Mosfet Cathode Follower

[Sheldon]

Hello,
I'd like to know if such a mosfet cathode follower can work for the output buffer of a guitar preamp stompbox?
The input is taken from the volume pot



Thank you

[bool]

Sure it can work

But dou you really need 200 volts?

Imho mosfets (like BS170 and also IRF510/610) really like voltages from 12V - 24V. Even 24V is in most cases a total overkill for guitar pedals ...

Note that with lesser voltages, the source resistor needs to be scaled down. Around 4k7 for a BS170/2N7000 and 1k-2k2 for the IRF.

[samhay]

You can also scale down R21 and R22. You don't need a 5M input impedance and it will just add quite a lot of hiss.

[duck_arse]

10M//10M across 200V = 100V, yes? from the spec sheet for the zvn0545a, under "ABSOLUTE MAXIMUM RATINGS", is
Gate-Source Voltage - V_GS ± 20 V
I think the 100V would seal the fate of the mosfet. it sealed my fate instead.


[edit :] I went and hadda look at this:
http://www.geofex.com/Article_Folders/mosfet_folly/mosfetfolly.htm

[Transmogrifox]

Probably better to make R21 and R22 something like 500k, put a biggish > 1uF cap between them then T off a 1 Meg for bias to the MOSFET gate.

Current through resistors = noise.  If you can make the gate resistor into something that doesn't carry bias current it will help.  The cap at the bias T is to filter off the noise you don't want to amplify.

This also keeps other noise like 120 Hz hum to a minimum.

Otherwise I don't see a problem here.  

Heat might be a concern if you play very many outside shows in a hot climate where your pedal is exposed to direct sunlight.  OTOH this device will be boiling water off its package before it exceeds its operating temperature

@duck_arse:  It won't ever see VGS > +/- 20V in this application.  It's a source-follower so VGS will always be pretty near the turn-on threshold of the device.

VDG and VDS are the voltages we're concerned about, and this device is rated to 450V, so no problems with the 200V rails.

[Sheldon]

put a biggish > 1uF cap between them then T off a 1 Meg for bias to the MOSFET gate.

sorry, I'm not sure to understand this part, can you please explain me a little more?

[PRR]

OP has a related thread which gives background to his questions.

http://www.diystompboxes.com/smfforum/index.php?topic=110111.0

[R.G.]

Some comments on using MOSFETs as you've shown.

Should work, however, there are two very important parts missing. You need a 12V to 15V, value not important, zener from source (anode) to gate (cathode) to absorb transients that would kill that MOSFET gate. You also need a series resistor of 100R to 1K, value not terribly important, to damp the urge the MOSFET will have to oscillate at a few hundred MHz. These should both be connected to the gate as close to the physical gate as you can get them.

Presumably you're doing this because you already have 200V.

[Sheldon]

So if I understand correctly, it should be something like this :



-What value can I choose for C15 and C16?
-For R21 and R22 biasing, I saw values from 1M to 10M on the internet, is there a particular reason to go as high as 10M ?
-For R25, R.G. recommend anything from 100R to 1M
-For R23, I assume the value depends on the biasing resistors and the output impedance I wish, but what value can I choose? I added a R24 resistor cause I  feel I'll have to attenuates the signal by a factor of 4:1

thanks a lot

[R.G.]

So if I understand correctly, it should be something like this :

Not exactly. R24 is not needed, make it 0 ohms. The cathode of the zener goes directly to the junction if the gate/R25.

-What value can I choose for C15 and C16?

The value of C15 at the input will be C = 1/(2*pi*R21||R22 * F) where F is the lowest frequency you have to pass at -6db down from the midband. Big values of R21 and R22 make for smaller values of C15.

The value of C16 depends on the input impedance of whatever you'll drive. Since you may not know that, make it as big as you can reasonably afford to. Put a 1M or so from the outside of C16 to ground.

-For R21 and R22 biasing, I saw values from 1M to 10M on the internet, is there a particular reason to go as high as 10M ?

It depends on what is driving the input. You want the parallel combination of R21 and R22 to be greater than 10x whatever the source impedance of the signal source is. If a guitar drives it directly, use 1M or 2.2M. The higher the resistance, the bigger the noise this creates from thermal noise.

-For R25, R.G. recommend anything from 100R to 1M

100R to 1K, anywhere in there.

-For R23, I assume the value depends on the biasing resistors and the output impedance I wish, but what value can I choose? I added a R24 resistor cause I  feel I'll have to attenuates the signal by a factor of 4:1

Actually, no, the result does not depend on the biasing resistors. They're determined by power and output impedance considerations. If you want a 4:1 reduction in signal level, R23 will be 1/3 of R24. The output impedance will be R23 paralleled by R24 to a close approximation. The *power* dissipated in the transistor and resistors will be a problem.

TheZVN0545A can only dissipate 0.7W. It would be wise to derate that by half to 350mW. You have to choose the amount of the 200V supply to be dropped across the transistor versus R23+R24. The same current will flow in them, so the power will be the current times the fraction of the 200V across each part. You can choose which part has how much voltage, but you need to leave enough voltage across the two resistors so the biggest signal swing you want out can be accommodated. If you have, for instance, a +/-90V swing, you really need to make the idle voltage across the transistor be 1/2 the 200V power supply. If your signal swing is less you can use less voltage across the transistor and more across the resistors. Resistors are much cheaper to dissipate power in than transistors.

Notice that your R21/R22 choice forces the gate to nearly half the power supply, so the source is approximately at half the power supply too. You could change the R21 and R22 values to bias at some other point.

