15 Jun
2013
15 Jun
'13
3:56 p.m.
Hi all,
In a tentative of giving more personality to the modular synth modules
of the avw.lv2 set, I have been looking at clipping.
I would like here to discuss here a few assumptions, the fruit of my
research and get a few feedback/comments to decide what direction to
take.
- in some cases (or let say modules of a synth), clipping is
implemented more to copie what an analogue system would do than a
mandatory part of the algorithm... Let's take an example: 2 sin waves
mixed together of amplitude -1/1 will just have an amplitude of -2/2
(as long as they are in phase)... A digital mixer without clipping
would be able to cope with that, but an analogue one wouldn't... and
that's why the analogue system would clip the signal......right?
- What method of clipping is used will give a "personality" to the
module: hard clipping, soft clipping, the method used for soft
clipping, etc...right?
- Hard clipping is something of the digital world - it doesn't exist
in the analogue world... right?
- Soft clipping will deform any waves of amplitude -1/1 even if it
doesn't exceed the accepted threshold, because just before reaching
the threshold the algorithm will take over and softly make the signal
reach the maximum amplitude and keep it there until the original
signal goes back under a set threshold.....right?
- Is there a preferred stage for clipping? In the case of a filter,
should we clip before filtering, after or both? Or are all these
options valid and that's what will give an additional personality to
the filter?
Thanks in advance for any comments!
Aurélien
15 Jun
15 Jun
6:22 p.m.
On Sat, Jun 15, 2013 at 02:56:13PM +0100, Aurélien Leblond wrote:
- in some cases (or let say modules of a synth),
clipping is
implemented more to copie what an analogue system would do than a
mandatory part of the algorithm... Let's take an example: 2 sin waves
mixed together of amplitude -1/1 will just have an amplitude of -2/2
(as long as they are in phase)... A digital mixer without clipping
would be able to cope with that, but an analogue one wouldn't... and
that's why the analogue system would clip the signal......right?
An analogue system with enough headroom wouldn't clip. A DAC will
clip in those circumstances.
- What method of clipping is used will give a
"personality" to the
module: hard clipping, soft clipping, the method used for soft
clipping, etc...right?
You'd also need to avoid generating harmonics above fs/2 to prevent
aliasing, unless that's part of the "personality" you're going for.
- Hard clipping is something of the digital world - it
doesn't exist
in the analogue world... right?
Wrong. An analogue amplifier with lots of negative feedback, eg. a
typical op amp or power amp, will usually hard clip.
- Soft clipping will deform any waves of amplitude
-1/1 even if it
doesn't exceed the accepted threshold, because just before reaching
the threshold the algorithm will take over and softly make the signal
reach the maximum amplitude and keep it there until the original
signal goes back under a set threshold.....right?
Yes, although clipping threshold can be any level you like (within
reason) in floating point.
- Is there a preferred stage for clipping? In the case
of a filter,
should we clip before filtering, after or both? Or are all these
options valid and that's what will give an additional personality to
the filter?
That depends on whether you want to modify the frequency response
before or after the clipping harmonics are added. Both can be valid
options.
John
8:50 p.m.
Le Sat, 15 Jun 2013 14:56:13 +0100,
Aurélien Leblond <blablack(a)gmail.com> a écrit :
Hi all,
In a tentative of giving more personality to the modular synth modules
of the avw.lv2 set, I have been looking at clipping.
I would like here to discuss here a few assumptions, the fruit of my
research and get a few feedback/comments to decide what direction to
take.
- in some cases (or let say modules of a synth), clipping is
implemented more to copie what an analogue system would do than a
mandatory part of the algorithm... Let's take an example: 2 sin waves
mixed together of amplitude -1/1 will just have an amplitude of -2/2
(as long as they are in phase)... A digital mixer without clipping
would be able to cope with that, but an analogue one wouldn't... and
that's why the analogue system would clip the signal......right?
- What method of clipping is used will give a "personality" to the
module: hard clipping, soft clipping, the method used for soft
clipping, etc...right?
- Hard clipping is something of the digital world - it doesn't exist
in the analogue world... right?
As already pointed out, an analogue system with enough headroom won't
clip the signal. It is also almost impossible to generalize what will
append when an analogue system go in clipping. This will depend on what
kind of hardware this is (class A preamp, class A, AB or B amp, ...), on
what kind of feedback it use ("normal" audio amplifier with a lot of
feedback, or power "operational amplifier" in open loop like a guitar
amp, ...), on the technology in use (bipolar, j-fet, tubes, ...), on
the actual implementation of the whole thing inclusive its supply, and
even on the quality or power of the components in some cases.
As example, when you push guitar amps in clipping at full volume, half
of the clipping you can ear is, with some brands, not the clipping of
the electronic, but the clipping of the power transformer. That sounds
very bad -:(, and that imply you may have to change often this
transformer when such amps are used to play blues or rock like styles
of music.
