Mosfet protection diode -- needed or not?

[mordechai]

I am making a circuit with an LPB boosting the front of a SHO...the signal goes from a trim pot at the end of the LPB to the front of the SHO section of the circuit.  Since the SHO stage comes second, can I forego the Zener protection diode?  Does it only need to stay in there if the SHO is receiving the input from a 1/4" jack, or does it need to stay in if it is taking the signal from the LPB's output trimpot's wiper, even if they're both part of a continuous circuit?

[R.G.]

The correct answer is - it depends.

If you can evaluate the circuit around the MOSFET and assure yourself that no voltage bigger than the gate-channel puncture can appear, that those voltages will be eaten or ameliorated by their path to the MOSFET, sure, leave it out. Note that people touching the back of the circuit board right on the gate connection is possible too.

Or, if you're a belt-and-suspenders kind of guy, put it in.

In the first case, the MOSFET will almost never fail from transients. In the second case it will never fail from transients. Almost never? Never?

Like the say in programming, almost always is almost always as good as always.

[jymaze]

I don't use a Zener but I use a LED or a 1N4001 wired in reverse polarity: they are rated pretty low for reverse polarity, which makes them equivalent to a 20 to 50V Zener.

It is cheaper, I have a bunch of them lying around, and it works just as well!

[R.G.]

I don't use a Zener but I use a LED or a 1N4001 wired in reverse polarity: they are rated pretty low for reverse polarity, which makes them equivalent to a 20 to 50V Zener.

It is cheaper, I have a bunch of them lying around, and it works just as well!

Not exactly. The gate isolation on most MOSFETs is specified at +/- 20V or so. The 1N4001 is only specified to withstand a minimum of 50V. It could be more, which doesn't matter, because at +50V, the MOSFET gate is long dead. It does protect well against negative-going transients, though, like any semiconductor diode with a low Vf would.

The LED is harder to figure out. The specs on reverse voltage on LEDs are loose, and the manufacturers' advice is "don't do that". Last I looked that was specified as a mimimum reverse voltage, too, so it could be bigger.

I think this works out to be halfway between "put in a zener" and "do nothing and hope". They probably do work well for you; you haven't hit the wrong combination of circumstances yet. Nothing wrong with your approach, but realistically, it's not a complete solution.

A simpler thing to do is to make note of the fact that modern NPN silicon transistors have a base-emitter breakdown of 5-7V. That's awfully low for power device use, but in the simple signal amplifier use the OP was doing, you'll never need 5V of enhancement on the gate. So clipping off the collector lead and using the base-emitter of an NPN as a zener from gate to source would be usable and cheap.

[mordechai]

That is a great idea (using the base/emitter of the NPN transistor).  If I do this do these transistor leads have to follow a particular orientation in going from the gate to source of the mosfet?

[Mike Burgundy]

A BJT as a zener, great idea! In this application, breakdowns will be rare, but does this do damage in the longrun, such as EC breakdown would, or can we assume the BJT is a zener-for-life, so to speak? Is anything known on max currents?
Is the specified datasheet breakdown value something approaching accurate or is it "It'll work up to here, beyond there be dragons, but nobody knows where"? I suspect the latter, although they won't spoend any more effort on it once they have an acceptable spread above specified value. They won't specify the width of that spread. And yes, I do realise it's trivial for this application, just wondering.
@ Mordechai: yes, approach the NPN as a diode with the base being the anode, emitter being the cathode (the "stripe" on the diode) and clip off/ignore the collector. Which leads correspond with those depends on the transistor used - see datasheet. Use it as if it was a regular or zener in that schem.

[rockhorst]

I recently bought some Zeners for this purpose. Not much of a price difference compared to a silicon transistor (probably even cheaper).

[R.G.]

A BJT as a zener, great idea! In this application, breakdowns will be rare, but does this do damage in the longrun, such as EC breakdown would, or can we assume the BJT is a zener-for-life, so to speak? Is anything known on max currents?

It's a zener-for-life. In fact, it used to be a common dodge to get a temperature-compensated reference voltage, as the tempco is nearly zero for transistors around 5-6V. Selected ones can be zero for limited ranges. The collector is open. It can't break down. Nowhere for the charge carriers to go. Max currents are as stated in the datasheet for max base currents. Kinda small compared to bigger zeners, but that's not an issue for references and protection zeners.

