diodes to ground have a hard ceiling.
tell me how to soften ,hard clipping.
how do you make the knee ...big...long...adjustable(my favorite word)
what can diodes do?@?
sorry. i get stuck on the little things.
diodes are complicated.
not simple as you would believe.
??? :D >:(
One very simple way is to put a resistor between the diodes and ground. As the resistor increases, the clipping softens.
thanks r.g.
i had thought that a variable resistance between a diode and ground acted to raise or lower the diodes threshold, but it acts as a sort of 'softness' or blend, and softens the hard clipping... interesting
it occurs to me that with a bunch of resistor/diode combinaitons you could have alot of control . like very minimal soft clipping for small signals
and added soft clipping for medium signals, and even more soft clipping for really big signals, and still retain the original signal dynamics...
the signal would never need to hit a hard ceiling and never really get squared... just rounded...or compressed?...
are there any circuit designs that do something like that ? i just don't recall seeing diodes set up in a row like that... set up to turn on one at a time depending on signal threshold...
Quote from: Quackzed on February 28, 2010, 09:53:32 AM
i had thought that a variable resistance between a diode and ground acted to raise or lower the diodes threshold, but it acts as a sort of 'softness' or blend, and softens the hard clipping... interesting
What it actually does is this:
A pair of clipping diodes to ground clip because they are (conceptually) open circuits to ground until the voltage across them gets to their conduction threshold, then they clamp the voltage to their conduction voltage. This is an oversimplification, of course, as the conduction of diodes is not abrupt, but has a knee, but go with it for a moment.
This means that diodes to ground are always used with some current limiting device in series from the signal source to the diodes. This is often a 10K resistor. The resistor limits the current that flows when the diodes conduct and their incremental resistance changes from almost an open circuit down to a few ohms. The signal is invariably taken between the junction of the two diodes and the limiting resistor.
If you add another resistor in series with the diodes, then any current that flow through the diodes also flows through the resistor, and that causes a voltage by ohm's law. This voltage is added to whatever the diodes are causing. This resistor voltage is a replica of the signal voltage, and is added back into the output signal by the resistor being in series with the diodes. The net effect is similar to some of the original signal being added back into the clipped signal. So the output becomes a sum of the clipped, diode waveform plus an addition of some amount of the part of the input signal that's bigger than the diode clipping threshold. This has the effect of making the diodes not clip as hard.
Another way to look at it is that the reason diodes can clip at all is that they are a voltage-variable resistor. Their resistance to tiny, tiny signals changes from very high, almost open circuit, if the DC voltage across the diode is near zero, to intermediate values as the diode starts conducting, to finally a quite small value, maybe 10 ohms, when the diode's forward voltage is pushed fully into conduction. Diodes can and have been used as voltage variable resistors. They're used this way in RF a lot, and some Vox solid state amps use them for a tremolo modulator. Anyway, inserting a resistor in series with them limits how low the diode+resistor resistance can go, no matter what the diode by itself is doing.
Usable values vary from zero ohms up to about the value of the current limiting resistor. At some point the diode clipping becomes hardly noticeable. The added resistor becomes a limit on how much the diodes can clip. The down side is that the output signal voltage grows as you increase the "softness" resistor, so it starts smallest and most distorted with the added resistor at zero, and the signal gets bigger and softer at the same time. This is very odd - your ears expect the opposite to happen.
Quoteit occurs to me that with a bunch of resistor/diode combinaitons you could have alot of control . like very minimal soft clipping for small signals
and added soft clipping for medium signals, and even more soft clipping for really big signals, and still retain the original signal dynamics...
the signal would never need to hit a hard ceiling and never really get squared... just rounded...or compressed?...
are there any circuit designs that do something like that ? i just don't recall seeing diodes set up in a row like that... set up to turn on one at a time depending on signal threshold...
It's been used. This is a simple form of a diode-clamp wave shaper, which was common in desiging waveform generators at one time. I used a series of diodes with resistors into nodes in the middle as a form of soft clipper. It works fine. You just need a bigger signal to work with, because the smallest step from diode to diode is the forward voltage of the diodes, 0.6V or so for silicon. This means for a soft clipper, you need on the order of 4-6 diode drops as the final clipper, and that means your signal needs to be 2-4V peak to have any noticeable soft clipping. But it does work, and you *did* just invent it, whether it's ever been used before or not.
