EEVblog #908 - Zener Diodes

[timb]

Actually, it's possible to use the Zener's junction capacitance to your advantage! All you need a fast diode in series with the Zener.

So, let's say you've got signal coming in that you need to clip (or limit) to 5Vpp. In this case, I'm using a 1MHz@10Vpp signal. We're using a 4.7V Zener.

Here's what the straight up Zener clipping circuit looks like:



Ouch, that GPIO.Vi waveform looks awful. However, because of the Zener capacitance, we expected that. So, how can we fix it? Add a diode:



Now look at that GPIO.Vi waveform, it looks *much* better! So, what exactly is going on here? Well, the first time the input goes over ~5V (4.7V Zener + 0.3V Diode), the Zener starts to conduct, then the square wave drops to 0V and Zener stops conducting, however it stays biased by the junction capacitance (that is, it holds the voltage up to just under the conduction point).

You can also use a single Zener (pre-biased from your VCC rail even) to protect multiple pins:



Either implementation works well and will allow the Zener to react within a nanosecond!

[VK3DRB]

Looks good, but you would not want to do that in some cases as you have added more parts to the circuit. More real estate. And more cost, albeit minor. The real estate can easily become an issue if you are protecting multiple lines, like in the case of an off-board data bus. In volume, a garden variety dual schottky like a BAT54 is roughly the same price as a single zener in a SOT-23 package. Depends largely on the application.

If EMI is an issue with off-board digital lines, sometimes simple series damper resistors can be used to lower the slew rate slightly, and often in so doing reduce overshoot and undershoot instead of using a zener or schottky pair.

What is the simulation package you are using on the iPad? Looks nice!

[David Hess]

Another advantage of using schottky diode clamps is that power dissipation of the forward biased junction is less than that of the reverse biased zener diode making it more likely that the later will fail under overload conditions.  Neither has good leakage though making them unsuitable for high impedance circuits.

[iromero]

The big drawback of using shottky clamps to the power rails is that if your power supply can't sink current, a big enough overload will cause the rail to soar, destroying everything attached to it. Timb's solution with a single clamp shared between several pins is a pretty elegant solution for moderately high overloads, pretty brilliant!

[VK3DRB]

The big drawback of using shottky clamps to the power rails is that if your power supply can't sink current, a big enough overload will cause the rail to soar, destroying everything attached to it. Timb's solution with a single clamp shared between several pins is a pretty elegant solution for moderately high overloads, pretty brilliant!

If your power rail is low impedance and your are using low ESR caps, it should not be a big problem. However it depends too on what you are trying to clamp. The recent revision of IEC-61000-4-2 requires 8kV contact discharge (with a peak current discharge of 30A!) and 15kV air discharge to anything accessible on a product. And that includes any exposed pins and gaps in the plastic casing. It is brutal. A zener diode on a signal line won't stand a chance of protecting against an ESD rise time of 0.8ns. The mitigation is isolation, low ESR caps, TVS diodes or physical barriers etc. Finding the cause of such issues can be painful because you cannot physically see where the discharge paths are within some complicated instruments.

[SilverSolder]

Thank you timb, great suggestion.  For ultimate performance, you could have decoupling capacitors to ground close to the high speed diodes - if the trace to the zener is long, it may impact the performance for fast rise time spikes due to inductance.

[tszaboo]

Actually, it's possible to use the Zener's junction capacitance to your advantage! All you need a fast diode in series with the Zener.

Sorry, but no. If you want to protect a GPIO (or anything alike), than the best is to use a TVS diode. It is a zener, with all the tricks they can put into it. They make sub picofarad capacitance devices for high speed interfaces. In one tiny package. Why would you even make it from discrete parts? ICs are invented, even if you "only" integrate diodes (some are even the way you re-invented).

[timb]

Actually, it's possible to use the Zener's junction capacitance to your advantage! All you need a fast diode in series with the Zener.

