16 Aug
2004
16 Aug
'04
7:09 a.m.
Hi,
I'm trying to convert an existing analog (hardware) synth to an open source
softsynth.
One of the components of the analog synth is a diode wave shaper. The
schematic is included (diode.jpg). The Inputs 2, 3, ... n and its resistors
are optional.
I'd like to know if there's a LADSPA plugin (or even better: a DSP IIR
recurrence relation) which emulates the included schematic as close as
possible. Unfortunately, I don't have the EE background required to do this
myself.
Thanks in advance,
Stanley.
16 Aug
16 Aug
7:38 a.m.
On mån, 2004-08-16 at 10:54, Stanley Jaddoe wrote:
Hi,
I'm trying to convert an existing analog (hardware) synth to an open source
softsynth.
One of the components of the analog synth is a diode wave shaper. The
schematic is included (diode.jpg). The Inputs 2, 3, ... n and its resistors
are optional.
I'd like to know if there's a LADSPA plugin (or even better: a DSP IIR
recurrence relation) which emulates the included schematic as close as
possible. Unfortunately, I don't have the EE background required to do this
myself.
It's a function that converts from linear to S-shaped. It bounds
-infinity <--> +infinity to be within -1.0 <--> +1.0 with fairly linear
characteristics around zero.
Thanks in advance,
Stanley.
8:05 a.m.
On Mon, 16 Aug 2004 10:54:04 +0200
Stanley Jaddoe <linux-audio-dev(a)terabytemusic.cjb.net> wrote:
Hi,
I'm trying to convert an existing analog (hardware) synth to an open source
softsynth.
One of the components of the analog synth is a diode wave shaper. The
schematic is included (diode.jpg). The Inputs 2, 3, ... n and its resistors
are optional.
I'd like to know if there's a LADSPA plugin
Possibly, but I'm not too familiar with what is available in LADSPA.
(or even better: a DSP IIR recurrence relation)
Nope, not possible with a linear recurrence relationship.
which emulates the included schematic as close as
possible. Unfortunately, I don't have the EE background required to do this
myself.
Its actually rather simple (first year EE):
static inline double
diode_clip (double input, double clip_level /* should be positive */)
{ if (input > clip_level)
return clip_level ;
if (input < -clip_level)
return clip_level ;
return input ;
}
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nospam(a)mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"Having a firewall that allows NFS to the Internet is like having a
seat belt that lets your head touch the dashboard." -- Marcus Ranum
8:29 a.m.
On Mon, Aug 16, 2004 at 07:49:36PM +1000, Erik de Castro Lopo wrote:
Its actually rather simple (first year EE):
static inline double
diode_clip (double input, double clip_level /* should be positive */)
{ if (input > clip_level)
return clip_level ;
if (input < -clip_level)
return clip_level ;
return input ;
}
This is a bit too simple. The diode circuit does not perform hard
clipping when driven at moderate levels.
Something like
out = tanh (in) (or tanhf() if available)
or
out = in / sqrt (1 + in * in)
with some scaling of in and out will come very close
to the real behaviour. Alternatively change the '1' in
the sqrt to some other constant to set the normal
operation levels.
--
FA
9:01 a.m.
On Mon, 16 Aug 2004 12:13:51 +0200
Alfons Adriaensen <fons.adriaensen(a)alcatel.be> wrote:
This is a bit too simple. The diode circuit does not
perform hard
clipping when driven at moderate levels.
Something like
out = tanh (in) (or tanhf() if available)
or
out = in / sqrt (1 + in * in)
Both of those would probably sound better than what I posted,
but the original circuit had a 12k source impedance. As soon
as the the input overcomes the forward voltage drop of the
diodes the impedance of the diodes drops to the low hundreds
of ohms.
If you Spice the given circuit I think you'll find that the
output looks a lot more like what I posted than either of
the above.
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nospam(a)mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"Lumping configuration data, security data, kernel tuning parameters,
etc. into one monstrous fragile binary data structure is really dumb."
- David F. Skoll
10:04 a.m.
