Bela Gliss not receiving clock

angstrom

Hi, Ive been trying all day to get the Gliss to show a yellow button flash on an incoming clock from an Intelijel Midi 1u with no joy.

The Gliss is in record mode () , with mode LEDs showing as per the video manual (green, red, yellow)

the clock output is definitely clocking, because it can successfully clock a range of other modules (clouds, morphagene etc) all of them showing clock indicators and keeping time.
but if I run a clock into the Bela Gliss there's no led indication and no response. If I arm the record (physical button, red LED) the recording doesnt work and the record arm cannot be unarmed (LED button stays red until i tap three times )

I can get the gliss to respond to a -5 / +5 LFO , it arms records and playsback in clock mode. clocking off the pulsewave LFO.

I tested the voltage ranges and I cant get the gliss to respond to triggers (or gates of any length) from 0 to 5v, and similarly ( as I said) its unresponsive to clock 0-5v

I tried setting the Gliss input voltage range to expect 0-5v ,
this does not help the situation though. Still no response to triggers, gates or clocks.

I tried trigger mode (orange mode) and the Gliss does not respond to triggers either

unplugging the gate/clock/trigger and setting to clock mode - the device will record as expected (free running)

Any suggestions?

angstrom

in addition I tried the "calibration mode" by pressing the top green selector for 2 seconds, the 4 orange LEDS started flashing around the touch strip. 15 minutes later, they are still going and its not reached the stage of "middle of the touch strip blinks orange"

It's just continuing to do the 4 LED thing

edit..

its now been calibrating for 1 hour. I think this unit is faulty

angstrom

OK, I flashed the firmware to the latest version and the problem persists
clock input or trigger / gate input (0-5 v ) doesnt do anything, no LED indication of clock, recording mode activation locks up the unit unless I do the three fast taps.

it will respond to any -5v to +5 v (zero crossing) voltage though, such as an oscillator of any kind.

giuliomoro

In signal mode can you get a signal input, pass it through and clip/scale it?

angstrom

giuliomoro

hmm, good idea.
The Signal mode works with an AC signal, such as the oscillator, or LFO.
I then tried the gate modification example , using the top output with Selector 2 on orange should pass the input gate or clock signal unaltered. Strangely it did not.

there was no output from the gate, and I couldnt see the gates represented on the faceplate (no led squarewave type VU response, or pulse type CV representation)

angstrom

I did the bad thing and pointed Claude at the codebase asking if there was any reason a clock or gate might fail where a -5v/+5v would succeed. The great water guzzlers opinions :

Yes. Record mode’s clock path is built around a ₂ V rising-edge detector on the raw −5 V…+10 V input, with hysteresis that only resets below ₁.25 V. A 0–5 V clock can sit above that reset level; a −5 V…+5 V clock cannot.

and I'll spare you the verbose diagnosis but it summed up with:

Why 0–5 V often fails here
Two firmware details make that likely.

Clock detect ignores calibration
Input calibration and voltage-range mapping happen after clock detection:

triggerInToClock(context);
// rescale analog inputs according to range
// TODO: don't do it if we are using this input for trig instead.
for(...)

context->analogIn[idx] = rescaleInput(inCal, context->analogIn[idx]);

The TODO is the tell: they know remapping is wrong for a trigger jack, and they also never run the trigger path through calibration. If 0 V on your module is a bit high in raw ADC units, 0 V can sit right on the hysteresis floor.

Hysteresis is only ₀.75 V below the trigger point If the “0 V” clock rest is really 1.5–2 V (very common with 0–5 V logic, mixers, buffered multiples, some sequencers):
first +5 V edge can fire
rest never drops below ₁.25 V
lastIn stays true
later clocks are ignored
A −5 V rest cannot do that.

Direct Control uses the same 2 V number, but only as a single-sample high check (tri.analogRead() > kTriggerInOnThreshold). Record/Clock also needs the falling reset, so it is stricter.

edit I tried several other gate / clock sources, such as my deluge Gate out. no luck

giuliomoro

I don't think ignoring calibration would break triggering: it's specifically ignored so that bad calibration wouldn't break triggering. How old is this module?

angstrom

giuliomoro

Its not very old, I got it earlier this year and it came with the V2 firmware, and when I flashed it I took it up to the latest (2.01?)

