Tried all three again and I had the same result.
-29 Is just silent in another slot and after reboot.

-25 and -27 Sounds very different, In patch 1, -25 have a second slow attack, and -27 have just one attack. I think that some envelopes are behaving different. Patch 1 should be a Piano sound, so it makes more sense what I hear in ver -27

-27 is with fixed attack/release EG so I suppose it should be better

Just to test -30 is the same as -29 but with "classic" C mod matrix implementation.

This one works.
It sounds like ver -27

Yes, as it should. Funny thing.

upd: Something is broken in FM48 - good test NTS-1 hangs immediately on select
Investigating, not likely the same as sporadic no sound with FM48, just a bug.

upd2: OK clear to test FM48 1.1-30 at dev size - fixed ASM code issue I left when was debugging %(

FM-48 1.1-30 is working, and it started to sound closer to a 4op synth

I tried reloading the one I made before, several times, with the same results - either silent, or sometimes a note held forever.

Same results, except I noticed voice 2 in bank 1 was silent, on 1.1-25.

1.1-30 Works well.

What are the other destinations for Shape Assign and Alt Assign?

I notice results with 1, 30 and 40-42 as destinations. The FM6x ones do something similar but I think there may be more.

All is documented there https://github.com/dukesrg/logue-osc
1 velocity
2 feedback
X1 op. X level
X2 op. X rate scaling
X3 op. X wave
(where operator numbers is in the order of processing, i.e. reverse to DX7 standard numbers)

Mmmm... Sombody is watching you...

https://youtu.be/unEjvOcGN8U


https://www.sinevibes.com/korgnode/

1.1-31 is ut. Nothing special, just algorithm source table optimization that will free a couple of cycles and other small optimizations.

Cool. Thanks. I can't immediately see how to map, that or the tables on github to actual numbers I can select on the synth, but I'll come back to it.

I've found FM48 1.1-30 isn't saving the settings properly - at least not for this example, on the XD:

Start with an Init patch
Set the user oscillator as FM48 (1.1-30, with the four factory banks loaded)
Set the level (VCO1,2 Off, Multi as needed)
Set Voice 18 (staying in bank 1), for a simple organ sound.

Now save it, and play it to hear the sound, then switch to the next preset and back. It no longer sounds the same (it's much hasher), but it can be fixed by selecting voice 18 again.

Saving, then switching to another patch and back again, repeats the same problem.

I get the same problem with 1.1-31.

I've attached it, with the four factory banks.

Maybe this is because parameter values for selecting bank and voice are filtered for change. Otherwise oscillator will be thrashed with voice load while twisting the knob os NTS-1. Since on logues parameters controlled with the encoder knob, maybe this should not be the case. I'll check that.

The other known issue is about the voice load on oscilator init/switch to, the voice sounds like initialized incorrectly.

Actually, I think only the algorhytm is initialized wrong...

Maybe you're right. This one directly affects the voice and have no way for panel value initialization from the voice.
Well, any assigned parameter and velocity could be affected (depending on which value change logue is processing first - parameters or Shape)

1.1-31 is working well here

I think if you select the patch and re-select the correct algorhytm will have the same effect as re-selecting the voice.
Try to select the saved patch then go to parameter 6 (algorhytm) and select 19 (the correct algorhytm for patch 18). It should sound as the original...

EDIT: Also try saving the patch with the correct algorhytm selected, then change the patch and re-call it to see if you get the right sound...

Lol! Artemiy does really nice work :0) completely different from Duke's awesome stuff.

I wonder how this 96khz claim works. Calculates two entire samples per frame, then averages them?

It was just a joke

It seems a nice oscillator, but much more limited than FM64...

Okay. The sounds it makes when I select voice 18 matches algorithm 32 - whether that's the correct one or not, that's hwt I've been hearing.

However, after selecting voice 18, then looking at the algorithm, it shows as 1.

What it saves as appears to be algorithm 1, but that's not what I hear until I save it, switch to another patch, and switch back again - or pick another algorithm, then pick 1 again.

If I select voice 18, then go to the algorithm (which shows 1), then set it to 32, I hear it the way I've been hearing it. If I then save it, and switch to another patch, and back again, it still sounds the same, and still has 32 selected for the algorithm.

So, it seems the problem isn't saving or recalling the patch, it's that it uses a different algorithm than the one shown, after selecting voice 18. Apparently a different one than intended too.

If I set the algorithm to 19, and save it, that works as it should (but 32 is better - kind of a Hammond sound).

