I’m a new V5 user, so forgive me if I’m missing something obvious.
Using Midi DIN and quite a simple snapshot with 5 instruments, shown below with their midi channel:
2. Noise Mak3r
3. ZynAddSubFX
4. Dexed
5. Helm
6. Noise Mak3r
Midi it receives:
2. is playing some short bass notes (1/8)
3. a bass with long notes.
4. a melody (1/8 max)
5. a bit more busy melody (1/16th and 2 notes at a time)
6. 1/4th backing with 3-note chords
The problem:
When all instruments are playing midi (from my external sequencer), both midi channel 2 and 3 notes are not fully playing. Some notes are, some aren’t. And it’s very consistent what plays and what doesn’t. When I mute channel 6 (the chords) in my external sequencer, everything plays normal. So those 3 notes every 4 steps seems to be disrupting the 2 lowest midi channels for some reason, which is odd considering it’s not that complex.
There’s nothing (e.g. weird artifacts/noises/lag) that indicates performance issues are ongoing.
Can multiple instances of Noize Mak3r be the cause?
I checked the midi monitor via the web ui. The notes are received and look normal.
I was initially looking for a screen that shows realtime CPU or Audio Engine feedback, or shows highest polyphony use, but I couldn’t find such a feature.
I’m not using any USB peripherals.
I updated from stable to staging, but the problem persists.
Is the ring mod turned up at all in Noizemak3r? It has caused pretty bad CPU load spikes for me if the ring mod is above zero. Much worse than what you’re describing (entire system nonresponsive with the CPU load pinned at 100% for up to a minute) but could be related. As long as I keep the ring mod at zero it’s efficient and reliable, though.
There is a CPU or there was a CPU meter that was the other option from PPM display, but I don’t know if it’s been tidied away in all the Oram unification.
A casual search hasn’t found it, and turning off PPM meters as a power save, simply removes the displays.
The normal give away of engine issues is Xruns which are reported in the webconf UI log under Interface.
failing that: using the terminal available in Interface webconf menu, typing.
journalctl -fu zynthian
tells a similar story, but by then you are onto a command line and htop is possibly the way to go.
just type
htop
Actually from a casual experiment htop seems to run quite well in terminal…
Hope that should be enough to reproduce it.
It’s mostly noticeable on channel 3 with the long notes. Easy to solo that on the mixer and mute/unmute the midi channels externally to see how it influences it.
I don’t know much about the Hapax, can it save standard MIDI files? If it can, when you’re able to reproduce your problem, save the Hapax sequence that’s making it happen as a MIDI file and see if you get the same Zynthian behavior when you play that MIDI file from a computer or anythiing other than the Hapax that can play SMF. If the Zynthian behaves the same way in both cases, then you can at least rule out the Hapax hardware (but not necessarily its firmware) as a factor.
That’s a shame but it’s not really that high priority, jsut something that would be handy.
Recording the output of the Hapax means recording any anomalies in the actual signal that’s carrying the data, which shouldn’t matter but I’ve had enough real-world experiences where it HAS mattered that I never assume it isn’t a factor when something goes wrong anymore.
Lots of more likely stuff to rule out first though.
However in this case, I multi-track recorded the midi into Logic, which converted it into whatever format Logic uses internally. Then I tested if the problem is there (muting/unmuting to reproduce it). It is there, a bit different (1 more note of the bass actually gets through for some reason), but definitely there.
Then I exported to a midi file. Then I created a new Logic project and imported the file in a similar setup. Probably was also still there.
So in this case I’m fairly sure it should be reproducible.
And while I type this I realize that I’m using the Hapax as a Midi interface, so the Hapax might still be the source of the problem. I can’t find my standalone Midi interface at the moment. I’ll try to find it later!
But still, everything looks the way it should on the Midi monitor as far as I could tell. And if the Hapax data would be corrupt somehow, I wouldn’t be able to make any music with other machines (I do, a lot) nor would Logic be able to record it properly. So yeah. I doubt the Hapax is to blame.
This isn’t necessarily the case, but these are edge cases and it’s really unlikely that’s what you’re dealing with, especially after the tests you’ve done. Sounds like you’ve got it covered, I read the original post late last night, should have reread it before I answered. I wish I had some answers for you!
I wonder if it’s a different symptom of whatever causes stuck notes for me and some other people.
Good that you say that.
Stuck notes are also a thing with the above midi file/snapshot. Not during playback, but whenever I press stop something usually hangs (ALL NOTES OFF command fixes it)
This really does sound like it’s related to stuck notes, except in rv5’s case it’s note-on messages that are being missed by some engines instead of note-off.
Doesn’t mean it’s actually related but it’s definitely suspicious.
For me it’s probably fine without stuck. Sequenced by Torso T-1, on the snapshot provided above in the discussion.
Don’t look for any art in this, I just quickly tweaked something on T-1.
EDIT: I also recorded the midi file directly in Zynthian. It’s not exactly what I recorded to audio, because the timing of turning on the individual channels is different. test-5x-dexed.001.mid (5.3 KB)
Another thing I’m adding to my mental list of tests to do when I have time is to play one of the stuck note producing MIDI files that people have shared into the Zunthian, and record it with the Zynthian’s internal MIDI recorder and on a computer from its MIDI Thru port at the same time, so we can compare the two recordings and see if there are differences.
yeah, just switched to staging again, still happening
just recorded an example. you first hear the long bass notes on top of the house music beat.
after the bass drum roll transition, another layer is activated that plays a melody. right away most of the long bass notes disappear. its quite drastic but also weird not more people are having such issues
I won’t be able to do any more testing for a few weeks unfortunately, but if it isn’t figured out by then I’ll also see if I can reproduce it with those files.