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

Feedback modulation (filter?) in FM8

Discussion in 'FM8' started by Alexandre Torres Porres, Nov 6, 2015.

  1. Alexandre Torres Porres

    Alexandre Torres Porres New Member

    Messages:
    4
    Hi there, I'm just trying to understand how the Feedback modulation works in FM8 and how it really is the same as in DX7.

    I'm actually also using SuperCollider, which provides a self modulating oscillator, but it just burst and blows up in noise way before the FM8 does with the same modulation index. Now, I know that in Supercollider there is no filter in the feedback, and I was researching around the web and found out about the DX7 patents where they described the usage of a mean filter in the feedback loop. Here's the patent http://www.google.com/patents/US4249447?dq=4249447&ei=7vK4T_m5LKOH6QHH2pTsCg

    It seems FM8 is supposed to be a perfect clone of the DX7, so I wonder if it has that same kind of filter in the feedback loop.

    My concern is that I'm teach music synthesis and sound design and I hope to give the correct information to my students on how Phase Modulation works and how the softwares I'm using works (be it SuperCollider, MAX/MSP and also FM8). I'm also talking about the DX7, but I don't have one. It'd be good for them to know in details if and how FM8 is actually a perfect clone of the DX7.

    Thanks
     
  2. Big Gnome

    Big Gnome NI Product Owner

    Messages:
    574
    There is definitely no filtering going on in the DX7's feedback loop (or elsewhere for that matter), as it predates the advent of real digital filters by several years (they didn't start showing up in commercial synths till the late 80's, and it seems to me if the technology were around when the DX line was being developed, it would have wound up in the direct audio path instead--witness Yamaha's later SY line). I'm having a little difficulty with the patchy English in the patent you posted, but it doesn't seem to indicate any such thing--check the block diagrams: it's all just memory and arithmetic representing a lookup table storing a sine wave which gets read through by both a phase accumulator and an arbitrary amount of the its own output. Filters are mentioned twice, once in describing the buildup of overtones as similar to the sound of a saw wave through a lowpass filter, and once which I'm having trouble making sense of, but it seems to me to be comparing the attenuation of the fundamental at high modulation indices to highpass filtering (?) Only NI could tell you for sure, but I strongly doubt that they've included any extraneous filtering in FM8. Try this: route Op F into Op Z, and then route Z back into F--you'll hear a very different tone than you would just feeding F back into itself, even without resonance and the filter(s) in their most unobtrusive settings.

    I couldn't tell you about SuperCollider, as I have only the most superficial experience with it. I imagine that feedback in SC and FM8 are scaled differently, or they use different ranges (e.g., 0-1 instead of 0-100).

    I can tell you that FM8 is not an exact replica of the DX7 per se--obviously, it's vastly more powerful than the hardware in too many ways to list; but the sound is very, very close if you don't deviate from the hardware's limitations. If anything, FM8 sounds "cleaner" than the DX7 (i.e., less noise and fewer digital artifacts), although the Digital parameter in the Master pane goes a long way towards this. I haven't A/B'ed a DX7 and FM8, but I owned one for several years, which I replaced with a DX5 (which had a noticeably more refined sound than the '7), which in turn I replaced with FM8 after about five years--the patches I imported from that were all but indistinguishable, once some erroneous mod wheel and aftertouch settings were corrected.
     
  3. Alexandre Torres Porres

    Alexandre Torres Porres New Member

    Messages:
    4
    thanks

    but here's the deal

    I was doing it wrong, I had the sample rate at 32Khz and didn't see it. When I changed it to 44.1Khz, voilá!!! The mean filter in the feedback loop made the oscillator sound pretty close to what FM8 does...

    So yeah, that patent reference hint might lead on to something :)

    Anyway, only NI can confirm it, but there is surely something in that feedback loop, cause nothing will make it blow up way sooner

    SuperCollider has a different range alright, I had taken that into account, it goes from 0-2pi whereas FM8 goes from 0-100

    Cheers
     
  4. Big Gnome

    Big Gnome NI Product Owner

    Messages:
    574
    That sounds to me like an issue in the design of the oscillator or phase modulation system with respect to sample rates; but like I said, I don't have much insight into how SC works.
    I'm still not sure where you're getting this notion of filtering taking place in the feedback loop--I'm quite certain that there's nothing more than summing/attenuation going on, and I haven't seen anything that persuades me otherwise.
     
  5. Alexandre Torres Porres

    Alexandre Torres Porres New Member

    Messages:
    4
    Samplerate affects the time delay of the single delay feedback, so at 32Khz the delay time of 1 sample was longer than in 44.1Khz.

    working with an even higher sample rate, say 96Khz, also changes the feedback time and alters the sound. So yeah, samplerate will quite much affect the way the feedback oscillator sounds.

    All I can say is that the SinOscFB object in SC has no filter in the feedback loop and it sounds way harsher than in FM8 with the same feedback amount.

    By applying a mean filter in the feedback loop as described in the patent I've posted, I can recreate the same behaviour as in FM8, so it seems they are doing something like that.

    I'm don't know which FM synths exactly used this filter in the feedback loop, but I've seen people mentioning about their usage in some of the yamaha synths, and there's this patent, and there's this experiment/comparison that I did. I can tell you that the raw signal is not fed back to the oscillator, if you implement such a thing yourself in Pure Data and Max MSP you can check this out. In SuperCollider you don't need to because ut already has this self modulating oscillator.

    I could make some sound files to show you, but you could turst me :)

    cheers
     
  6. tubeaudiogear

    tubeaudiogear NI Product Owner

    Messages:
    255
    Interesting, why dont you contact support and let them know that you teach music technology, maybe they are kind enough to give you some insight...
     
  7. Alexandre Torres Porres

    Alexandre Torres Porres New Member

    Messages:
    4
    I'm trying to reset my password so I cantalk to them since yesterday... site freezes and doesn't work :p