Ive used it a little bit this year but family stuff kept me from using it in anything other than unsynced "loop" mode . Though I got it for sync'ed modulation jamming.

this is the raw clock from the Intelijel uMidi , it seems fine to me and works on other modules. I kept thinking I must be doing something wrong - but the fact that the LFO triggers it (and triggers the button LED) and this clock doesnt makes me 90% confident its not a PEBKAK error

This LFO will trigger it

angstrom

OK, I'm not sure if this is expected or not but I noticed it was clocking on what seemed to be the negative portion of the LFO, (see img) but I think it may actually be when the voltage goes negative then its actually triggering on the little hump after the nega-spike

LFO Test 2

just so we all know Im not totally in the wrong mode 😉

giuliomoro

Can you try to set the second selector in RECORD mode to Loop (red) or Trigger (orange)? Then record a gesture and send a trigger input: the gesture should be triggered on positive edge crossing approx 2V. That should make it easier to verify at what point it is triggering exactly (especially if you can capture on scope or disk the trigger and Gliss's output at the same time).

angstrom

giuliomoro
heres a video of the module with the second selector in red mode "loop" and you can see the issue

bela-glis-clock-issue-small.mp4

9MB

hmm, that embed didnt work as I expected. heres a google photos link to a video
https://photos.app.goo.gl/eY5S7wFMo3ENLLwy5

giuliomoro

Thanks for confirming.
Is it possible that the trigger is too short ? I think the code can detect a clock as short as 5ms, or possibly even 1ms, but I wanted to check if you have an easy way of increasing the trigger duration.

TBH, the easiest explanation from what I see seems to be that the input is somehow not working and we are of course happy to replace the unit, I just want to make sure that it's no user or external error, or we'd both be disappointed if a new unit arrived to you and the issue persisted. I'll try to send you a firmware that passes through the input to the output so you can test this more in detail.

giuliomoro

@angstrom can you please update at https://bela.io/upgrade-gliss ? Download the file attached below, then at the firmware selection page, select the 'upload' option and select this .bin file instead of the firmwares listed there.

application-22c117ade5-dirty-ccdc08f146-dirty-80a64e5fc5-dirty-20b6fc4801-g491ke-trillrack.bin

192kB

With this fw installed, one of the outputs will be a copy of the input, so you can use it to test what the module is actually seeing at its input. It would be beneficial if you could then show a screenshot of what this looks like for various test signals, possibly showing the test signal at the same time.

angstrom

giuliomoro

thanks for this, I'll do that now. I've been a bit slowed down by my health (which is often a tricky beast) but I'll test this firmware

The Intelijel clock doesn't have a method of increasing the pulse length .

But I have a plan to test your theory of pulse trigger length by using Ableton's CV tools, and my MOTU Ultralite MK5 (which has DC coupled outs) . That way I can specify a clock pulse length, and also try some offsets.

The only hardware oscilloscope I have is one of those little kit ones which I've not used in ages, but I'll dig it out to test the clock pulse.
I'll be back with some results!

angstrom

giuliomoro

OK, here we go.

First some pics of how the Intelijel and Ableton (CV tools) clocks look, before we get to the meat of the vids showing how the test firmware responded.

INTELIJEL CLOCK LENGTH , seems to be about 4ms

ABLETONS CV TOOLS - clock length is at a minimum 16ms, which seems quite long. It can go up to 90% of a bar, which at 120bpm is around 1800 ms long !

And now, with the test firmware, I assume its a straight mirror of in to out, with no filter or anything. So I started with an audio signal

AUDIO SIGNAL https://photos.app.goo.gl/kSPXXsiK1fqjSKtW9

Then
INTELIJEL CLOCK (4ms)
https://photos.app.goo.gl/CjQohVziFK1VZuaF8

then
ABLETON CV TOOLS clock (16ms to 1000ms)
https://photos.app.goo.gl/NbDm7fL9bEnmLR9t8

giuliomoro

Out of curiosity, can you try the other output? Only one of the two has the loopback. I believe you are using the correct one, but it's worth double checking

angstrom

giuliomoro
Hi, sorry this took some time, I was working on a remix this week

here's the comparison between Input->output A , and Input -> output B, with the custom firmware
https://photos.app.goo.gl/vtMAmnKFk532tw4Z8

giuliomoro

ok well I think there's definitely something wrong there, then. Can you please email me at giulio at bela dot io with the order details?