How to tell if a JFET is reversible?

[commathe]

Hey everyone, I have a few experiments I have been wanting to try out with JFETs and how they are theoretically "reversible" but I have stumbled across various people saying that drain and source are not always interchangeable and to check the data sheet. The issue I'm having is I can't find a single datasheet that lists this as a feature or even mentions a specific JFET is *not* reversible.

So, how do I tell? Can anyone name a few? Part of me feels that any JFET that can act as an analog switch should be reversible, so ones that are listed as for analog switching applications should be my best bet. However, part of me is unsure of this too.

[MaxPower]

Looking in my old textbook I came across this: Although either end of most JFETs may be used as the source at low frequencies, this is not true at high frequencies.... It goes on to say that internal capacitances are minimized on the drain side.

I guess that's not helpful. Experiment I say, sacrifice a JFET or ten for knowledge and experience....

Hopefully Paul, R.G., or someone else who knows their stuff will reply.

[Lurco]

Example written on page 2: http://www.mouser.com/ds/2/302/BF245A-B-C-189056.pdf
another one http://www.mouser.com/ds/2/302/PMBFJ111_112_113-117069.pdf
more http://www.mouser.com/ds/2/302/PMBF4391_4392_4393_CNV-117377.pdf
http://www.mouser.com/ds/2/302/BF862-113763.pdf

[psychedelicfish]

Looking in my old textbook I came across this: Although either end of most JFETs may be used as the source at low frequencies, this is not true at high frequencies.... It goes on to say that internal capacitances are minimized on the drain side.

When people talk about valves/tubes, the Miller effect is often mentioned. The Miller effect is where any capacitance between the input and output of any inverting amplifier will be effectively multiplied by the gain of the amplifier. There seems to be a fairly common misconception that the Miller effect only applies to valves. This is not true, the Miller effect applies to any inverting amplifier, however it is less of a problem in solid state circuits as the transistors can be constructed to minimise input and output capacitance.

In JFETs, there is quite a significant amount (1-2nF) of unavoidable capacitance between the gate and the channel. We minimise the Miller effect by constructing the JFET so that most of that capacitance is between the gate and the source. If you reverse the JFET, using the drain as the source and the gate as the drain, you'll end up having a large Gate-Drain capacitance and the Miller effect will kick in, and you'll get lower gains with higher frequencies.

To put this into perspective, let's say we have a JFET amplifier with a gain of 4. That 1-2nF of capacitance (let's say 1.5nF) becomes effectively 6nF, and if we have an input resistor of 1M, we get an RC negative feedback network with a corner frequency of 1/(2*π*10^6*1.5*10^-9)=26.5 Hz, which is pretty close to the lowest frequency we can hear, and in our amplifier we are attenuating any frequencies above this.

So why doesn't this capacitance matter when it's between the Gate and the Source? It does matter, but not much as we don't have the Miller effect effectively increasing our capacitance and the RC network has a low R value, the source impedance of whatever is driving the FET. When dealing with high frequency switching applications we need a driver with a low source impedance to keep our switching waveform nice and sharp, minimising losses. When dealing with gainstages at guitar frequencies, however, this doesn't really matter.

[Lurco]

They only write about pF here: http://www.mouser.com/ds/2/302/BF545A_BF545B_BF545C-105462.pdf not about nF ?
http://www.fairchildsemi.com/ds/J1/J111.pdf

[commathe]

Example written on page 2: http://www.mouser.com/ds/2/302/BF245A-B-C-189056.pdf
another one http://www.mouser.com/ds/2/302/PMBFJ111_112_113-117069.pdf
more http://www.mouser.com/ds/2/302/PMBF4391_4392_4393_CNV-117377.pdf
http://www.mouser.com/ds/2/302/BF862-113763.pdf

Fantastic. At least I know it will be written.

However, will I break a JFET or experience no conductance at all for those that are not designed to be reversible? Generally, from an engineering perspective I tend to understand "high-frequency" to be outside of the range of human hearing, so I'm guessing if the only effect is a little extra capacitance (less than 10n) then it probably isn't going to bother me too much, might even add some character.

Thanks for all your help so far guys. Hope this can go on and get more in detail though!

[duck_arse]

I have datasheets for fairchild : J111/112/113, process 51: 2N5484/5/6, process 50: 2N5457/8/9, process 55: (and their appropriate mmbf numbers) and philips BFR30/31, which all state "the source and drain are interchangable".

I have to say I just stick them all in (including my j201's and j210's) without thinking, whichever pin is closer on breadboard or vero does the job. whether there is some effect I should be hearing that I'm not, or not recognising it when I do hear it, is a different matter.

[commathe]

So it seems that the general understanding so far is that *all* JFETs are "reversible", but they might not operate optimally when drain and source are swapped? The compromise is also mainly a case of capacitance, correct? Does anyone know if there is also a compromise in Idss? (or should it be Isds?)

[PRR]

The radio-frequency performance is slightly better if you use the D & S the right way.

Nearly NO audio circuit cares.

No damage will result.

[commathe]

Awesome! Exactly the kind of answer I am looking for. Thanks guys!

[commathe]

I'm back...

The same holds true for MOSFETs, am I right? Or at least it would if it were not for the body diode, yes? So if I were to get a 4-terminal mosfet then I could freely reverse the source and drain then? Probably with the same or similar capacitance issues?

This one is more a question of curiosity. I've never seen a 4 terminal MOSFET.

[R.G.]

The same holds true for MOSFETs, am I right? Or at least it would if it were not for the body diode, yes? So if I were to get a 4-terminal mosfet then I could freely reverse the source and drain then? Probably with the same or similar capacitance issues?