This is great to read for explaining clipping, particularly hard v soft:
http://www.generalguitargadgets.com/richardo/distortion/index.html
Also, what kind of diode you use will affect the clipping. the order it generally goes (from hard to soft) is:
Op Amp
LED
Silicon Diode
Germanium Diode
Tube Diode
You can see the difference here:
(http://mixonline.com/mag/701TFfig_1.jpg)
I read somewhere that you can use dual-triode tubes like 12AX7's in place of 2 diodes. I bet you get some pretty soft clipping from that!
Thanks again r.g.
I really appreciate your willingness to share your understanding of electronics. It is such a great asset to have someone with a deeper understanding of this stuff to shed some light when the path gets dark .
QuoteBut it does work, and you *did* just invent it, whether it's ever been used before or not.
:icon_cool: i confess that I've always wanted to hear you say that. 8)
time to do some reading, and breadboarding...
Quote from: Quackzed on February 28, 2010, 09:53:32 AM
it occurs to me that with a bunch of resistor/diode combinaitons you could have alot of control . like very minimal soft clipping for small signals
and added soft clipping for medium signals, and even more soft clipping for really big signals, and still retain the original signal dynamics...
the signal would never need to hit a hard ceiling and never really get squared... just rounded...or compressed?...
are there any circuit designs that do something like that ? i just don't recall seeing diodes set up in a row like that... set up to turn on one at a time depending on signal threshold...
Check this out (http://falstad.com/circuit/e-sinediode.html) for a visual example. Also gives a neat optical illusion effect, looks like the resistors are swaying back and forth but they really are not :icon_biggrin:
Quote from: trad3mark on February 28, 2010, 12:00:17 PM
You can see the difference here:
(http://mixonline.com/mag/701TFfig_1.jpg)
Hmmm.... To me, that looks a little exaggerated, my experience with different types of diodes has never shown that amount of difference. (mainly looking at GE vs. Silicon/LEDs)
I suppose I'll have to check it out myself today.
It would be more effective to adjust the signal amplitude of each signal before it hits the diodes so the diodes all start conducting at the same phase of the sine wave, it would give a better view of the knee's effect on the signal.
AMZ has a huge amount of info on diode clipping, including the 'softness' control, which they call a warp control.
http://www.muzique.com/lab/warp.htm
http://www.muzique.com/lab/sat.htm
http://www.muzique.com/lab/sat2.htm
http://www.muzique.com/lab/zenmos.htm
http://www.muzique.com/lab/tclip.htm
QuoteCheck this out for a visual example
that's almost exactly what i was talking about . ;D
i did a little falstad sim of this. just the basic idea, but you can see where each diode begins conducting at a different threshold and how each is set up to softly clip the signal. as r.g. mentioned you'd probably want more stages in order to produce a decent amount of clipping and as a consequence, need a bigger signal swing to work with, but for a simple visual on the idea i thought this showed adequately what is going on...
if you copy and paste this into falsad /file/import you should be able to see it working...very similarly to the triangle to sine circuit...
Quote$ 1 5.0E-6 11.558428452718767 69 5.0 50
v 64 208 64 64 0 1 40.0 1.0 0.0 0.0 0.5
g 64 208 64 256 0
w 160 160 160 208 0
a 160 144 304 144 1 15.0 -15.0 1000000.0
w 304 144 304 208 0
r 160 208 304 208 0 25000.0
r 64 208 160 208 0 10000.0
w 64 64 160 64 0
w 160 64 160 128 0
O 384 144 448 144 0
w 336 144 352 144 0
w 352 144 384 144 0
r 304 144 336 144 0 10000.0
r 384 144 384 192 0 5000.0
w 352 192 336 192 0
w 384 192 400 192 0
w 336 256 352 256 0
w 384 256 400 256 0
w 384 256 352 256 0
d 336 192 336 256 1 0.4
d 352 256 352 192 1 0.4
r 336 192 336 144 0 5000.0
g 336 256 336 288 0
d 400 256 400 224 1 0.4
d 400 224 400 192 1 0.4
d 384 192 384 224 1 0.4
d 384 224 384 256 1 0.4
o 0 64 0 34 1.1692013098647223 9.765625E-55 0 -1
o 9 64 0 34 1.167910376109888 6.103515625E-55 1 -1
Quote from: JKowalski on February 28, 2010, 01:38:38 PM
Hmmm.... To me, that looks a little exaggerated, my experience with different types of diodes has never shown that amount of difference. (mainly looking at GE vs. Silicon/LEDs)
Nevermind, that picture is accurate. :icon_biggrin:
R.G.'s explaination of adding resistance to the clipping diodes not being what your ear expects, explains why this is not used more often. Adding a cap instead of a resistor to block the lows from clipping would appear to have a similar effect. Thanks one more time... :icon_cool:
I've been fooling around with this idea, but in order to have several stages of soft clipping,as r.g. mentioned, you would need a big enough signal swing to accomodate several diode drops...
