FET Overdrive with Diodes as Source "Resistors"

[mac]

I really do not know when or where I read about using diodes instead of resistors at a fet source, and if there is something similar. I liked the idea so I tried it:



It's voiced for my LC30 so you may want to change a cap here and there. It is intended as a med gain circuit, that's why the 100K at the gate of fets #2 and #3.
Hope you like it.

mac

[John Lyons]

Hmm.

I've heard about diodes as source/cathode resistors but not in fet circuits.
Curious how this sounds. Any way for you to post a short clip?

John

[mac]

Hmm.
I've heard about diodes as source/cathode resistors but not in fet circuits.
Curious how this sounds. Any way for you to post a short clip?

I do not have a good recording system but I'll try.

mac

[Doug_H]

We experimented with using LEDs in the source circuit of the minibooster in one of Aron's Shaka pedals. That was a long time ago. It sounded good but I don't remember why. I think Jack may have an article on using diodes this way.

[WGTP]

The one I remember seeing used an LED.   

[Steben]

It seems they bias to the treshold with a very low impedance. I think the bias current is a major factor in how these diodes react.
Been thinking about this too, but rather in BJT's to enlarge the base-emitter treshold.

[Dragonfly]

Ive used this in FF style circuits on Q1...sounds good.

[Steben]

Ive used this in FF style circuits on Q1...sounds good.

And what is the difference? Bigger headroom?

[R.G.]

No magic here. JFETs need a certain amount of reverse bias on the gate. You can get this in one of a few ways, but the way everyone classically does it is with a resistor in the source to ground. The JFET current running through this pulls the source up and with the gate pulled to ground, you get a reverse bias.

The problems with this are that the value of the source resistance also controls the gain of the FET to a certain extent. You can bypass the source to ground to get higher gain back with a capacitor. But the DC setting and the AC gain setting are mixed up in that one part.

Enter the LED. The LED is a diode. As you increase the voltage across it, it goes from non-conducting to conducting over a small voltage range. That is, it goes from a resistance of almost open circuit to perhaps 10-100 ohms over a small forward voltage range. You can use small signals and actually see the change in resistance on the diode.

Using an LED for JFET bias gives you a volt or two of reverse bias over a wide range of channel currents, but a small resistance to signal. That gives you higher gain without using a capcitor to ground on the source. It also gives you modulation of the gain by the signal for signals larger than about 25mv peak. The signal itself becomes a factor in changing the resistance, so the LED junction resistance changes the instantaneous gain. It's a form of asymmetrical gain, and hence asymmetrical distortion, as well as a bias modulating gain effect.

Again, no magic. You just have to know how to look at LEDs on more than one scale: bulk effect of forward voltage on bias (i.e. couple of volts of reverse bias on the JFET), bulk effect of the forward resistance on gain, and incremental effect of the forward resistance change on signals.

[db]

Biasing a JFET in this particular way is also adding another variable to an already highly variable situation (not that I'm suggesting that it shouldn't be done).  The main hard-to-predict setting is what ID the JFET will bias at.  If you know both the V-I characteristic of the diode and the transfer characteristic (ID vs. VGS) of the JFET (measured - not from the datasheet) then the point at which it will bias is where the two curves coincide i.e. when VGS=-VDIODE and ID=IDIODE.

If you were to take an individual J201 which, for arguments sake, stops conducting when VGS is less than approximately -0.6V (VGS(OFF)=-0.6V) and an LED with a typical volt drop at normal operating currents (in the order of a few mA) of say 1.2V then the combination will bias at a very low ID (probably too low to be able to set a reasonable value of RD to bias the drain at somewhere near mid-supply).  The reason for this is that the LED current at around 0.6V forward drop will be extremely low as will the JFET current when VGS is around -0.6V.  We could be talking nA!  In this case, it would make more sense to use a "normal" doide or perhaps even a Shottky diode to achieve a higher ID.

[MartyMart]

  The reason for this is that the LED current at around 0.6V forward drop will be extremely low as will the JFET current when VGS is around -0.6V.  We could be talking nA!  In this case, it would make more sense to use a "normal" doide or perhaps even a Shottky diode to achieve a higher ID.