Is the specified datasheet breakdown value something approaching accurate or is it "It'll work up to here, beyond there be dragons, but nobody knows where"? I suspect the latter, although they won't spoend any more effort on it once they have an acceptable spread above specified value. They won't specify the width of that spread. And yes, I do realise it's trivial for this application, just wondering.

It's more like "beyond there be small lizards, but no one knows where". It's one of those things that just falls out of semiconductor physics and the adaptations to make fast, high-gain, low noise silicon devices. They all tend to come out well under 10V, generally 5-7V. I just looked at On Semi's sheet for the 3904: Base-emitter breakdown is a minimum of 6V, nothing else said.

However, the spreads are really not very wide on this one, as the base-emitter junction has a huge effect on lots of things, so it's tightly controlled, just for other issues.

The net is, you have to test them to find the exact voltage, but if you need a sloppily specified zener of about 6-7V, an NPN will do you proud.

Back when germanium devices were made by putting a slab of indium on each side of a slab of germanium and baking, the base-emitter and base-collector junctions were close(ish) to symmetrical, and EB voltages were about the same as CB. Gains, etc., were junky. Asymmetrical doping fixed that.

I recently bought some Zeners for this purpose. Not much of a price difference compared to a silicon transistor (probably even cheaper).

Not much difference. In my mind, it depends on (1) what you have on hand (2) whether you want to order stuff (3) whether you want a funny-shaped or sized part on your PCB, possibly to confuse the competition   

Mouser lists the 1N5242 (12V, which is what I'd get if I was buying them) for $5.00 per hundred. Various TO-92 transistors were $3.10 to $3.50 per hundred. Really not much difference. The decision should be on other factors than cost, all right.

[jymaze]

I guess I was lucky with my 1N4001... Although, as I understand it, it would be working if the aim is to protect from static discharge (very high voltage, with negligible current). I agree, against a lower voltage but still exceeding the gate-source rating and with a good current available it would not work at all since it would not trigger (but this case should not happen anyway). I should make some tests someday...

For a red LED, I tried to remember and I think I did read that the reverse voltage is consistently around 10 volts max (with high leakage too) so it would work much better.

I should just buy a lot of zener diodes and call it a day I guess... But they leak so much that I got paranoid about the noise (it is probably not a real concern I know...).

[R.G.]

I guess I was lucky with my 1N4001... Although, as I understand it, it would be working if the aim is to protect from static discharge (very high voltage, with negligible current). I agree, against a lower voltage but still exceeding the gate-source rating and with a good current available it would not work at all since it would not trigger (but this case should not happen anyway). I should make some tests someday...

The problem is that **any** voltage over the puncture voltage of the gate-source oxide layer damages it permanently, even at microscopic currents. So if the gate oxide was really 21.5V thick, and you put a source of 22V on it, the oxide punctures. So the difference between static discharge and lower voltages with pretty much any current available still has the possibility of damaging the gate. It's probably not every time, but the devices makers have clearly told you you're on your own for anything over the specified gate voltage.

The "should not happen" case was what I was talking about in my post where I noted that if you know your circuit will never let the gate get a damaging voltage, then your circuit prowess and canny design instincts have saved you the need for a protection device. I figure a $0.04 zener is a cheap way to pay for the mistakes I'll nearly always make at times. 

For a red LED, I tried to remember and I think I did read that the reverse voltage is consistently around 10 volts max (with high leakage too) so it would work much better.

I always check the data sheets or try the part. Note that high leakage on a protection device may be good - dissipative, damping RF rings, etc., but it directly negates the high impedance that you put a MOSFET in there for in the first place. It probably works fine - but some LEDs are in the 3-5V range of reverse as I think I read someplace, so I always check datasheets when I think I read something some place.

I should just buy a lot of zener diodes and call it a day I guess... But they leak so much that I got paranoid about the noise

Hmm. Modern zeners are actually quite good about having low leakage below their zener knee, and are not all that noisy above the zener knee.