I know. I did a double take when I saw this morning. But then there's lots of FM engines; the more the merrier! I'm still waiting on an NTS-1.

I think you are refering to patch 17 E ORGAN 1, a kind of Hammond organ that uses algorithm 32. Patch 18 is PIPES 1 tha is a tubes organ and uses algorithm 19... is it possible that you have the patch (Voice) numbers shifted? strange...

Anyway, when you select a voice it sets the correct algorithm, no matters what it is set on paarameter 6 "algorithm", until you go to parameter 6 and change it. The exception seems to be when you select a patch, in that case the voice is set to whatever voice you have saved but also set the algorithm, if you didn't change the algorithm, it is set to 1. But if you save the correct algorithm with the patch it should be recalled correctly...

I was talking about FM48 1.1-31 (with the four factory banks), bank 1 (or A), patch 18, which appears to be "16 8 4 2 F", assuming they're loaded in order.

I guess that is meant to be a pipe organ sound, but it sounds more like a Hammond, to me. The Hammond was originally intended to imitate the pipe organ though, so that seems fair.

There isn't yet a strong correlation between the names of the patches, and the sounds produced, so it's hard to tell, but that's the most organ sounding organ patch I've found on it, so far.

Bank 1, patch 21 should be "Farcheeza", and it's not as I remember it, but it is a biting organ sound, which I quite like, and does sound kind of like a Farfisa, from YouTube demos of them I've watched.

I had the same thing happen with bank 1 patch 21, more or less - saving it, after selecting the voice, then going back to it, gave a different sound. Manually setting the algorithm (to 32) then saving it etc, worked okay.

I seem to have seen it show different values for the algorithm, after selecting the voice, on different attempts, but maybe I just got confused by the process of trying to workout which algorithm it was supposed to be.

The problem is we can't set the parameter value from the inside of the oscillator.
From the one side, selecting bank or voice changes actual algorithm.
But from the one size, algorithm param shows the the value initialized by firmware on selection.

IMO parameters initialized in 3 ways:
1. on the first oscillator select after boot, parameters are zeroed, Shape/Alt get current positions (?)
2. on the subsequenct oscillator selections parameters are restored from previous session, Shape/Alt - ?
3. on patch recall parameters and shape/alt initialized from the values saved in the patch

So depending on the order of the parameter value initialization it could be
- restore voice & bank (new voice initialized 2 times)
- restore algorithm (those will mess algorithm in 1. and if algorithm was not altered in 2 & 3 too)
OR
- restore algorithm (actually will set the algoritm for the current voice)
- restore bank & voice (new voice initialized 2 times)

The similar funny thing is happening wit the assignable number & Shift/Alt initialization/restoring. I consider freeing the AC #1 (0) just not to mess up velocity to zero every time oscillator is first initialized.

Technically I can track the very first parameter change invocation with setting the initial internal parameter value to the negative. But I'm not sure this is the only case when oscillator is first initialized, if loading a preset with non-initialized oscillator won't initialize parameter's twise, this trick won't work correctly.

Ahhh, OK, I was refering to DX7 Rom1a as we usually do, my fault.

I think the initialization of the algorhitm is important, as having to match the algorhitm to save the patch is a bit confusing, but I think the AC is not a big problem, as is easy to set the velocity to taste and save the patch...

Tried different things - looks like there is no way to filter out initialization parameter change from the user change, so algorithm in param really breaks current voice.
Should I move it to the assignable controller then?

Is there something special about the behaviour of the algorithm parameter? Is it related to the position it appears in, in the menu, or do all of them do that? It seems it remembers the Voice setting correctly, for example.

It would be a shame to not have velocity assigned to the Shape knob by default, because that's the main timbral change in FM.

I moved velocity to AC 1 exectly to be used by default, but as a side effect velocity itself is reset to 0 on oscillator change.
All parameters do the same, presumably they initialized in order, so voice/bank is set before and last algorithm is set and ruin the default one just set bye voice/bank change. If algorithm moved to param 1, it wil always be superceeded by the dafult algorithm from the voice on oscilator change/preset load. And I don't like the idea moving it before voice select.
The only reason to move algorithm to the parameters is to visualize it on NTS-1. So we can sacrifice this feature and move it to the assignable parameter. With parameter display turned on on logues it will be easy to destinguish the algorithm number.
Also I can bind velocity for Shape exclusively, it should fix velocity recall on preset load. And keep Alt assignable for patch tweaking.