On Mon, Aug 16, 2004 at 08:45:49PM +1000, Erik de Castro Lopo wrote:
On Mon, 16 Aug 2004 12:13:51 +0200
Alfons Adriaensen <fons.adriaensen(a)alcatel.be> wrote:
The current in a silicon diode is an exponential function of
the voltage. Consequently, the small signal impedance is
inversely proportional to the current. A diode does not
have a fixed voltage drop, it still depends on the current.
This is a bit too simple. The diode circuit does
not perform hard
clipping when driven at moderate levels.
Something like
out = tanh (in) (or tanhf() if available)
or
out = in / sqrt (1 + in * in)
Both of those would probably sound better than what I posted,
but the original circuit had a 12k source impedance. As soon
as the the input overcomes the forward voltage drop of the
diodes the impedance of the diodes drops to the low hundreds
of ohms. If you Spice the given circuit I think you'll find
that the
output looks a lot more like what I posted than either of
the above.
The transition from the linear to the clipped region would
indeed be a bit sharper than with either of the functions
I gave. But it's still a function with a continuous
derivative. This makes all the difference for the spectrum
of a 'clipped' signal.
--
FA
10:55 a.m.
On mån, 2004-08-16 at 12:45, Erik de Castro Lopo wrote:
On Mon, 16 Aug 2004 12:13:51 +0200
Alfons Adriaensen <fons.adriaensen(a)alcatel.be> wrote:
Wouldn't the current be limited by the 12k resistor, so that the
impedance of the diode won't drop as fast as you suggest?
Anyways, here is a variation on the same theme, supposedly synthesizing
tape saturation:
http://home3.netcarrier.com/~lxh2/tapesat.html
http://home3.netcarrier.com/~lxh2/tapesat2.gif
/jens
This is a bit too simple. The diode circuit does
not perform hard
clipping when driven at moderate levels.
Something like
out = tanh (in) (or tanhf() if available)
or
out = in / sqrt (1 + in * in)
Both of those would probably sound better than what I posted,
but the original circuit had a 12k source impedance. As soon
as the the input overcomes the forward voltage drop of the
diodes the impedance of the diodes drops to the low hundreds
of ohms. 
If you Spice the given circuit I think you'll find that the
output looks a lot more like what I posted than either of
the above.
Erik
11:12 a.m.
On Mon, Aug 16, 2004 at 02:43:14PM +0200, Jens M Andreasen wrote:
Wouldn't the current be limited by the 12k
resistor, so that the
impedance of the diode won't drop as fast as you suggest?
OK, here are some real values, found with a hand calculator (real
engineers don't use Spice :-) View with gnuplot.
# Vin Vout
0 0
0.465 0.464
0.535 0.523
0.590 0.552
0.702 0.582
0.991 0.611
1.348 0.628
1.841 0.641
--
FA
1:05 p.m.
On Monday 16 August 2004 12:45, Erik de Castro Lopo wrote:
Both of those would probably sound better than what I
posted,
but the original circuit had a 12k source impedance. As soon
as the the input overcomes the forward voltage drop of the
diodes the impedance of the diodes drops to the low hundreds
of ohms.
The 12k resistor is just an example, it would be nice if it would be treated
as a variable.
How can I incorporate the 1N4148 diode characteristics in the final equation?
------------------------------
I did some SPICE emulation (Alfons: I know, I'm not a (real) engineer) and I
never saw a tangentfunction-like output...
Assuming one input and one output, (instead of the shown three inputs in my
first post), if the input signal is a triangle wave, I should get a (not
perfect) sine wave.
------------------------------
BTW: Does anyone on this list has any experience converting an analog hardware
synth to a softsynth using the synth's schematics?
Stanley
1:29 p.m.
On Mon, Aug 16, 2004 at 04:49:47PM +0200, Stanley Jaddoe wrote:
The 12k resistor is just an example, it would be nice
if it would be treated
as a variable.
The actual value doesn't matter, you'll get almost the same transfer shape
for any reasonable resistor value. Just make the input level variable, that will
determine how much effect you get.
How can I incorporate the 1N4148 diode characteristics
in the final equation?
All normal silicon diodes (and the 1N4148 is a very normal one) have the same
exponential voltage to current relationship. It is the combined effect of this
and the resistor that is approximated by the two formulae I posted earlier.