There are two "zener" mechanisms. True zenering only happens up to about ??4-6V??, and has a fairly sloppy knee and a steeper "leakage" slope just before the knee. Above that, the mechanism is avalanche, which produces lower leakage before the knee, a much sharper knee, and even lower impedance and noise above the knee. So the collected wisdom is to use zeners rated for less than the gate breakdown, but significantly larger than the gate-source voltage you want to put there if you can. For small signal stuff, where the gate-source may only be tens of millivolts above the gate threshold of a few volts and a breakdown of +/-20V, it's easy to just grab a 12V and be far away from the troublesome, leaky, noisy knee in both normal operation and in overvoltage spikes.

[MetalGuy]

I've seen different ways to protect the G-S:
1/ One zener from gate to source.
2/ Back to back (anode to anode) zeners from gate to source
3/ Back to back (cathode to cathode) zeners from gate to source
4/ Back to back (anode to anode) zeners from gate ground.
5/ Back to back (cathode to cathode) zeners from gate to ground.

So what's the deal with all those or some of the are just schematic mistakes?
Also what would you recommend for JFET protection as I had such incidents when it goes bad for no apparent reason.

[R.G.]

I've seen different ways to protect the G-S:
So what's the deal with all those or some of the are just schematic mistakes?

It's a question of expediency, intent, and foresight.
1/ One zener from gate to source.
Clamps the gate to no more than the zener voltage above the source and no less than minus one diode voltage drop in the negative direction (for N-channels, opposite for P-channels). It takes advantage of the fact that zeners conduct in the forward direction like any other diode. This is quick, cheap, and complete as long as you don't need to move the gate below (for N-channel, which is how I'll illustrate all these) the source.

2/ Back to back (anode to anode) zeners from gate to source
3/ Back to back (cathode to cathode) zeners from gate to source
Zeners in series, whether anode to anode or cathode to cathode have the same overall result at the ends: they clamp when the total voltage goes over the zener diode plus a diode drop. They're effectively the same thing. This is complete protection, but allows gate swings both directions, which allows some of the (admittedly odd) circuits which put a reverse voltage on the gate. It's more design freedom for the designer, at the cost of another (cheap) zener.

4/ Back to back (anode to anode) zeners from gate ground.
5/ Back to back (cathode to cathode) zeners from gate to ground.
As above, these are equivalent. In this case the designer has decided that although the gate-to-source voltage is what might kill the device, it's OK, or even desirable for some other reason to tie the protection to ground. Those other reasons may include something as simple as being easier to lay out.

In all cases, the idea is to keep the gate-to-channel voltage from exceeding the voltage rating of the gate insulating oxides. The oxide layer is a layer of very pure glass that is so thin that it can be ruptured by a voltage of around 20V. The differences are all in how the designer arranges the circuit to keep that voltage from happening.

Some MOSFETs have internal gate-protection zeners. These tend to be the back-to-back zeners from gate to source to allow you to pull the gate negative by the same mount it can go positive, which is useful in some circuits.

Also what would you recommend for JFET protection as I had such incidents when it goes bad for no apparent reason.

JFETs die in a different way. The gate of a JFET dies from overcurrent/overheating. I just looked up the specs on a 2N5485. The gate breakdown voltage is 25V (it's a forward biased diode the opposite way, no breakdown) and the maximum current is 10ma. The gate junction can be thought of as a 25V, 1/4W zener. Exceed 25V, it breaks over and conducts. Ramp up the current to over 10ma, it exceeds its power ability (or, in some cases, its bonding-wire current limit, maybe) and the junction or bonding wire dies.

JFETs tend to be rugged. I have killed a few, but it was only when I was doing something either malicious or dumb to them, or both. A series resistor to keep the current under 10ma protects against all input voltages up to whatever makes 10ma flow. It has no effect on the voltage gain of the circuit, because even a largish resistor in series with a JFET gate is in series with a seriously-much-larger resistance in the reverse-biased gate. The only disadvantage to this is that there is always a larger voltage that can get you over 10ma, the series resistance works with the gate-source and gate-drain capacitance to cut high frequency response, and the series resistor adds thermal noise.

A really solid protection scheme is to use a series resistor and silicon diodes, one to a positive clamping voltage (often the power supply) and one from a negative voltage (often ground or a minus power supply voltage). If the voltage limits are less than the breakover voltage, the resistor and diodes eat all the excess current until they burn out. You might want to do this when designing stuff for Tesla's lightning lab or something. In practice, keeping a couple of spare JFETs taped inside the box is probably good enough for civilian work.

