FETs as diodes

[amz-fx]

Check out the Mosferatu the use of the Mosfet diodes

The Mosferatu arrangement is the same as that used on the Shaka 3/5 projects...  the diode steers the current flow through the integral body didoe of the mosfet formed by the substrate to drain junctions. The body diode is fairly slow switching and that may account for its softer sound.

regards, Jack

[R.G.]

I'm always trying to see how soft I can make distortion, which probably accounts for why the body diode on MOSFETs just sounds like any other silicon diode to me. After hearing substantially every silicon diode clipper ever made, I just keep thinking that there must be something more.

That's one reason I'm always interested in clipping knees - the fun stuff for distortion is all there. A new silicon diode - or for that matter germanium or LED - clipper is just one more angel on the head of the pin.

Been there, done that, what's next?

[ahermida]

Has anyone looked at what Fender is doing in a few of their amps?  I think this is what they call "Dyna-Touch" or something along these likes.  Here's a pic of a section of the Champion 30 solid state amp.




I think their creating a "piece-wise" approach with this one.

Alf

[amz-fx]

After hearing substantially every silicon diode clipper ever made, I just keep thinking that there must be something more.

I'm quite in agreement with you on that...

regards, Jack

[R.G.]

Has anyone looked at what Fender is doing in a few of their amps?

That's funny. I wondered when any commercial outfit would finally tumble to the idea of putting (ideally) good sounding limiting before a solid state amp to emulate a tube amp. I've been telling anyone that would listen about that ever since I figured out what Thomas Organ was doing with their solid state amps in the 60's. I stumbled onto it in the 80's, only 20 years late. That's better than I usually do.

Now all they have to do is figure out the rest of it. Given their track record, that may be a long time.

[ahermida]

I've been studying the Fender designs and if you look at the pic I posted in a previous message they come up with something interesting but  they go overboard.  The have a pair of clipping diodes right before this more complex clipping stage (CCS).  What do they expect is going to happen?  They already chopped off a lot of "information" from the signal before it reaches the CCS  I've seen designs where they have multiple clipping sections that really are more about distortion than trying to emulate anything even remotely close to a tube sound.

I think one of your comments also hits it in the head:  most people focus on distortion, sustain, gain etc rather than dynamics, response, clarity etc.  It all depends on who their "market" is wrt these solid-state amp models.  I've been working on designs aiming for dynamics, transparency etc for the longest time and when people check my stuff out some ask me right off the bat if I can add more gain to it!

Alf

[MR COFFEE]

If you want really soft limiting, try using cadmium sulfide photocells biased partially on (with light, natch) to around 4K or greater instead of diodes. You can't get much clipping with a 9volt supply, but if you get the signal swing up to around 10 volts or better the nonlinearity rounds things off very nicely and smoothly. Think Bassman pushed gently.
 
You can make the distortion dynamic as well if you modulate the light intensity with the guitar signal just a little. Think "give". Vactec photomods like the VTL5C2 do well in this application. FWIW

[R.G.]

If you want really soft limiting, try using cadmium sulfide photocells biased partially on

Or even softer, the old voltage variable resistors (varistors) like they used in the Magnatone amps for vibrato. You need upwards of 70V across them for a big change in conductance.

Really, really soft.

[A.S.P.]

information lost...

[Nasse]

Been thinkin this note on Headwize limiter page...

"One way to vary clipping levels is to apply a bias voltage to the normally grounded side of the diode array"

But what kind of bias should it be and how much does it alter sound...

[WGTP]

See if I got this right.  Modified Aron's drawing and barrowed parts from Joe Davidson's Obsidian schematic.  This should be the high threshold soft knee version that we have been discussing.

Let me know if I messed it up.

http://aronnelson.com/gallery/WGTP/MosfetDiode

[johngreene]

See if I got this right.  Modified Aron's drawing and barrowed parts from Joe Davidson's Obsidian schematic.  This should be the high threshold soft knee version that we have been discussing.

Let me know if I messed it up.

You have it correct. Here's a simulation plot showing what this does for you. Half way through the high frequency tone is inserted and you can see how it is visable even when the lower frequency is clipped.



--john

[WGTP]

Very Cool.

[lovekraft0]

  OK, now I'm confused - it looks to me like WGTP's drawings has the diodes oriented opposite to the direction shown in John's sim schemo. Somebody check this out and tell me if this drawing is the correct setup for soft clipping, please:


One more silly question - am I missing something, or does it not matter whether you connect the diode to the source or the drain, so long as the correct polarity is observed?

[R.G.]

OK, now I'm confused - it looks to me like WGTP's drawings has the diodes oriented opposite to the direction shown in John's sim schemo. Somebody check this out and tell me if this drawing is the correct setup for soft clipping, please:

Here's how to tell on your own: the diode must let the MOSFET conduct from drain to source as though it were in normal operation. It *is* in normal operation, it's just feeding itself an odd gate drive waveform.

One more silly question - am I missing something, or does it not matter whether you connect the diode to the source or the drain, so long as the correct polarity is observed?

And that's a fundamental principle that no one will believe until they do it for themselves: as long as you don't connect to or look at what happens in the middle of the series string, the order of components in a series string is completely arbitrary. All 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.

[lovekraft0]

...the diode must let the MOSFET conduct from drain to source as though it were in normal operation...

OK, so the "polarizing" diode has to be wired opposite to the polarity of the body diode - thanks, that makes perfect sense, now that you've broken it down.

[WGTP]

In retrospect, I think I tried this before and didn't think it was working because it wasn't distorting much and the signal got louder.

It appears that applications using this method with the goal of a soft knee would benefit from a rail to rail op amp to minimize the harder distortion produced by the op amp.  Of course, mixing the 2 varieties of clipping could be desireable.   Maybe an LED/Mosfet combo. 

[johngreene]

Not surprising. Using MOSFETs as diodes introduces a bunch of addtional variables that you need to consider.
1. The clipping threshold is higher, depending on the MOSFET, it can be much higher. (therefore louder).
2. The diode you use doesn't add much if anything to the clipping so Schottkys or Germanium will help lower the threshold.
3. Using a low gain MOSFET will give a softer clip than a high gain one.
4. You can get low threshold MOSFETs but they tend to be high gain and therefore clip harder sounding not much different than a Si diode.

Just to name a few.

--john

[WGTP]

John, I don't know much about Mosfets, where do the BS170/2N7000 and IRF510 fit into that?   

[R.G.]

There are neato datasheets that you can find just by googling with the part number and "datasheet".

For the BS170, the gate threshold voltage is listed as 0.8 to 3.0V, with a typical value (whatever that is...) of 2.1V. The gain is the forward transconductance, which is listed as 320mSiemen, or 0.32 amps per volt.

For the 2N7000 the gate threshold is the same, but the forward transconductance (gain) is listed as 80 minimum, 320 mS typical.

For the IRF510, the gate threshold is 1.0 to 4.0, the gain is 1.2S minimum, or about four times the typical BS170 or 2N7000.