1. IMPORTANT:
    We launched a new online community and this space is now closed. This community will be available as a read-only resources until further notice.
    JOIN US HERE

Introduction-youtube-featreqs-bugs

Discussion in 'FM8' started by CS_TBL, Apr 19, 2009.

  1. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    Boo!

    Being an FM8-zealot I thought it to be thoughtful to sign-up here.

    If there are FM8-youtube hunters out here: I'm the creator of these FM8 remixes:
    Hymne
    Chariots of Fire
    Crockett's Theme
    Second Rendez-Vous
    Adagio for Strings

    (note, from the first vids one can obviously click the next ones, but they may not all be playing at HQ, these links here contain &fmt=18)

    A small excerpt of an upcoming FM8 remix for youtube: Pipeline.

    And a little improvisation/test from a few days ago:
    Improved Adagio strings (the improvement is the addition of the rough attack to the strings, most evident in the double basses and the cellos)

    Anyway, as you could've read in the video and the descriptions, these are all exclusively done with FM8 and sometimes FM8fx if, for ex. in Second Rendez-Vous, it would save a bit of my CPU to have a global reverb.

    Anyway.. perhaps the wrong place, but here're my featreqs and bugs. ^^

    Obviously: I'm not the one to go for freakish sounds, trance/dance sounds, I almost only use the filter as an alternative and scalable/movable EQ etc. Instead I try to recreate acoustic instruments and otherwise synthsounds featuring human quality. What follows should be read with the above in mind. Which means: if you don't understand the core idea of my feature requests, only because you aren't into this acoustic stuff, then there's obviously no point in telling me that I'm wrong.. :p


    Features:
    • Add random delay for each note. Quite in the same manor as that analog slider on the master page, but then for delay. This would also have to be functional within Unison (so, an unison of 8 will feature 8 different delays)
    • Add a smooth RND LFO or otherwise a slew filter for each LFO (kind of how it's done in the Novation Supernova (2))
    • Add frequency envelopes per operator (like the FS1r has)
    • make sure these frequency envelopes *and* the global pitch envelope work with operators having a 0.00 ratio and offset for pitch.
    • A new 'ensemble tab' in the navigator sidebar. This ensemble tab would have like 16 (or something) 'slots' in which you can define a fixed "depth" value, a random "depth" value and a destination. The idea is that you can define how certain parameters change when playing in unison. If you have a unison of 5 and a fixed depth value of +40 with destination 'panning' then what you get are 5 voices spread out at -40 -20 0 20 40. This is what we have already in the Master menu. The random depth is similar, but random. a value of +40 would give us 5 random panning values between -40 and 40, different each time you press a note. Having a random range with volume (or carrier output) as destination gives you random volumes at each note. A random range with filter cutoff as destination give you random filter depths each note. These ensemble additions must be additive, not absolute. You could apply the panning ensemble tweak to a sound which is already panned a bit. The most important aspect is that for destination you should be able to choose *every* aspect of FM8. Everything, even operator ratios, envelope node x/y changes, LFO freqs, modmatrix entries etc. The destination column in that ensemble screen has thus an enormous list to choose from. What I want is define realistic ensembles where players all play individually. Each his own panning position, each with a slightly different LFO speed, each with slightly slightly different envelope attack rates, filter positions, etc.
    • Highlight the notes being played in the scaling section. It's so comfortable to see where I'm going to adjust the scaling. Right now it's a bit guessing work, not at least because often octaves presented in hosts/plugins mix up -2 and 0.
    • Add octave shift in the virtual keyboard
    • If you want my vids to be more clear regarding keys being pressed: use colors for pressed keys. :)
    • At the moment, not all controllers in the modmatrix behave the same. Some add their value to a destination, while others overwrite their value to a destination. Make both possible for a source controller.
    • Add keychanging. With keychanging I'm talking about the keychanging you find in large sample libraries, to switch articulations. Instead of always using conventional controllers, I'd like to dedicate a few keys of my choice to change envelopes, operators, modmatrix entries and such.
    • As mentioned elsewhere: operator copy, including the envelope.
    • Scaling copy
    • More detail in the modmatrix or logarithmic scale. Often, my vibratos are juggling between radical values like 6 and 7, while the higher values are nice for sirenes but I doubt anyone hears the difference between 48 and 49.
    • More LFO's. They are important, and as I use one for random fluctuations already I'm left with only one LFO for typical LFO tasks.
    • Add bit-reduction and resampling into the effects section.
    • Add an ADSR-envelope in the LFO's, now it's just off-delay-on, which is unrealistic in terms of human playing.
    • operator waveform phase setting, just like in the SY77, a 360 degree shift
    • The custom samples has been discussed elsewhere. As I also have an SY77 I already know that FM'ing samples results in complex and perfectly worthless kinds of noise. I don't think 'large' samples really have a place here, but custom small waveforms could be practical.
    • More cabinet presets, or a way to make our own. I rarely use them for guitars 'n such, often they're practical for other (acoustic) instruments.
    • I rarely use the talk wah effect. It's quite a bit too harsh sounding to be practical.
    • As mentioned elsewhere, a compressor in the effects section would be nice, even if it's just for FM8fx, a gate would be practical too, for drums 'n such. But only if it's placed after the reverb/echo/chorus.
    • Speaking of DSP locations, it would be very cool to reposition effects in the effects chain anyway.
    • Add some convenient way to set
    • MIDI learn: add pitchbend as controller
    • MIDI learn: add aftertouch as controller
    • In the waveform/spectrum display, add a button to show the spectrum/waveform for the current midi note. Right now it's probably all based around a C 4 or something. With scaling and such it would be nice to be able to see how a waveform changes according to the key I press.
    • More operators, I'd surely know how to use them.
    • An extra filter, [Y]?
    • Make the whole effects section an 'operator' (kinda like the [Z] already is). This gives you the opportunity to have operators being sent out directly (dry) and sent to the effects 'operator'. You can send the output of this effects operator to the final output and thus have one sound containing dry and wet sound. More interestingly, you can route the output of the effects section to a normal operator again. Basically this is a bit like routing FM8 to FM8fx, but then without the hassle of having to add an FM8fx instance. In a way, it'd be the best to have 2 such effects operators, this way a complex sound made by FM operators and one effect operator can eventually still undergo a conventional effects section. Put simple: this gives you the opportunity to have an FM-operator 'with a phaser' FM'ing another FM-operator, while the final result still can go through some reverb.

