Hello,
I’v got few question about using my new Zynthian box:
Is there anyone who managed to succesfully use audioinjector card with the box. While building mine I’v got one but I cannot make it working. So I used Hifiberry.
Is there any way to correct playing of many the most beautifull, thick patches in ZynAddSubFx. The sound breaking, there are glitches and even whole box is freezing. My box is built on RP3 but even overlocking not solving the problem.
How can I change the tempo or timing in ZynAddSubFx. For ex: slowing or speeding arpegios to synchronize with the tempo of the song?
Is there a way to use RP3 built in wifi to manage MOD UI or SSH?
Can you explain a little more about your problems while configuring the audioinjector soundcard?
I’ve an audioinjector too, but i hadn’t time to test it yet. I hope to do it in these days. It shouldn’t be too difficult, but i’ve to try.
No. ZynAddSubFX can be very CPU-hungry. Some presets are really heavy and you must play with them very softly. If you play too much notes, you will reach the limits very fast. (for instance, Angel Piano
There are 2 solutions for this problem:
wait for RBPi 4 or 5 (or use another computer board more powerful)
rewrite the ZynAddSubFX engine for supporting multi-core. Currently it only uses 1 core ;-(
Yep. As far as i know, the “arpeggio” presets in ZynAddSubFX have the time “hardcoded” into the preset, so it’s not easy to synchronize with an external tempo. You can modify it by editing the preset using the ZynAddSubFX native GUI (take a look to the wiki!). Anyway, if you find a good method to do it, please, keep me informed
Yes, but it’s not configured by default. I hope to include this in a future revision. By now, you should configure it by yourself. You will find a lot of information about it if you “google” a little bit …
Yes, some sounds are a really CPU-time-horror - even on a RPi3. I think there is actual no way to solve this problem. “Maybe” one solution can be using a RT-Kernel and trying to give as much CPU-time as possible to the plugin. But the real problem is: A RPi[23] has four CPU cores but I think a plugin can not use them all really good in parallel. This is due to the calculation of the sound: You press a key and the plugin hast to fill a buffer with sound data calculated out of your input data (pitch, velocity). More keys, more calculations. At the end all pressed keys have to be ready with calculation at the same time: the time all voices have to be added together into the buffer - the slowest calculation wins. Perhaps you want some delay and reverb? Ok, the calculation for this is done when the sum of all voices is calculated. This depends heavily on the used algorithms, but there is not much space for parallel calculations. So the only way is massive CPU power.
Next problem: ZynAddSubFX is really a great synth but it has so much algorithms for nearly everything. That’s why it sounds so great - and why it cost so much CPU time.
What to do? Trying to avoid sound with “everything on”. Try to disable the Reverb or the delay. Try to reduce your used voices. Try to reduce everything else what may use CPU-time in the background (e.g. WLAN ) - but on Zynthian there is not much space for this.
I can only say this for MOD-UI: (Currently) You can’t. Further versions of MOD-UI may have a global time from mod-host. But the plugin must support this. Some plugins have their own tempo - but they are not synced against each other (like stepseq - which may work in sync when thes host has a global tempo). There is much to do - and much will be done, but it takes time.
I think this can be enabled. But it will take CPU-time… I don’t know if WPA2-scambling is done in the kernel driver… if so it will really take CPU-time. Perhaps anyone has tested this?
Thank you for the answers and informations.
According to audioinjector card: I have thought that I can simply replace hifiberry with audioinjector, uncomment dtoverlay in config.txt and it shall work. But no. I think that there are some diferences in gpio connections and there is a conflict with pitft. I know nothing about gpio connection so I gave up and simply use hifiberry. But if anyone smarter than I resolve the audioinjector connection I will be very interested
As far as i know, there is no conflict between Audioinjector and PiTFT. I’ve mounted a unit and it works perfectly, althought i had to change a GPIO pin, that now it’s the official wiring. See this post:
I believe arpegios (or sequences) are most of times simulated by LPF filter lfo with a ramp down wave, present in instrument’s global parameters, with more or less decay for simulating continuous seq…
Won’t almost for sure not be synced automatically, but the frequency knob i believe could be mapped into a sound page for manual timing… could be useful to add simil-seq effect to practically all presets…
.
This is the global parameter page of ADDSynth instrument of patch “Sequence1” you can find inside “arpeggios” bank. The seq is created by the parameters marked with yellow square
UPDATE: some sounds like ARPEGGIO8 have a echo efx on part insert effects, LFO lpf filtering is perhaps only about sequences
I think there is actual no way to solve this problem. “Maybe” one solution can be using a RT-Kernel and trying to give as much CPU-time as possible to the plugin. But the real problem is: A RPi[23] has four CPU cores but I think a plugin can not use them all really good in parallel. This is due to the calculation of the sound: You press a key and the plugin hast to fill a buffer with sound data calculated out of your input data (pitch, velocity). More keys, more calculations. At the end all pressed keys have to be ready with calculation at the same time: the time all voices have to be added together into the buffer - the slowest calculation wins. Perhaps you want some delay and reverb? Ok, the calculation for this is done when the sum of all voices is calculated. This depends heavily on the used algorithms, but there is not much space for parallel calculations. So the only way is massive CPU power.
