If acreil associated glitches with slew then i may have missed it.
But in any case, the slewing would be caused by components further down the tube. Basically any glitches short and big enough can cause slewing. But the cause stays the same, which is glitching. Insofar there is any since i'm talking out of my looking-at-schematics-and-thinking-up-theories-ass now. And i have the flue.,

I'd defo recommend Arturia's DX7 - sounds largely faithful to the original in my opinion. Little MIDI keyboard and hey presto . Used the Arturia DX7 liberally all over this track: YouTube

Has anyone done any critical comparisons to see if all DX7's sound identically the same? I would not be the slightest bit surprised if there is as much variation between "identical" DX7's as there is between good software equivalents and the real thing.

I was wondering about that. I put this together to see for myself. Dexed is amazing for free.

That's not going to affect the output. Most DAC designs produce some manner of glitch when the code changes, because the bits don't all change at the same time, but the sample and hold eliminates that.

Right, it would be the DAC buffers or the sample and holds that slew. But probably that's not happening to a significant degree.

Replicating everything a number of times on the same IC would take up too much space. There are only a few things that do that. Most things use a single phase accumulator, a single multiplier, etc. and time multiplex to achieve polyphony. Then the polyphony is limited only by the clock frequency, desired sample rate, and the number of registers. The only exceptions are cases where the hardware can't be clocked fast enough to do this, or when it's using clock divider ICs or other things that can't be time multiplexed.

The remarkable part is that it's pipelined well enough to calculate one operator per clock, doesn't use a multiplier, and greatly outperforms everything else that was available at the time. All the algorithms are implemented with just a couple registers and multiplexers. Modern romplers, etc. are still based on custom ICs that work in much the same way. So it's not all software even now.

Accurately emulating the internal workings of mostly undocumented custom ICs isn't "software". There are a number of things in there that are quite difficult to correctly reverse engineer.

The specific sample rate mainly just affects the aliasing. No one seems to notice particularly that the Yamaha V50 has a different sample rate from the TX81z (50 kHz vs. 55.930 kHz).

The DAC distortion will vary from unit to unit.

Second round, Arturia gives much more of that ugly "machine gun" effect.
Almost like if it was a single sample repeated.
Interesting.

DEXED tested against my DX7 mk1. Set up DX7 to output MIDI to DEXED in reaper. Recorded output of DX7 direct into mac. Only processing involved was normalizing the individual files.
These are various random patches.
Guess which is which. Some are obvious, others not so obvious. It shows the differences between the 2 but also shows how close they are.
I will reveal which is which later.
Excuse the poor playing, done in a rush.

Hmmm has the file uploaded? If not how does one upload a wav? I did add it to the post and it appeared to upload but i can't see it on my post!

Sometimes it is bugged.
But for me it works most of the time if i convert the wav to high bit rate AAC m4a.

By the way, from the comparison few posts above - if DEXED is not EQ corrected, then i think it will be easy to say which is which.

Second attempt

Nope.

I use this: qaac

Test:

Ill try that

Try again

I think there's a 10MB limit for attachments, so WAVs can easily go over that.

Good comparison, it shows that they're clearly not the same. It's not so bad on the sustained sounds but the percussive ones can be quite different. The slap bass one really sticks out, DX7 at 1.23, DEXED at 1.27 ? If you're playing DEXED on its own then you'd probably not notice anything, but put alongside the real thing and the differences show up. It becomes clear that the software can sound inferior. Patches don't sound quite how the programmer intended, they would have tweaked the brightness, the clicky attack, the body frequencies etc. I also managed to lock it up by sliding my ESQ1 data slider to change patches.

I'd guess the problem with percussive sounds is either the envelope attack or the initial operator phase.

It would be nice to have an uncompressed version (it makes it easier to identify differences in an audio editor).

The author of Dexed did mention that envelopes still need more tweaking. But not sure if this is only in OPL mode, or in all modes.

Well nothing is gonna replace the experience of a DX7 MK1. I do like the Reface. It looks like it would feel DX like.

I think the biggest diffenences between my dx7 and dexed are that dexed has a slightly sharper attack and is a little more sensitive to velocity. These differences can be tweaked out. But overall easily the best FREE dx7 emulator and editor

I can chop them up into smaller pieces to upload

That would make sense. As far as I know the DX7 (and probably most or all of the other Yamaha FM implementations) uses a weird feedback thing that modulates the attack rate with the current envelope level. This makes the attack stages quasi-linear rather than time-reversed exponential (which sounds weird, like a sound played in reverse). It's not described in any patents, so it would be quite difficult to reverse engineer accurately.

That would be appreciated.

...You are a little bit crazyyy!

My theory was that the mantissa and exponent are driven by separate pins and the conversion is done by separate ic's. Moreover, the exponent requires a stable mantissa signal to do its thing.
So i figured there is bound to be more/ longer glitching than what one would expect from just the dac.

The idea was that longer glitches will lead to bigger errors in the buffers.

But again, all theory

Also, not saying that this would change the sound dramatically or even at all.

Oli Larkin ran into problems reverse engineering the Casio CZ envelopes as well. Similar symptoms- percussive sounds translate inconsistently. I'll point him at this thread, your discovery might help for the V2.0 update.

I recall a debate at SDIY list back in late 90ies CZ attack portion was a stretched out S curve, ie deliberately wrongly designed (product cost cutting), ie reverse sounding.

Here are the audio files from the video. If anybody is interested.



DX7-Audio_Files.zip - Google Drive


Jim

Thanks.
Can you comment on what i said earlier about the machinegun effect in the second part of "DX7 V AUDIO#05.wav" ?
Am i the only one who finds it so ugly, compared to all others ?

FM8 is severely muffled while a real DX7 MK1 dont sound like that, pristinely clean and noise free.
My DX7 MK1 is noisy and distorty. You seams yo have a fake DX7!

The 4pole DAC filters is at -3db point c:a 16.5khz for DX7 and TX802, with significant more roll off on the 802 first pole .

I totally noticed it. I'm a big fan of old school industrial music like DAF, Front 242 and other bands of the sort, but that just sounds bad. lol...

Dexed is a winner for me for converting my giant catalog of rare DX files. The Arturia interface is clean and nostalgic and that is useful as well. It sounds good but for Sys-Ex, its Dexed all the way.

The sample and holds should still take care of it. It's not really a fundamentally different problem from any other PCM DAC. They've all got a settling time and usually some kind of glitch component. But if it's competently designed it won't affect the output.

I'm sure he knows that already. It's just that the details are hard to work out.

You're on the right track, though. The CZ series does seem to borrow heavily from the DX7's design (although phase distortion isn't really anything like phase modulation).