    Bugs:
    • Unison may work for basses, it won't work correctly for pads. It's not possible to really play nicely with slowly evolving pads. You'll hear envelopes being reset in an awful way.
    • Freshly saved presets don't show up in the preset list, only after browsing another directory and then going back into the relevant dir.
    • When going from unison to non-unison, all kinds of artifacts from the internal channels aren't being reset, panning positions and such. It can be cured by putting your arm on the keyboard, resetting everything in one go, but formally I'd call it a bug that has to go.
    • Have high ratios and lots of high frequencies played high up the keyboard, and chances are that FM8 is temporarily 'sick' of it. Sounds played lower will temporarily sound 'odd'. Will go away in time though.
    • In the modmatrix the distance to travel with the mouse (in pixels) is shorter than the amount values you can have. In short: not every value can be reached with the mouse. Please don't just extend it to 'just' match the amount of pixels 1:1, I'm on a HD screen here, a few pixels per value-change would be nice. It's a common problem that virtual instruments have an awfully small interface with ridiculously small knobs (and often enough they require radial mouse movement instead of linear mouse movement, I'll personally castrate each coder/designer who puts radial knobs into FM8/FM9), but this here is really a practical bug which is so simple to fix.
    • There's a bug with the LFO waveform display when hitting 'invert' the first time.
     
  2. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    Features:
    Add some convenient way to set tags. I've hundreds of untagged instruments. If I want tags I need to manually do that zillions of times + file > save, which is tiresome. Is there a way to speed this process up a bit? For instance by selecting a number of instruments and set tags for the whole group at once?
     
  3. Summa

    Summa Sounddesigner

    Messages:
    1,243
    Hi, if you add patch examples to your bug-list, I might be able to clear up a thing or two.
     
  4. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    The only thing that could be demo'ed was the fourth bug. I can't find an example of that one now (tried searching/tweaking for days), it's a bit unpredictable. As soon as I encounter that bug again, I'll drop it here. The rest of the bugs is easy to reproduce yourself.

    And here be space disco: http://www.youtube.com/watch?v=f-RfWtcyAX0&fmt=18
     
  5. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    Found something.. it's dead simple.

    Start with a new sound, a sine.

    Go to the mod matrix, to the [PB up]-[Pitch] junction. Change '2' to '100'. Test it by actually pitchbending upwards. Play a note while the pitchbend is up (you probably won't hear it anyway), let go of the pitchbend lever. Now hit new notes, and they'll be random'ish for a while.
     
