Biasing of jfet with bipolar power supply

[D45T]

Hi everyone. I am trying to make a morning glory clone, and having problems with changing its schematic for bipolar power supply like in latest versions. Op amp section converts easily, but how to do input biasing of jfet? Where should that 1M resistor go to: ground or -9V? Thanks in advance.

[antonis]

Hi & Welcome.. 

You can modify your CS amp taking into account some bias principles:

For considering a mid-supply Drain bias (+4.5V) you need a current of about 200μA (9 - 4.5 = 22000 X 0.0002)
Same current flows through Source resistor so Source voltage sits on about 2.4V..
Gate sits on 0V (negligible voltage drop across bias resistor) so a V_GS of -2.4V results into 200μA channel current..
(minus (-) sign means Source voltage is higher than Drain one..)

For bipolar supply (+/- 9V) Drain should lie at about 0V (for equal output swing..)..
To obtain this, you simply need double Drain-Source current (voltage drop between +9V & Drain -> 9V hence 400μA current)..
400μA current frowing through 12K Source resistor set Source voltage to -4.2V (4.8V drop..)
You now have to determine  which V_GS value results into 400mΑ current (relationship between V_GS & Drain current unfortunately isn't linear) and set Gate bias appropiately..
(meaning, you'll probably need a voltage divider instead of single bias resistor..)

P.S.
Of course, you may leave R102 as it is (grounded), make R16 47K & R17 33K to maintain V_GS & Drain current as it and also make R18 27K for same stage Gain..)
(not recommended for reasons beyond present scope..)

[Gus]

https://www.diystompboxes.com/smfforum/index.php?topic=93817.msg807917#msg807917

[D45T]

(meaning, you'll probably need a voltage divider instead of single bias resistor..)

Thank you a lot for such a detailed reply!
I have built a model of jfet circuit in LTSpice, and found out that according to simulations using huge bias resistor between transistor gate and minus of power (ground, -9V, -12V, etc.) always seem to work and increasing this resistor only makes signal louder and reduces low end loss.
Green: standard Morning Glory, 1M resistor between gate and ground
Blue: +9V/-9V Morning Glory, 1M resistor between gate and -9V
Red: +9V/-9V Morning Glory, 10M resistor between gate and -9V

Soo... LTSpice gets it wrong, or putting huge resistor between gate and minus is enough?

[antonis]

HUGE resistor between Gate and -9V results into considerable voltage drop (previously negleted) so you actually bias Gate closer to (negative) Source voltage..
(the lower the  V_GS - V_Source difference the bigger the Drain current..)

BUT counting on bias resistor voltage drop is of high risk due to inpredictable Gate leakage current and also its multiplication with temperature rise..

[D45T]

HUGE resistor between Gate and -9V results into considerable voltage drop (previously negleted) so you actually bias Gate closer to (negative) Source voltage..
(the lower the  V_GS - V_Source difference the bigger the Drain current..)

BUT counting on bias resistor voltage drop is of high risk due to inpredictable Gate leakage current and also its multiplication with temperature rise..

So... the safest way to bias is to put resistor between gate and -9V, and to keep its resistance as low as possible? In LTSpice it is approximately 930k for +9V/-9V, decreasing it even lower will make the output lower than the output of original schematic.

[PRR]

WHY do you need bipolar supply? Bigger output? Just because you have it?

The single-JFET amplifier is naturally single-supply. The input can already be at-ground. The output can't be designed at-ground.

There is (as I think Gus is pointing at) the technique of putting gate and the source bypass cap to ground, the source DC resistor to V-. This swamps-out major variation in JFET Vgs and allows loose-sort JFETs to bias all the same.

[D45T]

WHY do you need bipolar supply? Bigger output? Just because you have it?

It is used in original morning glory to increase headroom of op amp before this jfet, and although there is no schematics of latest versions with charge pump available online, I thought that there is no sense in doing whole pedal 18V and the last stage 9V.

