as I also stated in the quoted message making the blue2 example, the matrix doesn't necessarily need to be aligned and visually related to the operators above (although it'd look good) if it has labels for rows and columns.
My feeling is, by not having those labels for matrix rows/columns, the eye automatically checks around for visual reference and in this case finds something above, relates to it and notices the (mis)alignment.
If the eye checks around and sees labels for columns/row, it simply ends there.

If there are no plans for labels, then it would be nicer to have it aligned with the operators above for column reference (although it would still miss a reference for the rows).
Besides, putting labels for matrix rows similarly to how the operators have labels on the left, would also concidently line-up the columns with the operators.

[DarkStar]

I have already made a suggestion to Vojtech about this - just as you said: line up the columns and put "OP n" labels on each row (and Output on the last row)

[Chandlerhimself]

I think adding labels would be a good idea too. It would make things look better and it wouldn’t take mich work. I think a tabs for each of the operators would be best, but IMO that might be a lot of work for little payoff.

Another thing that might help is putting some of the less used parameters inside a pop-up for each OP. When you click OP1 for example you can have “noise”, “pwn”, “noise” etc controls and have maybe 3-5 of the most important controls on the main screen. It might help make the GUI smaller and easier to understand, while not taking away any features.

To be honest though, just the labels would be a big improvement.

[Kwurqx]

Of course I can compare with other synths that rely on phase modulation ,why should global feedback increase cpu usage ?
We're talking about simple phase modulation .
Maybe the developer can give some insight

Both Dexed and more flexible alternative Oxe FM are open source projects. So, sourcecode is available for analysis. Though Dexed only has the DX7 (fixed) algorithms and Oxe FM only has selfmodulation feedback.

Dexed
https://github.com/asb2m10/dexed

Oxe FM
https://github.com/oxesoft/oxefmsynth

Personally, I don't have any performance issues with MSF FM (yet?). My main confusion stemmed from the "reverse order" in the operator numbers...coming from Yamaha "FM"...

Since we are on the subject of MSF FM. I assume it's actually Phase Modulation? So a fixed carrier frequency of 0 Hz should be possible for "FM/PM wave shaping"?

I know the Yamaha DX/TX (and Dexed) are "limited" to a minimum of 1 Hz (and a bit of detune), resulting in non-stable carrier frequency rate spectral modulation cycle. But the SY series did offer 0 Hz (and Initial Phase), resulting in a stable spectrum.

Maybe the Oscillator Wave Shaping offers an alternative to this (Wave Shaping is how Kurzweil evaded the Yamaha "FM" patents in their mighty V.A.S.T. engine).

Maybe I'l look into this later....

[Kwurqx]

Coming from a Yamaha SY77...

Actually I guess MSF can emulate a lot of the Yamaha SY "AFM" synthesis. Still looking for the 16 SY AFM waveshapes/spectra though. The user manual is not detailed enough. It's the 8 waveshapes from the TX81Z + 8 more.

The Yamaha SY AFM algorithms have a semi-free feedback matrix with up to 3 feedback source operators.

There's is off course also the issue with the correct envelope times and curves. And the number of envelope segments, that can be looped as well.

Then there's the AWM / Subtractive part of the SY and the AWM/AFM interaction.

All and all, I guess MSF can emulate a lot of all this.

Maybe, in future, al sorts of emulations will become available as "instruments". Like the Yamaha TX81Z and SY77/99.

[Kwurqx]

Also coming from a Kawai K5 additive synthesizer.

I guess, with some creativity, much of the K5 can be emulated too. 2 x 64 harmonics (in parallel or serial up to 128), but only 4 (6 stage loopable) envelopes. But with those, individual harminics/partials amplitudes can be modulated. In MSF you'd have to set up groups of partials in an additive oscillator. And mixing spectra (same phase and/or phase inverted) or filter.

The K5 "formant filter" can be emulated using either filtering or by mixing spectra (same phase and/or phase inverted). It's actually more of an graphical EQ working directly on the individual harmonics levels (attenuating partials only).

Same goes for the "filter". You can just "draw" the "slopes" and "resonance". Very steep if needed. You can draw a Bandpass filter so narrow it will sweep through individual harmonics/partials (without "resonance" artefacts, since it is actually only attenuating partials).

Herbert Janssen has a site on the (rare and then revolutionary and generally underrated) Kawai K5.
http://herbert-janssen.de/k5.html

See also
http://www.kfuenf.org/
http://www.vintagesynth.com/kawai/kawaik5.php
https://www.matrixsynth.com/2018/11/syn ... ai-k5.html
https://soundprogramming.net/manuals/Ka ... Manual.pdf

Herbert Janssen has an ebook on SY programming too, by the way
www.herbert-janssen.de/doc/sy-prog.pdf

The Kawai K5 official sales pitch from the eighties....1987 or so. All the audio distortion/clipping is not the Kawai K5. Though it has "noisy" outputs.


And anothor audio example