Pushing polyphony!

  • Supposedly the virus TI2 can have up to 110 voices of polyphony. This is of course subject to the complexity of the patches used. It's not like I or anyone else will actually need all of these voices simultaneously but I wanted to see if the limit is really up there - maybe as a sort of test for the DSP chips and the general well-being of my synth!


    I did carry out the test - all 16 channels firing away a 6 chord simultaneously (so 16x6=96) using the init patch with no modifications whatsoever. Sure interesting enough, channels 1-4 produced 6 voices each while channel 5 and 6 barely produced 2 voices. Channels 7-12 produced no sound. Total voices = 4x6+2x2=28.


    Switching all 16 channels to osc balance all the way to play only osc1 produced no change, which verified my understanding that the virus has always both osc 1&2 on - it's just a volume fader.


    Switching now OSC1 to a sine wave ,naturally a simpler wave, produced again no change. I guess the reason is that it uses the same amount of memory as the init saw.


    Switching to a WT and turning off the LFOs, again same result.


    Carrying out the test live (no VC) I managed to have 6 non-init patches running at the same time, using 10 arpeggiated voices. If I remember correctly 4 patches had complexity 1/5 and the other 2 3/5, which I guess is similar to the 28 init voices as above. One of the 3/5 patches had arp mode=chord.


    From the manual: Patches using e.g. 1 or 2 simple oscillators, one or two mod routings and one or two effects can achieve upwards of around 64 voices. With the very simplest of patches you can achieve far more than this - for example, please check out ROM B-124 ‘Tutorial 5’, a nice, warm PWM pad with Delay and Chorus, capable of being played with more than 100 simultaneous voices.


    I loaded this preset, and the results were about the same, only change is that channels 5+ would randomly play either the full 6 chord, a single note or nothing at all. Definitely an increase -lets say 40 best case- but no way "more than 100". Even if we decide that each OSC is a voice, then here all voices are at the very best 80.


    Now the question - where are these 110 voices? Is there something wrong with my unit? I know that one dsp chip takes over even channels (2,4,6 etc) and the other dsp the odd channels (1,3,5 etc) so I know both dsps have the same power. Can anyone verify or comment?

  • Thanks for researching this. Very interesting. I wonder if Access meant the super high polyphony rates would be for single patches only, not multi's. (I may not have understood all this correctly though.)

  • tonstudio96, this is what I'm asking - whether it's an issue or not :) My TI2 is desktop.


    My main experiment was via my daw so yes, midi. The live experiment was naturally on multisingle mode.


    I think my question will be simplified if we learn the definition of a "voice", as per Access, first.

  • OK things are getting a bit weird now...


    Per the manual: · Unison Mode (Common Menu) – divides the voice count by the chosen value.

    I created an init patch with 8 voice unison. Let's call this preset "init8". With the advertised 110 voice polyphony I should be able to create a 13-14 note chord (14x8=112)


    Multi mode (live)
    I loaded init8 on multi mode. .I could actually pull off a 13-chord note (yeah I only have 10 fingers and a thick chin!) , so 13x8 =104 voices indeed.


    Multi-single mode (within daw)
    Playing init8 from my midi keyboard, again I could pull off a 13-note chord so again 13x8=104 voices. The weird thing comes via midi playback... Via midi playback, I can only do a 5-chord, so 40 voices, which is again as per my very first experiment.


    ... how is this possible? Now I guess my synth is fine, the limit is there but why does the synth behave so differently if I play the notes on my keyboard VS midi playback, both via the daw? Isn't the keyboard still sending midi messages?

  • How did you verify that those assumed 104 voices are indeed sounding all together?

    Bass Player and Synthesist.
    Virus TI2 Darkstar | Virus TI2 Desktop | Moog Sub 37 | Blofeld | Machinedrum | Monomachine | Analog Four | Digitone | MPC Live | NI Maschine+
    Mac OS X 10.14.6 (Mojave) | Cubase Pro 11.0 | Ableton Live 9.6 | Logic 10.4 | MainStage 3.4 | NI Komplete Ultimate 13 | RME Fireface UFX+

  • Hi Oli, using "init8" as per above: on the keyboard, I was pressing a single key and verified that sound is coming out. Then pressed on top a second key and verified that sound was being generated from both the first and the second key - simultaneously. I was doing that until I eventually reached the 13 keys - 104 voices sound. Additionally on 13 keys I was lifting a key randomly (so 12 keys) and the change was always reflecting the key removed.