QuoteYou just need a bigger signal to work with, because the smallest step from diode to diode is the forward voltage of the diodes, 0.6V or so for silicon. This means for a soft clipper, you need on the order of 4-6 diode drops as the final clipper, and that means your signal needs to be 2-4V peak to have any noticeable soft clipping.
but it occurred to me that this is only true if you have all the diode/resistor drops in a single stage. if you had a normal 1 volt signal, you could have enough swing for a diode soft clipper, with 2 anti-parallel diodes through a resistance . then you could re-amplify and use an identical diode soft clipper, but be able to adjust their 'relative' thresholds as compared to the signal swing.
or put simply, a 1 volt signal, softly clipped at .6v (barely clipped) then amplified x2 then softly clipped again at .6. (healthy amount of clipping)
so ,although the threshold of each diode stage is the same, they begin to affect the signal at different times, due to the relative difference in size of the signal.
so you could have many stages of gentle soft clipping and still be able to keep the signal within say an op amp's rails.
and though you cant change at what threshold a diode begins to conduct , you CAN alter the size of the signal relative to the diodes conduction voltage.4 of 1 or 2 + 2 of the other.
it also occurred to me that a 'soft clipper' could also be considered a 'noise floor amplifier'. ;D
with a single stage and 2leds and 2 si as diodes each through a 5k resistance, im not hearing distortion per-se, but rather a gentle limiting.
;Dthere is definately gold in them thar hills. ;D
Quote from: R.G. on February 28, 2010, 11:39:06 AM
It's been used. This is a simple form of a diode-clamp wave shaper, which was common in desiging waveform generators at one time.
Anybody know titles of old books that had coverd the theory of signal shapers ? This does not seems to be covered a lot in modern books...
Quote from: Quackzed on March 21, 2010, 11:18:37 AM
it also occurred to me that a 'soft clipper' could also be considered a 'noise floor amplifier'. ;D
Clever boy... :icon_biggrin:
Time to link to my favorite clipping thread. :icon_cool:
http://www.diystompboxes.com/smfforum/index.php?topic=38581.0
Quote from: trad3mark on February 28, 2010, 12:00:17 PM
This is great to read for explaining clipping, particularly hard v soft:
http://www.generalguitargadgets.com/richardo/distortion/index.html
Also, what kind of diode you use will affect the clipping. the order it generally goes (from hard to soft) is:
Op Amp
LED
Silicon Diode
Germanium Diode
Tube Diode
You can see the difference here:
(http://mixonline.com/mag/701TFfig_1.jpg)
I read somewhere that you can use dual-triode tubes like 12AX7's in place of 2 diodes. I bet you get some pretty soft clipping from that!
That's a great picture, it's explained something I never got, that the knees of the different diodes are fairly different, when scaled to be the same level as each other. I had assumed that people endlessly screwing with them was nothing more than the volume difference you get from the different turn on voltages, and it being impossible to A/B things at different volumes because louder almost always sounds better. Did you make that picture?
here's another good one that came up recently, visually showing the difference between inverting and non inverting opamp clipping stages:
http://www.diystompboxes.com/smfforum/index.php?topic=83077.0 (http://www.diystompboxes.com/smfforum/index.php?topic=83077.0)
Quote from: frank_p on March 21, 2010, 02:59:58 PM
Quote from: R.G. on February 28, 2010, 11:39:06 AM
It's been used. This is a simple form of a diode-clamp wave shaper, which was common in desiging waveform generators at one time.