It's really not worth the trouble to try and emulate this circuit with any
precision. For example, a change in temperature of 9 degrees (Celcius)
will have the same (minimal) effect as halving or doubling the resistor value.
--
FA
1:41 p.m.
On Mon, Aug 16, 2004 at 04:49:47PM +0200, Stanley Jaddoe wrote:
I did some SPICE emulation (Alfons: I know, I'm
not a (real) engineer) and I
never saw a tangentfunction-like output...
The formula I posted does not contain a tangent, but an *hyperbolic* tangent.
Assuming one input and one output, (instead of the
shown three inputs in my
first post), if the input signal is a triangle wave, I should get a (not
perfect) sine wave.
It _that_ is the purpose of the circuit, there is an even simpler and better way:
Let x(t) be a triangle wave between -1 and +1, then
y = x * (2 - x) if x >= 0
y = x * (2 + x) otherwise
will give you a very decent sine wave.
--
FA
2:35 p.m.
On mån, 2004-08-16 at 16:49, Stanley Jaddoe wrote:
The 12k resistor is just an example, it would be nice
if it would be treated
as a variable.
How can I incorporate the 1N4148 diode characteristics in the final equation?
One interresting thing to consider is that the diode pair is mimicking a
valve preamp, possibly with some rude Marshall characteristics.
The exact same solution can be found in the upper right corner in this
schematics, just before the tone controls:
http://home3.netcarrier.com/~lxh2/marshprb.gif
(The 1N914 share characteristics with the 1N4148)
So what you should be looking for is a valve overdrive/distortion
plugin, rather than a 1N4148 emulator.
My guess is that the *exact* characteristic is not *that* important, as
long as it has a reasonably soft knee.
------------------------------
I did some SPICE emulation (Alfons: I know, I'm not a (real) engineer) and I
never saw a tangentfunction-like output...
Assuming one input and one output, (instead of the shown three inputs in my
first post), if the input signal is a triangle wave, I should get a (not
perfect) sine wave.
For some value of 'not perfect', yes :)) Crank up the volume and it
should transform into a square, apply a second channel with LFO and you
get pulse-width modulation.
Now please tell us, what is this synth of yours that you have
disassembled? Is this circuit part of the output-stage or is it part of
the individual oscillators? Inquiring minds wants to know!
/jens
------------------------------
BTW: Does anyone on this list has any experience converting an analog hardware
synth to a softsynth using the synth's schematics?
Stanley
3:25 p.m.
On Monday 16 August 2004 18:23, Jens M Andreasen wrote:
Now please tell us, what is this synth of yours that
you have
disassembled? Is this circuit part of the output-stage or is it part of
the individual oscillators? Inquiring minds wants to know!
It's the TR-909 drum machine.
For the Bass Drum part, the wave shaper converts the triangle output of a VCO
to a sine.
For the Rim Shot part, it's in the output stage, where three sine waves of
three different oscillator are mixed (that's why I had three inputs in the
schematic of my first post) and shaped.
For the Snare Drum, Low-, Mid- and High Toms, diode clippers are being used to
transform triangle waves to sine waves.
3:27 p.m.
On Monday 16 August 2004 19:09, Stanley Jaddoe wrote:
On Monday 16 August 2004 18:23, Jens M Andreasen
wrote:
The full schematics can be found here (for your reference):
http://www.kolumbus.fi/janne.husu/specs/909va.gif
http://www.kolumbus.fi/janne.husu/specs/909vb.gif
http://www.kolumbus.fi/janne.husu/specs/909vc.gif
http://www.kolumbus.fi/janne.husu/specs/909vd.gif
Now please tell us, what is this synth of yours
that you have
disassembled? Is this circuit part of the output-stage or is it part of
the individual oscillators? Inquiring minds wants to know!
It's the TR-909 drum machine.
7200
days inactive
7200
days old
linux-audio-dev@lists.linuxaudio.org
13 comments
4 participants
participants (4)
-
Alfons Adriaensen -
Erik de Castro Lopo -
Jens M Andreasen -
Stanley Jaddoe