Recently I have embarked upon a quest to try to reproduce some of those oh-so-cool FM sounds that my old Sega Genesis's ym2612 chip spurts out. Well, pretty much any yamaha fm chip. For the most part I have been pretty successful in rigging up some multi operator FM synths in PD using subz3ro's Adlib Tracker II instrument editor as a sort of guide. The only real problem I have had is implementing modulator feedback. From what I can tell, I'm supposed to be taking the output of the oscillator, multiplying it by a feedback factor, and adding it to the frequency. So I try something like:

|sig~|
| ____
| / |
|+~| |
| |
|osc~| |
| \ |
| |*~|<-|-----feedback factor
| |_____|
|
|
etc...

No sound... PD says something about a DSP loop and tilde objects and pretty much shits all over my hopes and dreams.

So instead of running the signal directly from the |osc~|, I send it to a |snapshot~| being driven by a |bang~|. This time I a at least get sound, but nothing at all like I want. I also tried using a |delwrite~| and |delread~| and experimented with different delay times, again without satisfactory results. Finally I used good ol' |send~| and |receive~|, but it still sounded wrong.

So... Am I going about this the wrong way (likely)? Is my understanding of feedback FM completely incorrect (very likely)? Did I implement it properly but am just totally off base about how it should sound (I hope not)?

Any help is very much appreciated!

what range of values are you using for your 'feedback factor'? it's not like audio feedback where you multiply by a value of 0 -> 1. your feedback factor will determine the frequency of the pitch, so it needs to have a big range: 1 -> 20000 or something like that.

to avoid the audio loop error, i guess delread~ / write~ should be ok. there are other ways, and you might have run into a couple of them in your experiments already.

also, and i'm not too sure about this, but your [osc] will output values between -1 and 1, so you might want to add one to the signal to get it all in a positive range, and then divide by two to get it 0-1:

[osc~]
|
[+~ 1]
|
[*~ 0.5]

etc

Using [s~] and [r~] will give you a one-block delay (usually 64 samples) - you could try reducing the block size using [block~] but it would be more CPU-heavy.

You can get some quite fun sounds with pairs of oscillators modulating each other (with or without extra delays/distortion/filters) - rather hard to control in a musical way, though, as you can hear/see here:

http://www.archive.org/search.php?query=ClaudiusMaximus+Soft+Rock

Thanks for the quick replies! Well, from your responses it seems that though I am creating FM feedback with s~ r~ delreads~ etc.. its not the type of feedback I want. Maybe? I guess I might not have been asking the right question.

I attached a zip file with an example of how I implemented feedback and an mp3 of how I want it to actually sound.

http://www.pdpatchrepo.info/hurleur/feedback.zip

that sounds to me a bit more like waveshaping than feedback.

maybe have a look at this patch i did a while back, and see if that can give you any ideas:

http://puredata.hurleur.com/sujet-648-casio-sounds-gets-emulated

For this kind of (saw bass) sound you need the ratio of Fc and Fm to ne 1:1 so you get a harmonic series and a modulation index of about 0.3

If you're applying feedback to get the brightness then you probably need an offset so the feedback is added to an existing modulation index base.

see attch

http://www.pdpatchrepo.info/hurleur/fmfb-11-saw-plus-offset.pd