Anybody know titles of old books that had coverd the theory of signal shapers ? This does not seems to be covered a lot in modern books...
There is a chapter in "microelectronic circuits" by Sedra and Smith that has some info on using diodes + opamps to shape signals, although IIRC it does not have anything as fancy as the triangle to sine conversion in the falstad applet listed in this thread.
The book is in its 6th edition now so you can find really cheap 3rd or 4th eds on line on alibris or other used text sites. Its a pretty good book in general as well. (i think the author was or is the dean of engineering at Waterloo).
thanks for the link, thats a good read.
QuoteTime to link to my favorite clipping thread.
http://www.diystompboxes.com/smfforum/index.php?topic=38581.0
another weird idea popped in my head after reading about mosfet clipping arrangements and seeing this quote-
QuoteAll orderings of parts in a series string have equal effects at the ends of the string. You may swap series components around at will with no effect outside the string.
...
can you make a transistor out of diodes? or diodes and resistors?!?
varistors and calcium sulfide photocells i'll have to check into as well.
does running say 2 ge diodes in series alter the knee as well as the threshold?
i need to set up a 'diode lab' :D
ot... that reminded me that i just watched this short of a really drunk kid telling the story of teslas life. weird but it was good.it's on hbo ;'on demand'; 'funny or die'
hi quackzed.. i have been designing a simple power amp simulator using a gain stage with an opamp (or a 4049 inverter) and a simple diode ladder..
the diode ladder is formed by 4 back to back pair of diodes in series to ground and some resistors and caps. The first pair lower end is tied to ground by a 2K resistor and a 33nF cap, and connected to the second pair, which is tied to ground through a 1K resistor..
the signal line before the diodes have a 220nF cap and a 10kpot in series, so:
- when the signal doesnt clip, it´has a fairly flat response.
- if it´s higher than 700mv, the first pair conduct, so the 2K resistor to ground and the pot make a voltage divider (e.g. if the pot is set to 2kohms, the divider is 2k-2k, then the gain is divided by 2, that´s -6dB, the higher the pot, the lower the gain--> more limiting)
- when the second pair kicks in , the voltage divider is even lower, so more limiting is applied.. the relation between the diode resistors defines the "compression rate".
- also, when the first pair starts conducting, a band- pass filter is made by the 220nF & 2K, and the pot resistance with the 33nF cap, this should give a "transformer saturation" type of response..the higher the saturation, the lower the bandwith...
- the pot controls the hardness of the limiter
Quote from: slotbot on March 22, 2010, 06:35:41 PM
Quote from: frank_p on March 21, 2010, 02:59:58 PM
Quote from: R.G. on February 28, 2010, 11:39:06 AM
It's been used. This is a simple form of a diode-clamp wave shaper, which was common in desiging waveform generators at one time.
Anybody know titles of old books that had coverd the theory of signal shapers ? This does not seems to be covered a lot in modern books...
There is a chapter in "microelectronic circuits" by Sedra and Smith that has some info on using diodes + opamps to shape signals, although IIRC it does not have anything as fancy as the triangle to sine conversion in the falstad applet listed in this thread.
The book is in its 6th edition now so you can find really cheap 3rd or 4th eds on line on alibris or other used text sites. Its a pretty good book in general as well. (i think the author was or is the dean of engineering at Waterloo).
Yeah, I have bought the fourth edition used at thriftbooks.com. It's in chapter 12. I was looking for more configurations... Thanks for pointing out anyways. Thats a good book.
like what sort of shape changes are you trying to achieve?
you could play with something like a mcu using wavetables to generate many shapes easily or use it to change shapes by taking an analog sample then having the value of the sample being the index of a lookup table that you make. Like mapping an amplitude at the input to a new amplitude on the output. Otherwise maybe an OTA would do something interesting since you can have square or logarithmic relations between the input and output?
diode ladder... ok. i get it.
Quoteif it´s higher than 700mv, the first pair conduct, so the 2K resistor to ground and the pot make a voltage divider (e.g. if the pot is set to 2kohms, the divider is 2k-2k, then the gain is divided by 2, that´s -6dB, the higher the pot, the lower the gain--> more limiting)
- when the second pair kicks in , the voltage divider is even lower, so more limiting is applied.. the relation between the diode resistors defines the "compression rate".
