pouët.net

drum sample analysis

category: general [glöplog]
Has anyone ever done any proper analysis of the 606/808/909 drum samples? For instance I'm reading that the 808 cowbell appears to be saw-waves at 555hz, 835hz,1370Hz and 1940 Hz.

I'm googling around for a nice spectrum analyzer at the moment for this purpose and wondering if anyone has done this work before.
added on the 2010-05-18 20:07:29 by sigflup sigflup
hmm, i've been busy recreating all the samples with synths. but it was more trail and error than anything else.
a long time ago (while searching some info about how sounds synthesis and want to create my own sound synth for a 1/4k) i found an article explaining how 808 works internally

nothing was sampled or stored in memory most sounds where created using really simple generators.

there is was also some electronic schema showing how it was done internally (high/low pass filter implemented using a simple capacitor) ill try to found it back.
added on the 2010-05-18 20:53:48 by Tigrou Tigrou
Quote:
nothing was sampled or stored in memory most sounds where created using really simple generators.

To be precise, the tr-808 is an analog device, so it's no surprise that nothing is sampled. The tr-909 is partly analog and partly samples.
Btw, a nice realtime spectrum analzyer is f.e. Voxengo SPAN.
Has anyone ever found an analysis of the 909 open hihat?

As for the initial question, I bet that tons of people have done this before, but that most of it is held back as a trade secret, it's notoriously hard to find this stuff on the net.

I'll try the cowbell (Although I thought it used frequency modulation?)
added on the 2010-05-18 21:01:51 by AGL AGL
ahh http://www.soundonsound.com/sos/allsynthsecrets.htm looks a little promising- at least it sheds a little bit of light, for example http://www.soundonsound.com/sos/Apr02/articles/synthsecrets0402.asp
added on the 2010-05-18 21:04:36 by sigflup sigflup
the development is going further and further and (further and).

you people should get OFF your computers. and make it a real development.

AND what AND?
added on the 2010-05-18 21:08:25 by gentleman gentleman
Quote:
I'll try the cowbell (Although I thought it used frequency modulation?)


Maybe it does- maybe there are side band frequencies- the layout 1370Hz and 1940hz for instance seems kinda suspect.
added on the 2010-05-18 21:09:17 by sigflup sigflup
maybe this can help

take look at page 5, 6, 7 and 14. they describe how most instruments (BD, SD, ...) are done.
added on the 2010-05-18 21:09:29 by Tigrou Tigrou
@sigflup: I used wavesurfer for sound analysis during a course in speech technology. It's free / open source.
added on the 2010-05-18 21:15:27 by mic mic
Quote:
take look at page 5, 6, 7 and 14. they describe how most instruments (BD, SD, ...) are done.


helloooo nurse!!

Thanks mic, I'm downloading that now
added on the 2010-05-18 21:19:14 by sigflup sigflup
Those SOS articles look great. I've never read the magazine before the sampling revolution, and had no idea they did a series like that..

On the other hand, Watch Vince Clark do it
added on the 2010-05-18 21:20:07 by AGL AGL
Less useful, but Tr-909 Service Manual. I had always assumed it was all synthesized :/
added on the 2010-05-18 21:33:50 by AGL AGL
I studied the circuit diagrams a couple of years ago.

In a nutshell all of the sound generators are in theory super simple. Roland took textbook sound genrators and stripped them down to the bare minimum (to make them cheap). Unfortunately this will not help you for a proper emulation because the cheaper the circuit, the more non-linear it is. And exactly the non-linearities make up the charm of the sounds.

Sure, you can analyze what is going on inside the oscillators and VCA's. You can model them in software, but they will sound flat compared to the real thing because in the real deal each stage influences other stages and you end up with a highly complex non-linear system.

If you want to get a taste of this money-saving-makes-the-magic-thing, and you have the chance to play around with a 808/909 plus a 303 connected via Roland-Sync cables do so!.
The machines are syncronized so they ought to be running in sync, right? Not so! The machines will run hard synced every other beat (whatever the master dictates), but each slave machine will groove and wail to it's internal clock inbetween the hard syncs. It's almost the same tempo, but never exact.

You have to to experience that. Big mojo will await you!
added on the 2010-05-18 21:56:19 by torus torus
Quote:

I'm googling around for a nice spectrum analyzer at the moment for this purpose and wondering if anyone has done this work before.


Problem is, drums samples tend to resist spectrum analysis as they are so short and percussive.

I suggest to analyze them with iZotope RX and a lot of overlapping, it's very precise and will give you hints. Even then just recreating energy in these areas probably won't give you what you want control-wise.

I guess most bass drum emulation uses a kind of physical model for the sound. D16 Nepheton 808 bass drum "sounds" like a physical model to me. (T-Osc from xoxos is another example, too bad it has a lot of aliasing).

I don't think oscillators, filters and sample-accurate envelopes are enough i'm afraid. A problem of layered sound is that you get little correlation between layers ; physical models give you something strongly cohesive with a natural decay.
added on the 2010-05-18 23:47:02 by ponce ponce
i actually mostly wonder how they made the 909 open hihat sample. is it from a real drumkit? it doesn't sound very reall..
added on the 2010-05-19 08:17:36 by skrebbel skrebbel
I personally don't see the logic in including a (apparently more expensive) sample-player in the drum machine just to play back a synthesized sound, so I guess it would be a real drum, but you are right it doesn't sound like a 'proper' hi-hat at all.
added on the 2010-05-19 11:30:57 by ant1 ant1
Quote:
Has anyone ever done any proper analysis of the 606/808/909 drum samples?
Judging by the gazillion hardware or software 707/808/909 emulators and clones out there, I would have to say: yes.
added on the 2010-05-19 11:44:22 by gloom gloom
Skrebbel,

the 909 hats, crash and ride are samples on the rom.
added on the 2010-05-19 11:51:42 by rp^frstl rp^frstl
Ack, forget about it.

I once got close to the closed-hihat by using a short, filtered noise loop.
added on the 2010-05-19 11:53:49 by rp^frstl rp^frstl
rp, i know. everybody who ever heard multiple "decay" values for 909 cymbals knows (plain faster/slower sample replaying; you could play a melody with it)

what i wonder about is how that open hihat sample, on that rom, got to be. even if it's "just" a sample, and not synthetic, someone made the sample once. i find the fact that a huge percentage of dance music still uses that particular hihat sample, rather impressive. it's difficult to find a better open hat than the 909.
added on the 2010-05-19 13:19:35 by skrebbel skrebbel
i mean, was that japanese guy a genius? an awesome sound engineer? or did we just get so used to that particular sound between each kick that we like it just because it's so common?
added on the 2010-05-19 13:20:54 by skrebbel skrebbel
can't it be all tree?
Quote:
i mean, was that japanese guy a genius? an awesome sound engineer? or did we just get so used to that particular sound between each kick that we like it just because it's so common?


I'm personally convinced it's some sort of sonic crack. It's like a chemical addiction but with sound waves :P
added on the 2010-05-19 15:32:42 by sigflup sigflup
It was a happy mistake wasnt it? At the time of making the 909, the engineers were probably kicking themselves for making such a piece of crap. Cheap manufacturing, early sampling technology (6-bit or something?), the added complexity of providing timbre & envelope changes, and you get some convoluted noise that is luckily sonic crack a few decades later.

I'd like to see the process though, at 6 bits, I doubt it matters whether you're recording a hi-hat or two pans being hit together.
added on the 2010-05-19 19:29:59 by AGL AGL

login