[MetalGuy]

Thanks for clearing that out! So basically a resistor in series with the gate + a zener from (+) to gate and from gate to ground will do the job. It's good to have more protection for few more cents.
What was strange it happened in a simple JFET FX output buffer (12V, gate biasing resistor from 1/2V) so I was wondering if it was and ESD issue or something otherwise I don't see how it will exceed the max DS or GS voltage or current. The other incident was a mute circuit (a J175 in parallel with the sound line).

[MetalGuy]

I'm wondering if it's a good idea to use bidirectional TVS diodes instead of back to back zeners:

http://uk.mouser.com/Circuit-Protection/TVS-Diodes-Transient-Voltage-Suppressors/_/N-5g3g?P=1z0z819Z1z0x7msZ1z0x7shZ1z0x8grZ1yzv44xZ1z0x7zq&Keyword=diode+tvs+13v&FS=True

[R.G.]

Mayyyybe...

TVS devices are essentially the same as two zeners in one package, but are optimized for transient protection service. Maybe good.

Advantages:
- one device, not two, so less labor and board space
- possibly faster, or more rugged or whatever based on their design for the job


Disadvantages:
- less common
- less well known
- relative price?? I haven't looked. Zeners are pretty cheap.
- only one zener is needed in most cases, so the one-vs-two item above may not count
- there's not much real energy/power in static, so the TVS ruggedness item above may not count for much for gate protection.

It probably comes down to availability, price, and individual preference in the limited domain of pedal electronics.

[MetalGuy]

I'm not concerned about few more cents but what I noticed is that TVS have higher capacitance than zeners. I've seen TVS diodes in Mesa DR schematics in parallel with muting FETs.

[R.G.]

Capacitance in parallel with a high impedance input is not a good thing. It can eat up your high end response.

Still, the devil is always in the details. A 2N7000 gate has about 20-50pF capacitance to source.  A gate-source protection device will parallel that, adding to any cable and wiring capacitance from the guitar if that's what drives it. Vishay publishes a chart of zener diode capacitance versus zener voltage which shows a 12V at about 30pF, so a protection zener about doubles the gate-source capacitance. Average/good guitar cable is about 30pf or more per foot. A TVS chart I found at Mouser seems to show a capacitance of about 1500pF at nearly zero reverse voltage.

What do we make of all that?

Well the biggest thing is that cable length does affect guitar pickup tone, as the pf/ft capacitance adds up and that's easy to demonstrate. So a guitar with 300pf of cable capacitance will likely sound different driving a 50pF MOSFET protected by, on one hand a 30pF zener and on the other hand a 1500pF TVS. Without actually doing the experiment, I'd guess that the highs would be duller in the TVS case. Some people would say it would sound "warmer" or "browner" too. Same thing, different name.

Whether one likes that or not is, like everthing else in music, a matter of taste.

[MetalGuy]

The difference in capacitance would affect the sound for sure. It looks like in Mesa's case it was intentional because I can't assume they didn't notice it (pages 8 and 9):

http://www.fileden.com/files/2008/12/25/2237836/Mesa%20Boogie%20Road%20King%20II.pdf

Maybe it would be a better idea to hook those TVSs from (+) to gate and from gate to source to avoid the TVS capacitance.

[amptramp]

Back-to-back diodes in series with a zener diode would isolate the capacitance but still allow the zener protection to work.  Or, a silicon diode with cathode to gate and series silicon diode with anode to gate and cathode to zener would also work.

[R.G.]

Yes, there are techniques for isolating the ill effects of protection devices, too. For isolating a zener's capacitance, one trick is to note that capacitors in series get effectively smaller. So one zener is roughly 30pF. Two zeners head-to-head is half that, about 15pF. Adding diodes in series adds more series capacitance to cut the effective value more.

This rapidly leads to the question of how much is enough. That's the tough one.

Design of protection circuitry is a fairly complex mental exercise. The trick is to balance the cost of failures and repairs against the cost of the protection devices, both in failures and in normal operations (which they may degrade), and the additional possibilities of failure caused by simply adding more parts. All this requires predicting and visualizing the results of failures, and the probability of them happening.

Sheesh!

It helps if you can find some overweening cutoff factor that makes some of this mess not matter.