  6. Summa

    Summa Sounddesigner

    Messages:
    1,243
    I rather meant examples of the more serious stuff, like your unisono problem, since in most of the cases it works fine for me, besides an issues with the saturation operators amplitude envelope, that somewhat isn't taking in account when it comes to voice stealing. I'd guess most of the other problems can be solved with proper envelope programming and using the right Unisono/Voice Settings. In case of problems I recommend to turn off dynamic voice allocation and to set the polyphony to a multiple of the unisono voice count.
     
  7. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    http://pub.unreal64.net/boor/ObviousUnisonoBug.ksd

    http://pub.unreal64.net/boor/unisonbug.mp3

    First played with no unison, after that the very same notes played on the same instrument, but with an unison of 2.

    In both cases, there was a maxed polyphony of 64. In the unison case dynamic alloc was on, but it made no difference compared to when it was switched off.

    Also, watch that mp3 in a sound editor, and you can see how the right and left channels differ!

    I've a bit of an idea what may cause all this:

    Hit a note, keep it down while hitting a new note, and everything is fine. But let go of the first note and its envelope is capped when you hit a new note. I think that in the first case it's allocating new channels internally, but in the latter case it just starts allocating again from voice channel #1 (out of 64). And that's where the previous note was just playing!

    But it's definitely a bug! ;P
     
  8. Summa

    Summa Sounddesigner

    Messages:
    1,243
    Well, you set release to more than 11s, no wonder the voices start to cut after a while ;)
     
  9. Summa

    Summa Sounddesigner

    Messages:
    1,243
    They should, due to the free running Operators the timbre changes according to the random phase setting.
     
  10. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    'After a while' is not the same as 'the next note'

    With a unison of 2, and a polyphony of 64 I expect to be able to get 32 notes without problems.
    ---
    It's completely unrelated to that, completely. Did you actually bother to look at the mp3 in a wave editor at all? Because I don't think so.
     
  11. Summa

    Summa Sounddesigner

    Messages:
    1,243
    The problem isn't occuring right away, when you hit explanationmark symbol, to turn all voices off, before you start playing. Still, I'd rather recommend to programm release stages that are audiable most of the time, it may helps to split the stage to keep the same slope.
     
  12. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    This 11sec example was a mere example to exaggerate the problem. The notes-off [!] seems to help, but soon enough the whole thing is borked again. It seems to go wrong once all the voices of the set polyphony have been used.

    Long envelopes are unrelated anyway. I just tested the same sound using 2 midichannels sending two identical notes to the same FM8 instance. It works perfectly. Two tones together, no polyphony cuts, no envelope bugs, but obviously no unison detune/pan-spread either, as it's not an internal unison, it's external.

    It's a bug, and imho it should be fixed.
     
  13. Summa

    Summa Sounddesigner

    Messages:
    1,243
    Well, I do unisono pads with the FM8 for years, believe me, it's caused by your careless envelope programming. It's not just the release it's your attack segment as well, the flexibility of the envelope generators can lead to strange problems. If you model you envelopes more carefully everything works fine...
     
  14. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    I've done envelope pads and any other instruments with FM8 for years as well. Have you clicked those youtube videos from my first post?

    As I explained one post ago, use 2 midi channels (with identical data) to control 1 FM8 instance, and everything is alright. Use 1 midichannel and a unison of 2 and problems start. Conclusion: it has nothing to do with envelopes.

    But ok, I suggest you show me a unison pad of your own, I'll test it and report about it.
     
  15. Summa

    Summa Sounddesigner

    Messages:
    1,243
    This not even proves that single notes will be handled differently for the sake of parallisation, since it's very likely your sequencer is serializing Data send to the same port/device.
     
  16. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
    Whether via unison or via a second midichannel, whether parallel or serial, in all cases FM8 needs to allocate new internal channels. And in case of unison it goes wrong after all the internal voices have been used once.

    It's a bug, it has nothing to do with whatever envelopes. I've given enough examples and hints as for where the bug can be found, that ought to do it.
     
  17. Summa

    Summa Sounddesigner

    Messages:
    1,243
    I admit that FM8 on unisono mode is running against a wall as soon as voice limit has been reached, but unisono pads work fine as long as you don't reach that point.
     
  18. CS_TBL

    CS_TBL Forum Member

    Messages:
    144
  19. Summa

    Summa Sounddesigner

    Messages:
    1,243
    I guess, but I'm only one of the sound guys, so how should I know ;)
     
  20. CS_TBL

    CS_TBL Forum Member

    Messages:
    144