So the original posters 4148 use is quite a good real world option then ?
Does this mean that when there's no constant AC signal passing through the fet ( when using a diode from source )
that setting a 1/2 V at the drain would be difficult/fluctuate ?

[db]

So the original posters 4148 use is quite a good real world option then ?

Well firstly, I wouldn't argue with a proven circuit whatever my theory says but it does seem to bear out if you look at the specs.  I've got the BF245A and 1N4148 datasheets in front of me.  The BF245A shows typically 2mA for VGS=-0.6V.  The 1N4148 doesn't start conducting properly until around 0.6V and it rises to about 0.8V at 25mA.  Again, take these values with a pinch of salt 'cause they're all given as typical.  So I would expect around about 2mA ish so the drain trim would be set typically at around 4.5V/2mA = 2.25K.  (Actually, looking at the note on mac's circuit he found that the BF245A's biased at 10K and 3K9).  With a 20K trim, it shouldn't be difficult to adjust for a given drain voltage but the gain of the stage wouldn't be very predictable.

Does this mean that when there's no constant AC signal passing through the fet ( when using a diode from source )
that setting a 1/2 V at the drain would be difficult/fluctuate ?

No, it shouldn't be difficult with the diodes shown but if you used a high forward voltage LED (say a green one with a drop of 2.5V) then biasing would become more difficult.  There shouldn't be much fluctuation (only that due to temperature which I don't think would have too much impact).

[mac]

Well firstly, I wouldn't argue with a proven circuit whatever my theory says but it does seem to bear out if you look at the specs.  I've got the BF245A and 1N4148 datasheets in front of me.  The BF245A shows typically 2mA for VGS=-0.6V.  The 1N4148 doesn't start conducting properly until around 0.6V and it rises to about 0.8V at 25mA.  Again, take these values with a pinch of salt 'cause they're all given as typical.  So I would expect around about 2mA ish so the drain trim would be set typically at around 4.5V/2mA = 2.25K.  (Actually, looking at the note on mac's circuit he found that the BF245A's biased at 10K and 3K9).  With a 20K trim, it shouldn't be difficult to adjust for a given drain voltage but the gain of the stage wouldn't be very predictable.

I have BF245A from different manufacturers. I used them because I can not get J201. The ones I used need a 10K to get 4.5V with 4148, but others from ¿fairchild? need lower resistors. MPF102 may need even lower values. But 2SK117 that have low Vgsoff goes from 50K to 100K with 4148. These are higher gain fets than BF245A. No matter what fets you have, I'll use 3 with similar bias.

RG,
Does the diodes smooth the source, or drain, current? Or as they are on, they just let pass the signal without changes?

mac

[db]

It won't "smooth" the drain current (which is the same as the source current give or take a few pA) as such, but the source voltage won't vary by very much and hence it doesn't provide much negative feedback and therefore higher voltage gain than you would get with an unbypassed source resistor.  If you look at a diode characteristic you will see that a large current change centred about the knee of the diode corresponds to a fairly small voltage change i.e. low dynamic resistance.

As there isn't much negative feedback in this circuit, you will get asymmetric distortion anyway courtesy of the FET transfer characteristic (mainly 2nd harmonic).  This will generally dwarf any effect of the diode.

I hope this helps.

[WGTP]

Silly question, but would this work along with the diode compression schem of Joe Davidson's for the Jfet?   

[mac]

db, thanks.

Silly question, but would this work along with the diode compression schem of Joe Davidson's for the Jfet?   

I guess that a diode in series with the 47K after the 1M, and some big resistor to vcc like JD's Vulcan may work:
a) 1M or 100K to gnd, |< diode, resistor to vcc, 47K, fet
b) 1M or 100K to gnd, |< diode, 47k, resistor to vcc, fet
or something more complex,
c) 1M or 100K to gnd, |< led, resistor to vcc, >| led, 1M or 100K to gnd, 47K, fet


mac