I'm getting gain from a FET buffer (gate tied to 4.5V). What gives?

[midwayfair]

I thought a source follower couldn't have gain. But there it is. I've got a louder signal, and the source is all the way up at a whopping 7V.

What gives? Am I imagining it?

[dschwartz]

Maybe the next stage has a very low input impedance?

[amptramp]

Some FET's are bidirectional - the drain and source can be interchanged.  The difference between boost and follower is the polarity of the power supply.  If you have a follower with the gate biased to 4.5 volts, running on one polarity gives you a follower and the opposite polarity gives you a grounded-source amplifier.

[Transmogrifox]

Probably need to see the schematic.  It is possible to unwittingly create something like a Sallen-Key type of filter between the input cap and gate-source parasitic capacitance.  You may also get a wide-band resonant hump if bootstrapping.  I assume this is a JFET if you're talking about the source being at 7V vs a 4.5V bias at the gate.

Certainly is a weird one...

[midwayfair]

Probably need to see the schematic.

Just a basic FET buffer with Vr.



Vr is 4.5V, V+ is 9V. Using a 2SK170, but it also works with a 2N5457, 2N5952, 2N3819, and MPF102. Source measures 7V and that's true for all the transistors.

Only thing after is the amp, and less than 20ft of George L's cable (~21pf/foot) total. The signal isn't hugely louder, but it's definitely noticeably so. I also popped some into a fuzz that has BJT buffers and there was a noticeably increase in distortion when substituting any FET for the BJT input buffer. (The BJT would have lower output impendence, so if the signal were unity, I would think that the resulting signal would be mildly less distorted with a FET source follower.)

Amptramp might be right that it's the fact that the FET's gate and source are interchangable and it's using ground as the voltage source. But I would have thought I'd need a P channel FET to do that!

[anotherjim]

Do you know the impedance of the amp input? They're not all as high as 1M.
Usually, you might expect a unity buffer to give an apparent boost, simply because it unloads the pickups closer to the guitar which gives a bit more level and highs. Might notice it more with humbuckers.

[midwayfair]

Do you know the impedance of the amp input? They're not all as high as 1M.
Usually, you might expect a unity buffer to give an apparent boost, simply because it unloads the pickups closer to the guitar which gives a bit more level and highs. Might notice it more with humbuckers.

It's a HRD. Schematic says 1M impedance.

This is a clear increase in level as compared to ground referencing the gate. The amp's input impedance is not part of the equation.

(I did forget to mention that my source resistor is 10K, but I don't think that would make a difference.)

Does anyone want to throw this on a breadboard to confirm or conclusively deny this?

[Transmogrifox]

Amptramp might be right that it's the fact that the FET's gate and source are interchangable and it's using ground as the voltage source. But I would have thought I'd need a P channel FET to do that!

Most JFETs are symmetrical.  The designation "source" and "drain" are arbitrary since there is no physical difference between one or the other.

What amptramp is hypothesizing is that if you connected the power supply backwards you have a common source amplifier (if it were biased correctly).  Based on the 4.5V bias and no drain {source} resistor it is unable to work as such with a JFET because it would be biased into saturation.  Simulation of this configuration confirms with a wide range of FETs this would just sound bad (gain loss and poor frequency response) if you got +9V and ground connections reversed somehow -- and you definitely wouldn't be measuring those voltages.

As a thought experiment, apply a 4.5V bias on a JFET gate.  Put a 10k in the drain and a 15k in the source.  Configured as an output off the 15k, you get approximately unity gain.  Swap the 9V and gnd and now your drain becomes source, source becomes drain, and you get a 10k in the source and 15k in the drain, a configuration for gain if biased correctly.  The only way I could get this to work in simulation was to drop the bias a lot lower in the amplified configuration or it's saturated and you don't get gain.

I am very certain that this is not due to a reversed power supply connection, and definitely not a mix-up between drain/source pins.

Per a standard text-book description of this type of circuit based on classic behavior of a transistor (or tube) configured as a source/emitter/cathode follower gain is always <1.  If the device is perfectly idealized with a transconductance approaching infiinity then the gain asymptotically approaches 1.0, so a high gain device can't push it over that unity mark.

The ways I have seen gain >1 is when you add some reactive components that can provide positive feedback.  In this case I might imagine a low-Q (wideband) resonance that provides a boost to a group of frequencies that creates a true gain over a wide frequency band.

As a proof-of-concept I came up with something in a simulation that looks a bit like a sallen-key filter.  This implementation gives +3dB boost at its peak and gain is >unity from about 350Hz up to about 3 kHz.  You would definitely notice loss of high end...also I couldn't come up with any combination of cable capacitance, faulty connections, or anything else capable of any kind of gain in simulation. 

Guitar pickups will resonate with cable capacitance but this isn't unique to a JFET buffer.  If you somehow added capacitance to ground in parallel, it might bring that resonant peak lower which could cause a perception of gain but you would also hear the treble loss.

The input impedance of a BJT buffer with a 1Meg and 10k in the emitter comes to about 666k assuming beta of 200.  Higher gain devices can get you up close to the 1Meg mark.  Still I don't think the difference between having my guitar terminated into 500k vs 1Meg is distinctly audible.  Simulating against a realistic pickup model (RLC network matched to one of the strat curves: https://courses.physics.illinois.edu/phys406/Lab_Handouts/Electric_Guitar_Pickup_Measurements.pdf) shows the BJT follower with a 2N3094 about 0.25 dB louder than the FET follower except at the resonant peak between the cable and pickup.  Then the FET allows the resonant peak to go 1 dB higher than the BJT follower.  None of this seems to indicate an explanation for a distinct perception of gain (I can detect a 1 dB change on a 1/3 octave band at 4.5 kHz when adjusting EQ on my stereo IF there is significant high content like cymbals, but such a change does not yield any perception of increased loudness).

That said I have exhausted whatever I can imagine to explain it 

Does anyone want to throw this on a breadboard to confirm or conclusively deny this?

Definitely has me curious.

[anotherjim]

Assume all is right with the buffer and the laws of physics have not been changed. Therefore the buffer really does only have nearly unity gain. In that case, what I said about the "Unloading" boost effect is a highly likely explanation. I see the HRD has some 68k switching on the input jacks. Try a dummy jack in the unused input and see what happens then.