  • But it could be that the Virus played those notes with, say 4 instead of 8 voices. You wouldn't have noticed the difference.... at least I'd probably not hear it with a 13 note chord.

    Bass Player and Synthesist.
    Virus TI2 Darkstar | Virus TI2 Desktop | Moog Sub 37 | Blofeld | Machinedrum | Monomachine | Analog Four | Digitone | MPC Live | NI Maschine+
    Mac OS X 10.14.6 (Mojave) | Cubase Pro 11.0 | Ableton Live 9.6 | Logic 10.4 | MainStage 3.4 | NI Komplete Ultimate 13 | RME Fireface UFX+

  • well... not really. I'd like to know whether SEQ Mode will provide the full voice count, and if not, why. One thing I can imagine is that in SEQ Mode, the two CPUs are more strictly bound to the MIDI channels in order to achieve a predictable voice stealing behavior.

    Bass Player and Synthesist.
    Virus TI2 Darkstar | Virus TI2 Desktop | Moog Sub 37 | Blofeld | Machinedrum | Monomachine | Analog Four | Digitone | MPC Live | NI Maschine+
    Mac OS X 10.14.6 (Mojave) | Cubase Pro 11.0 | Ableton Live 9.6 | Logic 10.4 | MainStage 3.4 | NI Komplete Ultimate 13 | RME Fireface UFX+

    Edited once, last by oliAtBass ().

  • You are probably right, but provided your concern, I have no idea how to reliably measure polyphony now. I guess I have to load other presets and compare between the different play modes. I'll revert :)


    p.s. it's somewhat funny and awesome at the same time that this 7 year old synth now is still not fully explored :)

  • Hello guys.
    In my case i created a simple saw wave patch with unison 8 voices.
    When i trigger 1 cord via midi and add the same cord two times, each one octave higher, the virus stops playing usually the lower notes, when i reach 24 voices. The funny thing is, when i delay all cords for one 16th i can hear all notes played. Same case when i push as many keys i can on the keyboard of my virus polar. I tried it in ableton 8 and cubase 8 pro. With unison 2 voices i can run usually more notes via midi. If this someone could explain to me?
    Thanks and have a great day

  • I'm bumping this thread, because I have the same un-resolved issue with polyphony count.
    I'm always using a TI2 /61 in sequencer mode. No USB connected, I prefer the old fashion way with midi din connectors.
    I use various akai MPCs as sequencer (MPC 1000/ 2500/Live).


    No matter the complexity of the sound, the virus always fails to deliver a simple 4 parts piece of music.
    2 monophonic arps, 1 mono lead sound and a simple 4 notes chord will usually put the Virus to its knees and produce crackles or voice stealing. No patch has more than 2 bars of "CPU ressource".


    All in all, the multitimbral mode is barely usable and far from what's advertised. I begin to wonder if there is something wrong with my unit, a dead DSP ? or maybe a DSP allocation error ? Don't tell me about the usual tricks (unison, turning off EFX, amp release etc.. etc...) I've tried them all.


    Psylence's experience is quite interesting, having a different polyphony count between single and multi mode points to an obvious software issue.

  • In my experience, the Virus seems to also have a problem when many sounds, especially with fast envelopes/LFOs/FM, are triggered at exactly the same time. It seems that although the CPUs could calcualte all the voices, they are not fast enough to handle the occurrence of all those transients at once. This would exlplain why you can play more voices manually than triggering from a DAW.


    Also, the complexity bars are only a rough indicator. And dynamic as well: Observe them as notes are played... all of a sudden a one bar patch can raise by one or two bars (!) depending on played polyphony or note repetition speed. I've ran into several situations, where changing a patch just slightly (type of osciallator, slower attack, faster release, mono instead of poly for a mono bass sound) allowed to play two more other sounds although the indicated complexity stayed the same.