Sorry, but no. If you want to protect a GPIO (or anything alike), than the best is to use a TVS diode. It is a zener, with all the tricks they can put into it. They make sub picofarad capacitance devices for high speed interfaces. In one tiny package. Why would you even make it from discrete parts? ICs are invented, even if you "only" integrate diodes (some are even the way you re-invented).

A TVS is designed for ESD or other surge events. My solution is designed for clipping the input level on groups of GPIO (or general purpose digital and analog I/O) that might see repeated long duration over voltage events (like driving a 12Vpp clock into a 5Vpp input by accident). It has the side benefit of mitigating ESD.

Also, keep in mind that low voltage TVS diodes typically don't start clamping until around 7V to 20V, while my circuit can clamp at 5V or 3.3V (possibly lower, though I haven't tried it).

So yes, there is a reason you'd want to do it my way!

In addition, a similar arrangement can also be used to prevent damage from large ESD events and lighting strikes at potentially less cost (and with better protection) than putting a TVS diode on *every* pin. In fact, free_electron used just such a setup to protect the digital inputs of some telecom equipment!

[timb]

Thank you timb, great suggestion.  For ultimate performance, you could have decoupling capacitors to ground close to the high speed diodes - if the trace to the zener is long, it may impact the performance for fast rise time spikes due to inductance.

Thanks! It actually works quite well. What I ended up doing was staggering an additional Zener and cap for every 5 pins it protects. I agree that a small cap (100pF or 1nF) at each high speed diode could increase protection. You could also increase the protection resistor's value if needed; you'd want to calculate pin, diode and trace capacitance first and just make sure it didn't affect your signal.

I wish I could take credit for coming up with the idea, but I'm sure others have done it long before I thought of it. (I've come across the concept several times since, as you can see by the page from the ON Semi app note I posted earlier and in free_electron's circuit above.)

[VK3DRB]

Your statement about no other connection between system ground and safety ground other than ferrite L2 is important. Creepage and clearance must also be considered between the system and safety ground. If an ESD strike hits the system ground but not earth, it may find an alternative path to frame ground due to the presence of ferrite L2 if these clearances are not maintained.

It would be interesting to know how petrol bowsers assure there is no electrostatic discharge from a car to the fuel nozzle as the fuel pump is connected, igniting the fumes. You cannot guarantee insulation of the metal nozzle, and earthing of the nozzle would be bad. I don't know how it is done.

[David Hess]

The big drawback of using shottky clamps to the power rails is that if your power supply can't sink current, a big enough overload will cause the rail to soar, destroying everything attached to it. Timb's solution with a single clamp shared between several pins is a pretty elegant solution for moderately high overloads, pretty brilliant!

I have never actually encountered damage beyond the diode clamp itself although I have seen a couple of damaged diode clamps.  I have seen far more shorted zener diodes and a few shorted transient surge suppressors which were used where an SCR crowbar would have been more appropriate; of course the shorted transient surge suppressors did do their job by ... shorting ... and maybe that was the intention; they acted as reverse fuses by permanently closing.

If the impedance of the power supply rail is going to be too high which would be pretty unusual, then a separate rail for the clamp can be made with a transistor which will sink current to ground.  I have occasionally run across this in old instrumentation where the clamp voltage differed from ground or the supply voltage.

[GK]

If the impedance of the power supply rail is going to be too high which would be pretty unusual,

Not that unusual. If the current injected into the supply rail via the protection diode exceeds the current drawn by the circuitry powered from said rail then the supply voltage will typically/likely rise. A typical linear regulator, for example, will not sink the additional current and thus maintain regulation of the rail voltage if it is not designed to operate in both quadrants. Consider a positive regulator with an internal NPN emitter-follower series pass device. It can only source current, not sink it, so it will have no control over the rail potential under such a fault condition.