Even without going to such extreme cases, the actual implementation,
which is always a compromise between money and quality, will determine
the sound of any analogue system. And this is the most difficult part to
simulate -:)
Dominique
- Soft clipping will deform any waves of amplitude
-1/1 even if it
doesn't exceed the accepted threshold, because just before reaching
the threshold the algorithm will take over and softly make the signal
reach the maximum amplitude and keep it there until the original
signal goes back under a set threshold.....right?
- Is there a preferred stage for clipping? In the case of a filter,
should we clip before filtering, after or both? Or are all these
options valid and that's what will give an additional personality to
the filter?
Thanks in advance for any comments!
Aurélien
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev
--
"We have the heroes we deserve."
10:34 p.m.
On Sat, Jun 15, 2013 at 08:50:39PM +0200, Dominique Michel wrote:
As example, when you push guitar amps in clipping at
full volume, half
of the clipping you can ear is, with some brands, not the clipping of
the electronic, but the clipping of the power transformer. That sounds
very bad -:(, and that imply you may have to change often this
transformer when such amps are used to play blues or rock like styles
of music.
An output transformer will saturate if the frequency is low enough, but
the signal level required to saturate it is directly proportional to
frequency. In a properly designed guitar or bass amp there will be some
transformer distortion at the lowest frequencies but not much above that.
If you lowered the frequency enough to fully saturate the transformer it
wouldn't sound very good, as you say. (I design guitar amps among other
things).
John
11:24 p.m.
On Sat, Jun 15, 2013 at 09:48:40PM +0100, John Rigg wrote:
If you lowered the frequency enough to fully saturate
the transformer it
wouldn't sound very good, as you say.
And it could turn out be somewhat destructive... unless the rest
of the circuit is designed to handle this.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
11:33 p.m.
Le Sat, 15 Jun 2013 21:48:40 +0100,
John Rigg <ladev6(a)jrigg.co.uk> a écrit :
On Sat, Jun 15, 2013 at 08:50:39PM +0200, Dominique
Michel wrote:
Me too, and I repair them too. I was talking here about cheap power
transformers used in some brands of commercial guitar amplifiers, not
about their output transformers. The main frequency is low enough to
easily saturate them when they are not properly dimensioned, and this
saturation will go through everything to the speaker.
A typical example are the old Peavey Mace, good transistor preamp and
driver stage, 6x6L6 for the output, but a too small power transformer to
drive such a power (160 w RMS), and a bias circuit for the power stage
that kill the dynamic when it is in saturation. The power transformer is
definitely too small to drive the tubes at full saturated volume. I
measured such an amp, the maximum power is the same with a clean sound
and at full saturation. The sound is very good when the power stage is
not saturated, but very bad when the power stage is saturated, that
not only because of the lack of dynamic, but also because of the
saturation of the power transformer.
Excellent amplifier for jazz music - due to its large power, you can
even use it in small places with a mixer and put a whole gang on it -
but not for rock or blues. Even a small Fender Champ give a much better
sound at full volume. On the other hand, the output transformer of the
Champ is too small, and you will change it as often than the power
transformer of a Mace when playing loud, but the sound will be much
better. It was some years ago, one of my clients was "using" one
output transformer each year with his Champ.
Dominique
As example, when you push guitar amps in clipping
at full volume,
half of the clipping you can ear is, with some brands, not the
clipping of the electronic, but the clipping of the power
transformer. That sounds very bad -:(, and that imply you may have
to change often this transformer when such amps are used to play
blues or rock like styles of music.
An output transformer will saturate if the frequency is low enough,
but the signal level required to saturate it is directly proportional
to frequency. In a properly designed guitar or bass amp there will be
some transformer distortion at the lowest frequencies but not much
above that. If you lowered the frequency enough to fully saturate the
transformer it wouldn't sound very good, as you say. (I design guitar
amps among other things).
John
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev
--
"We have the heroes we deserve."
16 Jun
16 Jun
12:59 p.m.
On Sat, Jun 15, 2013 at 11:33:45PM +0200, Dominique Michel wrote:
Power transformer saturation only occurs if the voltage applied to the
primary is too high. It is not affected directly by the load on the
transformer.
An output
transformer will saturate if the frequency is low enough,
but the signal level required to saturate it is directly proportional
to frequency. In a properly designed guitar or bass amp there will be
some transformer distortion at the lowest frequencies but not much
above that. If you lowered the frequency enough to fully saturate the
transformer it wouldn't sound very good, as you say. (I design guitar
amps among other things).
Me too, and I repair them too. I was talking here about cheap power
transformers used in some brands of commercial guitar amplifiers, not
about their output transformers. The main frequency is low enough to
easily saturate them when they are not properly dimensioned, and this
saturation will go through everything to the speaker. A typical example are the old Peavey Mace, good
transistor preamp and
driver stage, 6x6L6 for the output, but a too small power transformer to
drive such a power (160 w RMS), and a bias circuit for the power stage
that kill the dynamic when it is in saturation. The power transformer is
definitely too small to drive the tubes at full saturated volume. I
measured such an amp, the maximum power is the same with a clean sound
and at full saturation. The sound is very good when the power stage is
not saturated, but very bad when the power stage is saturated, that
not only because of the lack of dynamic, but also because of the
saturation of the power transformer.