    Bass Player and Synthesist.
    Virus TI2 Darkstar | Virus TI2 Desktop | Moog Sub 37 | Blofeld | Machinedrum | Monomachine | Analog Four | Digitone | MPC Live | NI Maschine+
    Mac OS X 10.14.6 (Mojave) | Cubase Pro 11.0 | Ableton Live 9.6 | Logic 10.4 | MainStage 3.4 | NI Komplete Ultimate 13 | RME Fireface UFX+

    Edited once, last by oliAtBass: Fix typo ().

  • Still thinking about the different polyphony counts between single and multi. I was wondering if by chance there was a difference between multimode and seq mode regarding global polyphony?


    The multimode has the ability to disable unused parts to spare some midi channels. It might also save some precious computing cycles.
    I reckon most DAW users are running the TI on seq mode. I do.


    Could the sequencer mode be some kind of CPU hog, with all 16 parts running at the same time and only a few actually producing some sound ?

  • Still thinking about the different polyphony counts between single and multi. I was wondering if by chance there was a difference between multimode and seq mode regarding global polyphony?


    The multimode has the ability to disable unused parts to spare some midi channels. It might also save some precious computing cycles.
    I reckon most DAW users are running the TI on seq mode. I do.


    Could the sequencer mode be some kind of CPU hog, with all 16 parts running at the same time and only a few actually producing some sound ?


    single mode = 2 DSPs for your sound


    multi mode = 1 DSP for even channels, 1 DSP for odd channels and some special trickery if you just use MIDI channel 1 which might improve the polyphony further. and you can adjust the priority of a particular part in multi edit. priority means the likeliness of a voice getting "stolen" by a new note. it doesn't mean that you automatically have more polyphony to your disposal. so, enabling it every part is a zero sum game.


    sequencer mode = 1 DSP for even channels, 1 DSP for odd channels. and you can also adjust the priority of a part. context click on "D" in the part bar


    so yes, in single mode you get more polyphony. but parts which don't produce sound in sequencer and multi mode do not consume calculating power.


    hth, marc

  • In my experience, the Virus seems to also have a problem when many sounds, especially with fast envelopes/LFOs/FM, are triggered at exactly the same time. It seems that although the CPUs could calcualte all the voices, they are is not fast enough to handle the occurrence of all those transients at once. This would exlplain why you can play more voices manually than triggering from a DAW.

    I found the difference between MIDI triggering in Studio One and in FL Studio when I trying to play the 7th chords with two hypersaw oscillators in 8x unison mode. Studio One skips most of the notes in the chords, while FL Studio plays them clearly. I think it's because of difference in their algorithms of latency compensation, but I hope that somebody knows how to solve this in S1 :)

  • I found the difference between MIDI triggering in Studio One and in FL Studio when I trying to play the 7th chords with two hypersaw oscillators in 8x unison mode. Studio One skips most of the notes in the chords, while FL Studio plays them clearly. I think it's because of difference in their algorithms of latency compensation, but I hope that somebody knows how to solve this in S1 :)


    If possible, please try the following experiement: Use your chords to play a sound with a very short transient not using the Virus but a sampler (plugin) in both Studio One and FL Studio and record that. Also worth trying is sending MIDI events from both DAWs to another recorder or to a MIDI monitor and check out their timestamps.


    If my assumption is correct, then Studio One is sending the note events at exactly the same time while FL Studio has a slight difference between the events.

    Bass Player and Synthesist.
    Virus TI2 Darkstar | Virus TI2 Desktop | Moog Sub 37 | Blofeld | Machinedrum | Monomachine | Analog Four | Digitone | MPC Live | NI Maschine+
    Mac OS X 10.14.6 (Mojave) | Cubase Pro 11.0 | Ableton Live 9.6 | Logic 10.4 | MainStage 3.4 | NI Komplete Ultimate 13 | RME Fireface UFX+

  • If possible, please try the following experiement: Use your chords to play a sound with a very short transient not using the Virus but a sampler (plugin) in both Studio One and FL Studio and record that. Also worth trying is sending MIDI events from both DAWs to another recorder or to a MIDI monitor and check out their timestamps.


    If my assumption is correct, then Studio One is sending the note events at exactly the same time while FL Studio has a slight difference between the events.

    So i made a very short click in sylenth and recorded a big chord with it in FL and then in Studio One. The waveforms are absolutely identical.


    UPD: I've changed the USB port and it stopped to play chords fully even in FL. Seems like it's more an USB issue.