RazorBlade osc question for Spogg and/or tulamide

[k brown]

I kept hearing an odd couple of notes jump out with the PIANO wave, so I checked on a scope and found that for just a couple of notes around the middle of the keyboard, the wave has a considerable gap (flat line) in the waveform, that is very audible as a bright zing in the sound. I checked SQ8L and it's not there.

Also found it in the STRING wave, but for that, the flat spot is only not there for about the top octave.

I'm pretty sure both those waves were samples; maybe a connection?

Considerable flat spots in REED2, but that's probably an additive wave and they're supposed to be there.

But PIANO and STRING clearly have something going on. I fear there's a ghost somewhere in the DSP.

Piano.jpg (31.23 KiB) Viewed 31981 times

String.jpg (24.56 KiB) Viewed 31981 times

Reed2.jpg (24.73 KiB) Viewed 31981 times

[tulamide]

I kept hearing an odd couple of notes jump out with the PIANO wave, so I checked on a scope and found that for just a couple of notes around the middle of the keyboard, the wave has a considerable gap (flat line) in the waveform, that is very audible as a bright zing in the sound. I checked SQ8L and it's not there.

Also found it in the STRING wave, but for that the flat spot is only not there for about the top octave.

I'm pretty sure both those waves were samples; maybe a connection?

Considerable flat spots in REED2, but that's probably an additive wave and they're supposed to be there.

But PIANO and STRING clearly have something going on. I fear there's a ghost somewhere in the DSP.

Piano.jpg

String.jpg

Reed2.jpg

My part of the work was to get the waveforms from the original rom and separate them into .wav files. We took all care needed to make sure that start and end points of the waveforms were correct, and that was not as easy as it may sound. Once we "cracked the code", so to speak, and figured out how the waveforms were put together one after the other and still be distinguishable [this is a word I always have to look up for spelling, I fear this word], it was easy then. We had to correct a few, where the markers were mixed with actual sample data, but we tested all of them in the end.

That's why I think, that something must have gone rogue, while you experimented with them. I'm not the expert for that part - all the really difficult sound engine parts were done by Spogg - , but I think there's a ramp or counter or something involved, that Spogg uses, and maybe the length there got mixed up somehow?

But I'm sure, Spogg will have more helpful ideas!

[k brown]

Couldn't have been anything I did (or Spogg, doing the phase input); the screen shots were done inside RazorBlade, so nothing to do with the recent phase input mod.

Good info, though - Since I didn't understand a single word in the lengthy thread(s) during development, I've never been really sure who did what. Spogg is very humble in his user guides.

Just noticed the 'x' in thread title, any way to correct a thread title, or only mods can?

[Spogg]

Just noticed the 'x' in thread title, any way to correct a thread title, or only mods can?

I’ll have a look at this later. In the meantime I’ve corrected the title spellings.
As a side note, you can edit your own posts but not other people’s.

[Spogg]

OMG! Going back to such a complex synth was quite a shock to my poor old brain, but I found the issues.

Each multi-wav “wave” has custom DSP code to tell the wav reader the sample size according to which wav is indexed in the wav array. I called that index the “Tag”. Turns out I’d made mistakes in the Piano and String DSPs, giving 512 instead of 1024 for example.

In fact the Reed2 was fine and you can tell because the waveform is never flat.

When I was testing the Piano, I found the transition from B6 to C7 (top of the 5-octave keyboard with C4 being middle C) was truly dreadful. It changed from a nice high note to a vile sawtooth with aliasing. So I checked the original spreadsheet we used and that was correct, according the table. It could well be a mistake when the table was created, so I took the liberty of changing from a nasty sawtooth to a much more pleasant “more aggressive sine.wav” and to my ears the transition is much improved, to the point where it sounds almost continuous.

I’ve attached the updated schematic. If you can confirm there are no more nasty surprises in store I’ll update the zip download and publish it, and the update will be a good excuse for reminding people about it.

Many thanks for the feedback Kevin.

[tulamide]

So I checked the original spreadsheet we used and that was correct, according the table. It could well be a mistake when the table was created, so I took the liberty of changing from a nasty sawtooth to a much more pleasant “more aggressive sine.wav” and to my ears the transition is much improved, to the point where it sounds almost continuous.

Well, while I'm glad, that my thought process of the lengths being the issue wasn't off, the real issue shows again, how difficult it was to extract the waveforms from the ROM. I'm pretty sure, the error happened, because once again a certain length was pre-defined for the ROM, while the actual waveform was 0-padded. It was such a pita! We obvioulsy didn't find all of the candidates, although I vividly remember the ones we found.

[k brown]

Aha - A code violation!

Thanks tons for the investigation. Those two waves are particularly nice ones.

Been having a blast building things around the new phase input for the oscs. I thank you both profusely in the Notes pane of the projects I'm using them in; I know it was an exhaustive process for you both. I also put a hot link to Rex's site on the GUI for the Synclavier-inspired project, which I think is pretty well ironed out now. It's a bit of a monster and can eat up quite a bit of CPU with some settings (when all four Mod/Carr pairs are used and they all have the 'Chorus' function on, there 16 RB oscs running!). I tried what Rex did with RazorBlade - having switches to turn off each LFO and EG that are not being used, but even with them all off, I couldn't see a consistent drop in CPU (?). After watching a video on how the MJ 'Beat It' intro was constructed I was able to dial it in fairly close on my plug. When they played the unmodulated Carrier waveform, it sounded exactly like the SQ-80 'SAW2'. I called the preset "Can't Beat The Thrill". As they point out in the video, it has to be played by hitting one key (it's a monophonic sound) and immediately releasing it, to get the expected sound; if you hold the key down, the sound dies very quickly.

I'm including this PDF (from the dulcimoo.com website) in my download folders - it has great info/guidance on the SQ-80 waveforms (for example, that the 'Piano' wave sounds best when dropped an octave, and how to use the 'Kick' wave most effectively). You might want to have it in RB's folder as well?

ENSONIQ SQ-80 Waves.pdf.zip

(226.57 KiB) Downloaded 1285 times

I'm wondering what you both think about incorporating the phase-input oscs into RazorBlade - having the FM option modulating the actual SQ oscs, rather than just a sine as we originally did. If so, I can take care of it, and you guys can QC it, or if want to do it yourself that's of course OK too.

Thanks again so, so much.

[Spogg]

I totally forgot I had that SQ8L plugin so I checked the high Piano notes using a scope and they weren’t using the aliasing sawtooth wav. Digging deeper I realised I’d originally not loaded the SAW2 wav but the raw sawtooth one, intended for the lowest pitches.
In fact the “more aggressive sine.wav” sounds just as good to my ears so I’m inclined to leave it in place.

I’ve always said that anyone can do anything with any of my stuff, and that’s still the case, but from my position the RazorBlade is complete (unless it needs a fix, like just). Of course, you can go ahead and make any changes you see fit, but it might be best to rename it RazorBlade FM or something, just to avoid confusion and to differentiate it from the original. That’s my opinion and maybe tulamide has a different idea…

[k brown]

I thought the 'Piano' wave was a sample.

[tulamide]

I’ve always said that anyone can do anything with any of my stuff, and that’s still the case, but from my position the RazorBlade is complete (unless it needs a fix, like just). Of course, you can go ahead and make any changes you see fit, but it might be best to rename it RazorBlade FM or something, just to avoid confusion and to differentiate it from the original. That’s my opinion and maybe tulamide has a different idea…

I see it just the same.