This one is more a question of curiosity. I've never seen a 4 terminal MOSFET.

They used to all be 4-terminal. Look up some 3N... datasheets if you want to look. These were pretty crude by today's standards, but did exist. RCA was an early producer of discrete MOSFETs, as well as inventing CMOS, which is pretty much the entire basis for today's computer logic chips. RCA exists only as a brand of TV today, if that brand exists at all. Haven't looked recently.

FETs of all stripes are a doped bar of silicon ( ... or gallium arsenide, or silicon carbide, a few other exotics) with a gate region overlaid. The nature of the raw conductivity in the semiconductor bar, the length versus width of the channel, the depth of the conductivity into the bar, all the other minutae set up how the bar will conduct.

On top of this there is a gate region that is (MOSFETs) or can be made to be (JFETs) isolated from the conductive bar, and when a voltage is impressed on the gate overlay, the electrical field (hence FIELD effect) modulates the conductivity of the conductive bar. It does this by pulling in more charge carriers or repelling them away from the gate by sheer electrical field. It's the purest form of the "garden hose effect", as it modulates conductivity by squeezing the conducting channel more or less closed.

JFET gates are isolated by the gate being the other polarity from the channel, and having a reverse voltage applied with respect to the channel. Exactly how this works in some situations like self bias are fairly complex. I'm trying to think if I've ever seen an enhancement mode JFET, or if it's possible. Haven't thought about that for many years now.

MOSFETs are a conductive layer of metal or epitaxial semiconductor which is laid over an insulating layer. In silicon this is one of the oxides of silicon; it's very pure glass, about 20-40 volts thick. Not needing the reverse bias, MOSFETs can be and are made which are enhancement mode, depletion mode, or dead in the middle with Vth = 0 ish.
Most MOSFETs are enhancement mode. This means that the conductive bar isn't very conductive at all. Putting a few volts of gate-to-channel on the gate pulls in charge carriers and opens up the hose. JFETs are depletion mode - they conduct all they possibly can if the gate is at 0V compared to the channel, but as the gate is made more negative, the depletion region from the gate pushes further into the thickness of the conductive bar and squeezes the hose closed.

If you could do a MOSFET with no substrate diode, you could use them symmetrically. This is economically impractical, but may someday be. The substrate is being etched away from regions of some semiconductor logic chips because of the rise of leakage currents in the zillions of tiny, tiny, tiny FETs that make them. Silicon On Insulator (SOI) is one of the technologies that are competing for the next extension of Moore's Law. If a way around leakage and the incredibly fine photolithograpy needed today are not found, we may be nearing the end of the doubles-every-few-years rise of computing power, as a sidelight.

[commathe]

So ultimately, though MOSFETs are theoretically reversible, they are *not* reversible in reality due to limitations that mean they need the substrate diode. So it wouldn't be possible even if I got a 4 terminal and left the substrate connection floating?

Also, out of curiosity (again), what would the advantages to a 4-channel mosfet be? Why have they become almost obsolete? I imagine most of the answers to this wont really affect pedal building but knowledge is always a good thing.

[R.G.]

So ultimately, though MOSFETs are theoretically reversible, they are *not* reversible in reality due to limitations that mean they need the substrate diode. So it wouldn't be possible even if I got a 4 terminal and left the substrate connection floating?

The conduction part is reversable, as you note. Manufacturing processes always have a substrate today. I've never seen a SOI single device. The 4-terminals have different connections to the 4th terminal. Some of them have two gates, one for signal and one for gain control.

I went looking for the old datasheets. Seems like the source was connected to the substrate. The 4th lead was usually a second gate. It may not be possible to separate the substrate. And with processes that old, it may not be that the substrate forms a body diode.

Modern CMOS chips like the unbuffered 4001 can sometimes be used for no-body-diode transistors as long as the applied voltages are entirely within the + to - range of the applied power supply.

Also, out of curiosity (again), what would the advantages to a 4-channel mosfet be? Why have they become almost obsolete? I imagine most of the answers to this wont really affect pedal building but knowledge is always a good thing.

The screaming advantage to dual gate MOSFETs was always in RF circuits. The very high input impedance and separation of the two gates let you have modulation without feedthrough, and far less distortion (which meant less spurious cross modulation products to get you into trouble with interference). The screaming disadvantage is that the processing wasn't all that good, so the gains weren't necessarily high, and the current they could conduct was quite small compared to any of today's many-section power MOSFETs. They were also excruciatingly sensitive to static shock. They were sold with all pins shorted, and you were to solder them into the PCB before removing the shorting wires.

You could make an OK tremolo with a dual gate MOSFET. If you could afford them. I don't know when or if production stopped, but they're rare. I did find some listings for some 3N1xx and 3N2xx devices, but they seem to be $10-$50 each.

[PRR]

JFETs are "always" made horizontally, across the surface of the crystal. You pick two points for D and S, dope a junction over the space between to be G. The G is either centered or *slightly* offset. There is little to no difference D or S.

The vast majority of modern MOSFETs are "Vertical". One end (S?) on top, a lattice of oxide-insulated G, and the other end (D?) down under that. So there is a real physical difference between S and D: one is the bulk crystal, the other a bit of fluff on top, and the G may be a lot "closer" to one than the other.

I am not sure how this affects symmetry. My impression is that for small voltages and low frequencies, it isn't very asymmetrical. But most of the parts we can get are "big", several Amps. And the capacitances go up with current-ability; OTOH our signals are often low current. If we are managing mA signals with Amp devices, capacitance is likely to be a problem well inside the audio band. Not to mention that the vertical construction "always" has a hidden diode, so you can't put much reverse voltage D-S or it just sucks.

[commathe]

Thanks guys! Massive help!