Author |
Message |
françois
Joined: Dec 23, 2006 Posts: 139 Location: Paris (France)
Audio files: 1
|
|
Back to top
|
|
|
StephenGiles
Joined: Apr 17, 2006 Posts: 507 Location: England
|
Posted: Sat Dec 15, 2007 8:27 am Post subject:
|
|
|
Thanks, but |
|
Back to top
|
|
|
frijitz
Joined: May 04, 2007 Posts: 1734 Location: NM USA
Audio files: 54
|
Posted: Sat Dec 15, 2007 9:49 am Post subject:
|
|
|
I think an exponential final decay would usually be considered to sound better because that is how most natural processes decay. In fact, I often actually use a *truncated* decay when I'm making something like a chirp transient, since in that case the exponential decay already has too long of a tail.
Also, I don't recall anyone ever saying that the peak of the pulse of a traditional AD generator sounds bad. Usually you set the peak at a few percent below the asymptote and it sounds OK. I have an envelope generator that is actually an AR generator driven by a monostable pulse. This gives a pulse that is fatter at the top, somewhat along the lines of what you are suggesting, and it is sometimes useful for getting a more meaty sound.
To me, a bigger problem is the click produced by the discontinuous slope at the start of the envelope. A more s-shaped attack curve is what helps the most in getting a more natural sounding pulse -- at least to my ear.
But of course whatever sounds best to you is what you should use. Do you have any sound clips to demonstrate your results?
Ian |
|
Back to top
|
|
|
Tim Servo
Joined: Jul 16, 2006 Posts: 924 Location: Silicon Valley
Audio files: 11
|
Posted: Sat Dec 15, 2007 8:01 pm Post subject:
Transient Generator |
|
|
Hey Francois,
It's nice to see someone try a new approach to envelopes. I'm not sure if you've really nailed it yet, but your paper is an interesting read.
One thing that I think isn't really "problem" is this:
"Most other envelope generators use a capacitor that is loaded “exponentially” up to some level, and then a comparator triggers the decay segment. Up to what level ? certainly not 10V since exponential loading never (in principle) reaches the full voltage. So what ? 9V ? 9.5V ? 9.99V ? or something else ?"
I'm not 100% sure off the top of my head, but I would think a very easy way around this would be to simply charge the timing cap to something slightly over 10 volts, but then set the "end of attack" comparator to still flip states at 10V.
And yes, I also think some sound samples comparing this to normal expo envelopes could be illuminating
Tim (flipping states as we speak) Servo |
|
Back to top
|
|
|
yusynth
Joined: Nov 24, 2005 Posts: 1314 Location: France
|
Posted: Sun Dec 16, 2007 2:55 am Post subject:
|
|
|
I think that Bernie Hutchins published such a comparable design based on a state variable filter architecture. Check the Electronote site, this articleis in a freely downloadable Electronotes copy if I remember correctly.
http://electronotes.netfirms.com/free.html _________________ Yves |
|
Back to top
|
|
|
françois
Joined: Dec 23, 2006 Posts: 139 Location: Paris (France)
Audio files: 1
|
Posted: Sun Dec 16, 2007 4:10 am Post subject:
|
|
|
Thank you all for your comments. This is precisely what I was expecting.
First, my Transient Generator (TG) has a longer decay than a truly exponential would be. This is seen on the curve. I set the time constant at 1 second, so the decay phase starts at 1 second. After 4 seconds of decay the envelope is almost at 1V (about 916mV actually), while a trruly exponential would be at only 18mV. So this may seem "unnatural" from Ian's point of view. I personally don't care, because I'm not trying to emulate natural instruments, but that may be an issue.
Also I agree that the attack has a discontinuity in slope, which does not happen in natural phenomena. A more S-shaped attack could be better. But the curve would have at least two inflection points, and a simple second order circuit cannot do that (not easily, I mean).
Also, I never claimed that the AD peak sounds badly ! I've been using standard ADSRs happily for years. To speak frankly, my concern was to have an AD envelope that evolved naturally from a trigger pulse, with as little arbitray parameters as necessary. The function I suggest is a simple possibility. Now this is essentially a theoretical concern, and in practice standard ADSRs do a very good job.
I simulated the circuit with LTSpice. Since LTSpice can produce a .wav file I amplitude modulated a pulse wave through a simple VCA. Although the result is not exactly the same as with a classical (exponential) AD because of the TG's long decay tail, it seems the TG does not bring significant improvement - nor significant drawbacks. I'll rerun the simulation, maybe including a SVF this time, and post a demo sound.
Yves : Thank you again. I couldn't find the ciruit you mention at Electronotes, but there are plenty of pages, I'll check that later.
-- françois |
|
Back to top
|
|
|
yusynth
Joined: Nov 24, 2005 Posts: 1314 Location: France
|
Posted: Sun Dec 16, 2007 5:07 am Post subject:
|
|
|
Quote: | A more S-shaped attack could be better. But the curve would have at least two inflection points, and a simple second order circuit cannot do that (not easily, I mean). |
This can be simulated with a diode waveshaper but would cost a lot of components though... The sketch would be a linear integrator to generate a linear ramp (attack) then feeding that ramp with adequate offset and amplitude to a tri2sine diode (or OTA based) waveshaper and then offset and scale it to the right levels and amplitude... At least two or thee OPA would be necessary.
Cheers _________________ Yves |
|
Back to top
|
|
|
frijitz
Joined: May 04, 2007 Posts: 1734 Location: NM USA
Audio files: 54
|
Posted: Sun Dec 16, 2007 8:19 am Post subject:
|
|
|
yusson wrote: | This can be simulated with a diode waveshaper but would cost a lot of components though... |
At one time I made an EG with a little "foot" at the beginning. I can not remember how I did it -- some kind of simple diode shaper, I think. I found that I could get a faster attack with less click using it.
Ian |
|
Back to top
|
|
|
Scott Stites
Janitor
Joined: Dec 23, 2005 Posts: 4127 Location: Mount Hope, KS USA
Audio files: 96
|
Posted: Sun Dec 16, 2007 8:33 am Post subject:
|
|
|
Quote: | At one time I made an EG with a little "foot" at the beginning. I can not remember how I did it -- some kind of simple diode shaper, I think. I found that I could get a faster attack with less click using it. |
Rack you brain, man! I hate that click....
Seriously, sometimes the click is OK, but there are a lot of times I'd sure like to do without it, and I hate slowing down the attack to get rid of it. For me, that's one of the charms of Vactrol filters and VCAs - they're just too slow to be bothered by the click. But, it would be nice if it could be done easily with the non-Vactrol stuff, too. _________________ My Site |
|
Back to top
|
|
|
Ricko
Joined: Dec 25, 2007 Posts: 251 Location: Sydney, Australia
Audio files: 27
|
Posted: Tue Dec 25, 2007 7:09 pm Post subject:
|
|
|
One thing that can be quite nice is just to cross-mix the gate signal with the AR or AD generator. This gives a more definite start and a more reverb-y tail.
The trouble with conventional attack is that if you need to play the note ahead of time in order for it to sound on time (in organ terms, it is "slow to speak"). That makes playing by hand more difficult, and indeed horrible if you get MIDI involved and latencies.
The trouble with conventional decay is that it doesn't simulate reverb characteristics at all. If you have a long decay you lose note articulation, and if you have a short decay you lbring the sound right up into the foreground. |
|
Back to top
|
|
|
|