There is a lot of room here for figuring out where to bias this. If you make 1/2 the voltage across the transistor and the source resistors, the voltage across each one is 100V, and the current cannot be more than 350mW, our derated power in the transistor divided by 100V, or 3.5ma.  So R23+R24 must be 100V/0.0035A = 28571 ohms. R24 is 3/4 of that and R23 is 1/4, or R23 = 7.1K and R24 is 21.4K. You could also increase these resistances to standard values and reduce the current and power dissipation in the resistors and transistor a bit. But decreasing them will make the transistor very hot. Your output impedance then can't be lower than 7.1K || 21.4K, or about 5.33K. 

Notice that your output impedance is driven by the power dissipation in the transistor, which is probably unexpected for you.

[Sheldon]

Hello,
thanks a lot
Here is a corrected version including R.G. advices



R19 and R20 attenuate the signal with a 4:1 ratio.
For R21 and R22, as R19||R20=114k, R21 and R22 must be at least 10*114k = 1.14M => 1M is too low =>I put R21=R22=1.5M, which is somewhere between 2 common values I saw on the internet (1M and 2.2M).
For C15, when R21=R22=1M, I often saw C15=220n, when R21=R22=2.2M, I often saw C15=100n, so I guess 150n will do the trick.
For R24 and R25, I took the standard values to have R24+R25>28.6k and R24=3*R23
I guess the preamp output should be R24||R25 = 5k
For C6, I put a 1uF value (3k reactance @ 50Hz)
I'm not sure how works the 1M R26 resistor and if it raises the output impedance or not.

[R.G.]

R19 and R20 attenuate the signal with a 4:1 ratio.

You also have R24/R25 attenuating the signal in a 4:1 ratio. There is now a 16:1 attenuation. Did you mean to do both of these?

For R21 and R22, as R19||R20=114k, R21 and R22 must be at least 10*114k = 1.14M => 1M is too low =>I put R21=R22=1.5M, which is somewhere between 2 common values I saw on the internet (1M and 2.2M).

Same question.

I'm not sure how works the 1M R26 resistor and if it raises the output impedance or not.

It does not change the output impedance significantly. I does load the output a bit, but this is insignificant.

[PRR]

Zener must go Gate-to-Source, NOT Gate-to- some place low on the Source resistor string.

The plan you drew might actually work up to 25V-30v supply, but with 200V supply the Zener will conduct all-the-time and defeat the MOSFET.

[Sheldon]

R.G. : Yes the gain stages are quite "high gain" and I need a total 16:1 attenuation.

PRR : you mean I should better do something like this?

[Sheldon]

Apart from the Zener, which I suppose is in the correct position now, I noticed I have something else wrong.
I can't modify the circuit before the A1M volume pot.
I assume the volume pot is 1M because it's in the output of a tone stack.
I calculated R21 and R22 to be more than 10* the source impedance. But with the 1M pot, the source impedance can be VR3 + R19||R20 max => around 1M.
So I should take R21=R22=10*1M=10M (which can create more noise)
How can I improve the circuit?

[merlinb]

I calculated R21 and R22 to be more than 10* the source impedance. But with the 1M pot, the source impedance can be VR3 + R19||R20 max => around 1M.

No, the source impedance can't possibly be more than R20, in fact it must be less. (I make it 124k max, for the circuit you posted).

So I should take R21=R22=10*1M=10M (which can create more noise)

Increasing R21 R22 will result in less noise, not more.

I suggest you revert back to 10M. Remove R19 and R20 and instead attenuate the signal more by adjusting R24/R25.

[R.G.]

Increasing R21 R22 will result in less noise, not more.

I'm a little confused by that. The thermal noise of a resistor goes up with resistance. The bias circuit looks like R21/22 in parallel to a voltage of half the power supply, and the thermal noise from those resistors in parallel with whatever other resistances are there adds to the input noise, doesn't it?

Granted there are issues of current noise versus thermal noise voltage. Increasing the resistances lowers the current, so it decreases the current noise, but this is more simply cured with "noiseless biasing" where the bias voltage is generated by a divider of lower-value resistors and bypassed by capacitance to ground, then coupled into the gate by a high-value resistor. The high-value resistor still generates its own thermal noise, but since the current is essentially zero in this setup, the current noise vanishes.

[Sheldon]

I admit I'm a little confused too!

* so if I want a 16:1 global attenuation, a simple solution to lower the output impedance of the preamp could be to keep R19 and R20, lower R20 value to 39k, and take directly the output from R19/R20 junction (maybe keeping C16 and R26), with no use of mosfet cathode follower.
The output impedance would be around 33k, which is not great, but not too bad.

* if I want an output impedance below 10k, the mosfet cathode follower is necessery, but I'm confuse with the R21-R22 values (R.G. solution vs Merlin's one)

[samhay]

Seeing as you are driving this with a 1M pot, you could alternatively achieve your 16-fold attenuation by swapping the pot to a 100k pot to ground in series with a 1M5 resistor.

[merlinb]

Increasing R21 R22 will result in less noise, not more.

I'm a little confused by that. The thermal noise of a resistor goes up with resistance.

RG, you know as well as I do that increasing resistance in series with the signal adds Johnson noise, but increasing resistance in shunt with the signal reduces noise. Consider if you removed the shunt resistors altogether: you would have essentially infinite resistance (air), but I hope nobody would accuse all those 'air resistors' -which are everywhere- of adding infinite Johnson noise!

If the source resistace is 124k then a pair of shunt 1.5M resistors adds 2.2uV of Johnson noise (20-20kHz). A pair of shunt 10M resistors adds 1uV.