--so at each intersection of diodes is an opportunity to reduce or divide the signal and have an 'increasingly' soft-clipped signal? i've been using 2 diodes (.7) and a resistance then another 2 diodes (1.2) and a resistance, one after the other in parallel to ground, but essentially it seems i've missed that the diode ladder is basically a stepwise (voltage drop by voltage drop)..er ladder!
-so if a signal gets through the first pair, it then becomes twice as hard to get through the second pair.as well as reducing volume to the output .
o.k. diode ladder... say 3 antiparallel diode pairs in series to ground .each have a threshhold v of .6 and a signal thats say 4v so it can clip... 2 nodes ; at each node there is a resistor 1k to ground(no cap, for sake of simplicity)
+<=>-<=>-<=>-g
\ \_r__g
\_r______g
at .6v first pair begin to conduct, and untill the signal gets above the second pair's threshold the only active part of the ladder is this.
+<=>
\_r_______g
once the signals reaches 1.2v it looks like this
+<=>-<=>--r--g
\_r______g
is there anything actually different between that and doing it in parallel?
+<=>---r---g
+<=>-<=>---r---g
other than saving the cost of extra diodes?
Quote from: Quackzed on March 23, 2010, 08:32:38 AM
o.k. diode ladder... say 3 antiparallel diode pairs in series to ground .each have a threshhold v of .6 and a signal thats say 4v so it can clip... 2 nodes ; at each node there is a resistor 1k to ground(no cap, for sake of simplicity)
+<=>-<=>-<=>-g
\ \_r__g
\_r______g
at .6v first pair begin to conduct, and untill the signal gets above the second pair's threshold the only active part of the ladder is this.
+<=>
\_r_______g
once the signals reaches 1.2v it looks like this
+<=>-<=>--r--g
\_r______g
is there anything actually different between that and doing it in parallel?
+<=>---r---g
+<=>-<=>---r---g
other than saving the cost of extra diodes?
yups! that´s how it works..
is not the same to put it in parallel, unless you change the value of the r´s..
try simulating it on ltspice or livewire
Quote from: slotbot on March 23, 2010, 01:35:34 AM
like what sort of shape changes are you trying to achieve?
you could play with something like a mcu using wavetables to generate many shapes easily or use it to change shapes by taking an analog sample then having the value of the sample being the index of a lookup table that you make. Like mapping an amplitude at the input to a new amplitude on the output. Otherwise maybe an OTA would do something interesting since you can have square or logarithmic relations between the input and output?
Diode ladder configurations. As far as I can see there are not a lot of stuff on that topic in the Sedra. I don't know why but I have an impression that there are many types of ladders that can be done and that the theory had been discussed in older books. But I may be mistaken.
first, thanks dsshwartz for the insight into diode ladders as i had never really dug into them, being a bit alien to me and most of the circuits i am familiar with.
i did a flastad sim of the diode ladder setup as well as the parallel setup, but i didn't see any differences.
i tried using different limiting resistors to get a better visual reference for any differences, and so long as i kept the resistances the same as well as the diode conduction voltages, the 2 signals appeared to be equally effected.
at first i thought that having one pair of diodes passing signal for both diode/resistors to ground would be where the difference came in...
as opposed to the parallel setup, where the signal is split before the diode pairs/resistors to ground, but i couldn't find any difference.
i suspect that it's another 'black box' where you can swap around parts of a series string but the end result is the same.
it does seem that a diode ladder has all the advantages of less parts and more exact conduction voltage matching, without any drawbacks.
i DO feel like I'm learning something as well, a picture is worth a thousand words.. sadly i crashed falstad before i could save it to post as a reference... ??? :D :D :D I'll have to try ltspice some time! falstad is a bit light. but good for fast analysis
How do you determine resistor values for such a diode ladder? :icon_question:
basically you can think of it as a voltage divider...
so if you use a 10k resistor from the output, and a 10k between the diodes and ground, half of the signal above the diode threshold will be clipped...half will be unaffected, sort of a 50/50(10k/10k) clean/clipped blend...