Bulk bypass/filter capacitance on the rail will only protect against surges, but are effectively open circuit to DC. So imagine that you accidentally connect your protected (diode-clamped to the supply rail) signal input to an external source of DC power of higher voltage.

So much for that protection scheme then. I've lost count of the amount of crap I've repaired over the years that was killed in part by ill-considered diode-to-the-rail "protection".

[thm_w]

It would be interesting to know how petrol bowsers assure there is no electrostatic discharge from a car to the fuel nozzle as the fuel pump is connected, igniting the fumes. You cannot guarantee insulation of the metal nozzle, and earthing of the nozzle would be bad. I don't know how it is done.

The nozzle is earthed in all cases, unless its a drum or portable can:
https://www.reddit.com/r/askscience/comments/1yqbxw/why_dont_you_have_to_ground_cars_before_fueling/

[Brumby]

Also, in regards to fuel trucks refilling service stations - I have made a direct observation of an LPG tanker doing just that.

Before any fuel lines are touched, a separate grounding wire is run from the truck to the tank installation at the servo.  It stays there during the whole process.

[David Hess]

If the impedance of the power supply rail is going to be too high which would be pretty unusual,

Not that unusual. If the current injected into the supply rail via the protection diode exceeds the current drawn by the circuitry powered from said rail then the supply voltage will typically/likely rise. A typical linear regulator, for example, will not sink the additional current and thus maintain regulation of the rail voltage if it is not designed to operate in both quadrants. Consider a positive regulator with an internal NPN emitter-follower series pass device. It can only source current, not sink it, so it will have no control over the rail potential under such a fault condition.

Bulk bypass/filter capacitance on the rail will only protect against surges, but are effectively open circuit to DC. So imagine that you accidentally connect your protected (diode-clamped to the supply rail) signal input to an external source of DC power of higher voltage.

So much for that protection scheme then. I've lost count of the amount of crap I've repaired over the years that was killed in part by ill-considered diode-to-the-rail "protection".

For external interfaces, there should always be a series resistance to limit the current if shunt protection is used.  Otherwise the shunt protection potentially lowers reliability.

Oddly enough, many of the older equipment designs I have studied have power supplies which will handle the situation you describe just fine; they include SCR crowbars on the various supply rails including the logic rail.

[UncleSam]

Nice video, Dave.

On the circuit showing a zener diode regulating the input voltage to an IC.....wouldn't it make more sense to put the resistor in series with the zener diode, but in such a way that the conducting path to the IC DOES NOT PASS  through the resistor?

[Fungus]

Nice video, Dave.

On the circuit showing a zener diode regulating the input voltage to an IC.....wouldn't it make more sense to put the resistor in series with the zener diode, but in such a way that the conducting path to the IC DOES NOT PASS  through the resistor?

No.

The idea of the Zener is to protect the IC. The path to the IC must pass through the resistor or the Zener won't be able to do its job.

[SeanB]

Also, in regards to fuel trucks refilling service stations - I have made a direct observation of an LPG tanker doing just that.

Before any fuel lines are touched, a separate grounding wire is run from the truck to the tank installation at the servo.  It stays there during the whole process.

If you look carefully at fuel tankers you will see as well on many of them a chain, or a carbon fibre bundle, that is connected to the chassis and touches the ground. this reduces the static charge on the vehicle relative to the ground. They still have the stainless steel cable ( a lot more wear resistant than a copper wire, and the little bit of extra resistance is not a worry here) with the phosphor bronze grounding clamp or pin, which is either clamped to a post inside the refuelling post or to the metal surround, before connecting the hose, and which is removed only after the hose is removed.

For cars the hose is conductive, but has a high resistance, so providing current limiting in the construction of a large carbon composition resistor. The inner braid is grounded though, and the entire pump housing is metal, and all parts have a grounding strap or wire, even the panels that cover the control switches, so that they all are bonded to the filling station ground mat.

[dwayne]

Had a friend who called his car the Zener, since when it broke down it was avalanche breakdown ...