The effect you are describing is due to the internal resistance of the
transformer windings and other power supply components, not transformer
saturation. When more current is drawn the supply voltage drops due to
resistive losses. If there's a tube rectifier the effect will be more
pronounced. Some people like that effect but not me. I agree that power
transformers in many commercial designs are undersized.
John
9:15 p.m.
Le Sun, 16 Jun 2013 12:13:30 +0100,
John Rigg <ladev6(a)jrigg.co.uk> a écrit :
The Mace have silicon rectifiers. And the supply voltage drop on the
output transformer was at least of 100V, that between the full volume
point (160 w with a clean sinus, I know its hard...), and the fully
saturated point, when at the same time, the output power was almost the
same. At the output, the voltage was dropping with the saturation.
I don't think resistive losses alone can explain such a huge voltage
drop on the supply voltage.
Dominique
On Sat, Jun 15, 2013 at 11:33:45PM +0200, Dominique
Michel wrote:
Power transformer saturation only occurs if the voltage applied to the
primary is too high. It is not affected directly by the load on the
transformer.
This is true for a transformer that is not or normally loaded. But when
it is overloaded, we can get a behaviour similar to a coil: the
saturation is fixed by the current, not by the voltage, due to the
knee of the hysteresis curve. That imply the voltage drop will not only
be a function of the resistive losses, but also of the saturation of
the flux in the iron.
An output
transformer will saturate if the frequency is low
enough, but the signal level required to saturate it is directly
proportional to frequency. In a properly designed guitar or bass
amp there will be some transformer distortion at the lowest
frequencies but not much above that. If you lowered the frequency
enough to fully saturate the transformer it wouldn't sound very
good, as you say. (I design guitar amps among other things).
Me too, and I repair them too. I was talking here about cheap power
transformers used in some brands of commercial guitar amplifiers,
not about their output transformers. The main frequency is low
enough to easily saturate them when they are not properly
dimensioned, and this saturation will go through everything to the
speaker. A typical example are the old Peavey Mace, good
transistor preamp
and driver stage, 6x6L6 for the output, but a too small power
transformer to drive such a power (160 w RMS), and a bias circuit
for the power stage that kill the dynamic when it is in saturation.
The power transformer is definitely too small to drive the tubes at
full saturated volume. I measured such an amp, the maximum power is
the same with a clean sound and at full saturation. The sound is
very good when the power stage is not saturated, but very bad when
the power stage is saturated, that not only because of the lack of
dynamic, but also because of the saturation of the power
transformer.
The effect you are describing is due to the internal resistance of the
transformer windings and other power supply components, not
transformer saturation. When more current is drawn the supply voltage
drops due to resistive losses. If there's a tube rectifier the effect
will be more pronounced. Some people like that effect but not me. I
agree that power transformers in many commercial designs are
undersized.
John
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-dev
--
"We have the heroes we deserve."
12:56 p.m.
Hi Aurélien,
- What method of clipping is used will give a
"personality" to the
module: hard clipping, soft clipping, the method used for soft
clipping, etc...right?
from my experience, there are two main factors determining the
character of waveshaping distortion: one is the hardness of the
clipping transition, the other its +/- symmetry.
A perfectly symmetric clipping function will generate only odd-order
harmonics. Breaking the symmetry gives rise to even-order harmonics,
easily achieved for example through the simple addition of a DC offset
before clipping. At high gain, the degree of symmetry of the clipping
function becomes much more important to the "personality" of the sound
than the softness of the clipping transition.
- Soft clipping will deform any waves of amplitude -1/1
even if it
doesn't exceed the accepted threshold, because just before reaching
the threshold the algorithm will take over and softly make the signal
reach the maximum amplitude and keep it there until the original
signal goes back under a set threshold.....right?
You should also consider asymptotic functions, atan for example, that
can be used for clipping with a 'soft' ceiling, never reaching total
saturation. I prefer their sound, but the difference can be subtle.
- Is there a preferred stage for clipping? In the case
of a filter,
should we clip before filtering, after or both? Or are all these
options valid and that's what will give an additional personality to
the filter?
The heart of many distortion pedals is a clip|filter|clip chain, but
there are so many ways to vary or extend the scheme. For example, an
interesting approach to digital ladder filters wedges saturation
circuitry between the filter stages.
1:58 p.m.
On Sun, Jun 16, 2013 at 12:55:37PM +0200, Tim Goetze wrote:
The heart of many distortion pedals is a
clip|filter|clip chain, but
there are so many ways to vary or extend the scheme. For example, an
interesting approach to digital ladder filters wedges saturation
circuitry between the filter stages.
Or the (usually soft) clipping can inside the filter itself.
This happens e.g. in the 'Moog ladder' filter, where the signal
dependent emitter impedances of the transistor pairs act as the
resistors in a series of four simple RC filters.
The soft clipping that results follows a tanh() law.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
3971
days inactive
3972
days old
linux-audio-dev@lists.linuxaudio.org
9 comments
5 participants
participants (5)
-
Aurélien Leblond -
Dominique Michel -
Fons Adriaensen -
John Rigg -
Tim Goetze