--->----[ 10k ]-----,--->
/
---
^ v
-,-
l
10k
l
G
with a ladder setup, whatever is over the first diode pairs threshold will be clipped 50%(10k/10k)...as mentioned but...
whatever makes it over the second diode pairs threshold will be clipped an additional 50% ...
so below the first diodes threshold you have 100%signal
above the first diode pair threshold you get 50% signal
now from that 50% that is left,whatever part makes it over the 2nd diode pairs thresh will be half clipped again, leaving just 25% of the original signal above THAT threshold, and on and on...
--->----[ 10k ]-----,--->
/
---
^ v
-,-
/---[10k]---G
---
^ v
-,-
l-------[10k]---G
if we leave the top 10k but use 20k resistors after each diode pair then
instead of 50% clipped then another 50%clipped -10k/10k
we'd get 1/3rd clipped then another 1/3 of that 1/3rd clipped... -10k/20k like a voltage divider...
--i hope that makes sense, not an in depth explanation, but at least the basic idea of whats happenin'
thanks again to R.G. for taking the time to explain it to me, i hope i 'm not mucking it up ...
Sounds good to me. :icon_biggrin:
Thanks!
I'm fairly positive that I've seen a diode ladder in one of Fred Nachbaur's circuits but I can't seem to find it. I always wanted to try it. He called in a limiter circuit.
http://www.tubecad.com/2006/05/blog0066.htm (http://www.tubecad.com/2006/05/blog0066.htm)
It took me a little while to find this one again, I remember looking at it years ago and trying to wrap my head around the idea. The circuit is interesting to me - in this case how it uses a balanced signal instead of just dumping everything to ground. The principle is the same, though. Has such a nice graph illustrating the effect!
http://gilmore2.chem.northwestern.edu/projects/limiter_prj.htm (http://gilmore2.chem.northwestern.edu/projects/limiter_prj.htm)
This is another link, again from the perspective of Hi-Fi. Covers a few different schematics, too.
http://sound.westhost.com/articles/soft-clip.htm (http://sound.westhost.com/articles/soft-clip.htm)
What I like about this link is that it brings up the idea that I didn't initially "get"... that diodes have resistance that decreases as the applied voltage goes up. It has numbers, more colorful graphs, and again talks about the concept of limiting power via soft clipping (diodes) to protect the power amp/headphones/etc.
Saul t
Hi all
Resurrecting an old thread :D
Quote from: dschwartz on March 22, 2010, 08:32:50 PM
hi quackzed.. i have been designing a simple power amp simulator using a gain stage with an opamp (or a 4049 inverter) and a simple diode ladder..
the diode ladder is formed by 4 back to back pair of diodes in series to ground and some resistors and caps. The first pair lower end is tied to ground by a 2K resistor and a 33nF cap, and connected to the second pair, which is tied to ground through a 1K resistor..
the signal line before the diodes have a 220nF cap and a 10kpot in series, so:
- when the signal doesnt clip, it´has a fairly flat response.
- if it´s higher than 700mv, the first pair conduct, so the 2K resistor to ground and the pot make a voltage divider (e.g. if the pot is set to 2kohms, the divider is 2k-2k, then the gain is divided by 2, that´s -6dB, the higher the pot, the lower the gain--> more limiting)
- when the second pair kicks in , the voltage divider is even lower, so more limiting is applied.. the relation between the diode resistors defines the "compression rate".
- also, when the first pair starts conducting, a band- pass filter is made by the 220nF & 2K, and the pot resistance with the 33nF cap, this should give a "transformer saturation" type of response..the higher the saturation, the lower the bandwith...
- the pot controls the hardness of the limiter
Is such a diode ladder supposed to be in the inverting loop of the opamp or tied to ground after the opamp like as an hardclipper would be?
Regards, anthony
It depends a lot on the impedance driving the diode ladder and the impedance to be driven by the ladder.
The advantage of using a diode ladder in a feedback loop is that the input impedances are known quantities on each end of the ladder. Anything in the feedback loop of an opamp is effectively driven as though the output of the opamp was a current source and the inverting input ground/0 ohms.The opamp acts to drive the output voltage to make the current through the feedback impedances cancel any currents at the inverting input.
You can still do diode ladders to ground, but getting what you want for clipping up front gets trickier because whatever drives the ladder has to be capable of driving it, but not so capable that it can force the ladder voltage to be equal to the signal voltage. As a practical matter, that means putting some resistance between the driving source and the ladder to ground.