[PD] TR909 Synthesis

Roman Haefeli reduzierer at yahoo.de
Sat Apr 5 19:49:04 CEST 2008

On Sat, 2008-04-05 at 18:56 +0200, raul diaz wrote:
> Hi list!
> I would like to open a thread about drum synthesis
> I'm trying to make a tr909 emulator with an interface made with
> processing communicated with pd via OSC.
> I would like to make drum synthesis on pd based on chapters from
> fabulous sound-on-sound "synth secrets".

i once tried to emulate the 808 drumset in pd. however, due some stupid
copy-some-stuff-around-in-order-to-backup-mistake i lost most of the
instrumentes (basically only kick and clap are left).

a few words on the methods i used in order to model them:

- i downloaded a set of samples which different settings for each

- for sounds that consist of of several parts (e.g. snare: noise part
and two-oscillator part), i tried to isolate all the parts by
subtraction (e.g. subtract the snare sample with 0 'snappy' from the
sample with full 'snappy')

- i tried to figure out what kind of waveforms are used. 

- i tried ot figure out what kind of envelopes were used. those are
probably the easiest part, because you can measure the times by looking
at the signal graph (you don't necessarily need to hear them).

- i tried to figure out what kind of filters were used (e.g for the
snappy part in the snare). i found frequency analysis to be helpful in
order to guess the correct filtertype (order, low/hi/bandpass, etc.)

i didn't have a very close look at the 909 drums, but i could imagine
that those are more difficult to remodel. the diagrams are at least far
more complex than the ones of the 808, as far as i can see it. 

> I'm not an expert on drum synthesis, so I'm making this work in order
> to learn about that. I know there is really expert sound designers on
> this community and I would thank a lot their help.
> Now, I'm trying to emulate a bass drum. I've take this scheme as
> example:
> http://www.soundonsound.com/sos/Feb02/images/fig13tr909bd.l.gif
> But I don't know how important for the final sound is the block VCO
> (sawtooth wave) + Waveshaping, instead of a simple sine oscillator. In
> synth secrets articles says:
> "The output from the oscillator then passes through a waveshaper. This
> removes almost all the overtones, transforming the sawtooth into
> something very close to a sine wave."

the best would be to compare them. only the ears can tell, whether the
difference is remarkable or not.

however, i think that the waveshaper part is critical. the interesting
property of a waveshaper is, that the outcoming signal morphs its shape
depending on the amplitude of the incoming signal. if i remember the
kick of the 909 correctly, it sounds punchier at the beginning and gets
softer while fading out, which is most probably the result of using the
waveshaper. also when more than one oscillator is fed to the waveshaper,
the outcoming signal is quite complex (though, this is not the case
here, according to the diagram you posted).

> Another doubt is how accent has influence on amplitude envelope attack
> and decay, what's the range of pitch sweep (from 400 to 40-80 Hz?) and
> its envelope decay value (200msg?).

yo. not that i am an expert, but i would assume, that the smaller the
parts you're trying to remodel, the easier it will be. i think, that
finally only the ears really can judge if it's done right.

> And, finally I'm thinking about replaced noise generator + pulse
> generator with the trick used by roman haefeli on his 808bass
> (http://lists.puredata.info/pipermail/pd-list/2006-04/037663.html).
i remember, that i found a solution for the snare using simply noise and
a filter. it turned out, that i could use one of the iemlib filters
(cannot recall which, but it was sure a higher order one). the result
sounded quite similar to the original, though the scope didn't look that


Der frühe Vogel fängt den Wurm. Hier gelangen Sie zum neuen Yahoo! Mail: http://mail.yahoo.de

More information about the Pd-list mailing list