Since Xotic EP booster has identical jfet stage and it works fine with regular 18V power supply, it looks like just placing 1M resistor between gate and -9V is fine as well.

There is (as I think Gus is pointing at) the technique of putting gate and the source bypass cap to ground, the source DC resistor to V-

LTSpice simulation shows that there is no output if 1-10 M resistor is between gate and ground.

[D45T]

So, I've done some math and tested various combinations of values, and found out that as far as gain resistor (the one between source and capacitor to ground) is bigger than half of resistor between source and minus of power, bias of input/gate is not playing significant role and putting 1M resistor between gate and minus is fine. If we make gain resistor really small and enter the territory of extreme outputs, it is still fine and makes interesting asymmetric clipping. If someone wants to make transparent ultraclean jfet booster, there is schematic for proper biasing:

Green is clipping schematic on the left
Blue is clean schematic on the right
(crazy capacitors and voltages are there for testing purposes and can be reduced to whatever you have)

[antonis]

Ok.. just to settle things down..

Paul asked you the reason for using bipolar supply for a single stage JFET CS amp..
Your answer is for increasing head room which is logic enough but develops some practical issues.. 
e.g. how you can be sure for Source resistors bypass capacitors polarity..??

[D45T]

e.g. how you can be sure for Source resistors bypass capacitors polarity..??

Joyful bipolar electrolytic caps, approved by audiophiles!
Or sorrowful ceramic caps.

[antonis]

Money talks...

[PRR]

LTSpice simulation shows that there is no output if 1-10 M resistor is between gate and ground.

Gus is smarter than SPICE. Look again at his plan:



R4 is that "1-10 M resistor is between gate and ground". R3 is twice the size of R1 and forces good biasing. R2 plus JFET 1/Gm, relative to R1||load, sets audio gain.

Tip: the lower the DC current, the higher the audio gain, until the external load becomes a drag.

In guitar amp work, we do not need more than 1V-2V to smack the snot out of any guitar amp, so a 6V rail is marginal and a 9V rail seems ample. OVER driving a g-amp input with absurd large signal risks damage; on solid-state amps, perhaps a pain to repair.

[D45T]

we do not need more than 1V-2V to smack the snot out of any guitar amp, so a 6V rail is marginal and a 9V rail seems ample. OVER driving a g-amp input with absurd large signal risks damage; on solid-state amps, perhaps a pain to repair.

Power is not affecting gain directly. 2N5457 is not going to make more gain than some super hard mosfet at 9V no matter how much power you feed it.
And besides, there are plenty of boosters with insane voltage on the market — advertisements say that it is for "clarity and dynamics" (no idea what it should actually mean):
TC Integrated Preamp (it is officially recommended to use 24V or 30V power supply), Fortin 33, Fortin Grind, Free The Tone booster (I forgot the name, but all their analog pedals are 30V), etc...
Tomorrow I will take my breadboard and test all possible bias methods, starting with Gus schematic. If I don't report the results, it means that wrong polarity cap C1 has exploded and killed me.

[PRR]

....starting with Gus schematic. If I don't report the results, it means that wrong polarity cap C1 has exploded and killed me.

C1 polarity is correct as drawn.

But don't trust anybody. Set up without C1 and measure the voltage polarity where C1 goes.

(In an extreme case, R3 could pull Source negative a part-volt. But the JFET is about to stop working. And a part-volt won't blow-up an e-cap.)

[Gus]

Fender Harvard schematic note + - supplies in the preamp section
This kind of biasing has been around for some time
https://el34world.com/charts/Schematics/files/Fender/Fender_harvard_reverb.pdf

[PRR]

> Fender Harvard schematic

Yes, R3 R4 do put the source below common and needs its source cap the other direction. Many ways to skin rabbits.

> This kind of biasing has been around for some time

At least. It was old when Sherwin wrote this tip-sheet. (It says source-follower but mostly you can stick a load in the drain leg and take gain there.)