Sysex feeling like a punishment

[nathanscribe]

So. I've got a 27 byte sysex configuration file that I can't find any way to send to the device from my old Mac running 10.9.5 (I know, I know)

I've tried Sysex Librarian (doesn't read the file properly, thinks it's only 3 bytes), Logic 9 (impossible to work out how to do literally anything with sysex in this frankly dreadful nightmare of an application), and Mixbus doesn't do sysex at all. The file itself is fine, it opens in a hex editor with no trouble.

Any other options? It's driving me up the wall.

[muzines]

Snoize!

https://www.snoize.com/sysexlibrarian/

Never had a single problem with it.

Can you upload the file to see it?

[nathanscribe]

Snoize!

https://www.snoize.com/sysexlibrarian/

Never had a single problem with it.

Can you upload the file to see it?

Sadly Desmond, this is one of my bugbears. Snoize has worked flawlessly with almost everything I've pointed it at/at it, but this? Nope. It thinks this 27 byte file is 3 bytes. No idea why. It only shows as 3 bytes in the list, and 'view contents' only shows 3 bytes. It's definitely not 3 bytes, it's 27.

[muzines]

Like I said, can you upload the file? I can check here...

[Eddy Deegan]

Snoize has worked flawlessly with almost everything I've pointed it at/at it, but this? Nope. It thinks this 27 byte file is 3 bytes. No idea why. It only shows as 3 bytes in the list, and 'view contents' only shows 3 bytes. It's definitely not 3 bytes, it's 27.

This sounds as if Snoize is having difficulty parsing it. Are you certain that the file is complete in and of itself and not merely the 'payload' to be used with a device-specific SysEx header?

Does the data in the file start with 0xF0 and end with 0xF7, and is the second byte in the data the correct ID for the manufacturer of the hardware?

[muzines]

Or maybe there's a >7-bit value in there which is breaking the sysex string.

All of which is why I was asking to look at the file, as it seems weird to me, I've never found Snoize anything other than totally reliable for anything I've ever thrown at it...

[nathanscribe]

This sounds as if Snoize is having difficulty parsing it. Are you certain that the file is complete in and of itself and not merely the 'payload' to be used with a device-specific SysEx header?

Does the data in the file start with 0xF0 and end with 0xF7, and is the second byte in the data the correct ID for the manufacturer of the hardware?

It does. It's for a DIY piece of gear so I wonder if Snoize expects something from a set of manufacturers. But I do know other people have used this file with no trouble.

[nathanscribe]

Here you go, just copied from the user guide, but I believe the file to be accurate. It's just a string of key assignments for a drum module, with a data byte (that 00 near the end) for sustain behaviour:

F0 00 90 90 23 24 25 26 28 27 27 29 2B 2A 2C 2D 2F 2E 2E 30 32 31 31 33 33 00 F7

[Eddy Deegan]

Hmm. A manufacturer ID of 0 is reserved for ID extensions, which are 3 bytes long.

That might be the problem. As a sanity check, if you change that second byte to 0x01, does it solve the issue of Snoize only seeing it as 3 bytes?

[nathanscribe]

As a sanity check, if you change that second byte to 0x01, does it solve the issue of Snoize only seeing it as 3 bytes?

Nope, but it now thinks the device is made by Sequential.

[Eddy Deegan]

OK, well it was worth a go (and Sequential's ID is 0x01, so that makes sense).

I see a couple of 90s in there. They are > 7-bit values. Are they decimal values? (In which case they should be specified as 5A). Valid SysEx only contains data bytes up to a value of 7F (127).

[nathanscribe]

I see a couple of 90s in there. They are > 7-bit values. Are they decimal values? (In which case they should be specified as 5A).

Ah, that is a very good point. Hang on a tick while I faff them around.

[muzines]

The two 90's are the problem, bigger than 7-bit values, as I suggested above.

(Verified in Snoize, change them to something under 128 (decimal) and the string works as expected).

Basically, Snoize is telling you that was a completely invalid sysex string and was terminating it at the invalid bytes with an F7 (entirely correctly as per the MIDI spec I think you'll find).

Like I said - always reliable!

[Eddy Deegan]

The two 90's are the problem, bigger than 7-bit values, as I suggested above.

+1

[nathanscribe]

Thanks chaps, Sysex Librarian now sees the full thing and will send it as desired.

The device isn't responding correctly though so that's another battle, and it may well end with fire and a kazoo.

I really hate MIDI.

[muzines]

Are you still battling with your note-off thing?

[Eddy Deegan]

You did change that ID byte back to 0 right? (just checking!)

[The Elf]

There's nothing wrong with MIDI! MIDI is our friend!

How about pointing us at the manual, and explain how this piece of gear is responding, to see if we can deduce what's going on?

Oh, and tell us what you're trying to do!

[nathanscribe]

OK so this device is an Introspectiv 9090, which is a clone of the TR-909 sound circuits developed by Colin Fraser and Trevor Page, and it's designed as a module with no onboard sequencer. The analogue sounds trigger at Note On commands as you'd expect, and Note Off has no effect on them, they just run to their natural decay as determined by the knobs. The sampled sounds, however (crash, hats, ride) are gated by Note Off, and by the use of Velocity 0 as transmitted by some devices in lieu of Note Off. Clearly this is not something we necessarily want, and certainly if you're sending it signals from another drum machine, it's simply not possible to adjust the note length, so all you get from the 9090 is a tiny 'click' as the sample starts and immediately ends.

In order to get these samples to run fully, this sysex file contains data to allocate sounds to MIDI notes, and a single byte to define sustain behaviour: when byte 26 is 0, sustain is gated off, but when it's non-zero, note off is ignored and the sustain runs fully.

Plenty of folk have reported no trouble at all getting this to work, but clearly I'm not among them. Thanks to the previous posts, I've got Sysex Librarian (Snoize) now seeing the full file, and I can edit the hex, so that's now not a problem. The problem is getting the device to recognise it.

Basic method is: hold the config button on the device while powering up, causing the MIDI LED to flash (yes, that's OK so far) – send the file – my interface flashes to show the file is being transmitted – but then nothing happens at the other end. MIDI is getting to the 9090, I can play the sounds, but it doesn't like the config file.

[zenguitar]

Have you tried sending it slower using the settings in SysEx Librarian?

Andy

[nathanscribe]

Are you still battling with your note-off thing?

Indeed, this be it.

[nathanscribe]

You did change that ID byte back to 0 right? (just checking!)

Yes indeed! I suspect the receiving device is primed to look for 90 90 though, is that a possibility?

[nathanscribe]

Have you tried sending it slower using the settings in SysEx Librarian?

Andy

I have, and to no avail!

[Eddy Deegan]

I suspect the receiving device is primed to look for 90 90 though, is that a possibility?

Not unless the device was engineered to break the MIDI standard. I guess that's not impossible but it seems like a very daft thing to do if it does. Is there anywhere you can verify what those two bytes should be?

[muzines]

Yes indeed! I suspect the receiving device is primed to look for 90 90 though, is that a possibility?

Not in a sysex string, no. Data bytes in a sysex string (actually, in MIDI, *any* MIDI) *can only be 7-bit*. Anything >127 is a Status byte, not a data byte.

So you either need $5A, or the message you have is wrong (I'd try things like $09 instead of typos, but it depends on the sysex documentation which is unknown), or the value *is* $90 but needs to be encoded over multiple 7-bit bytes in the sysex string.

Can you point us to the source of where you got this info? Because it seems suspect to me, or something's lost in translation (and clearly whatever you have set up isn't working, so therefore must be wrong)...