Just a thought - if one provides I/O protection on a devices pins to the outside world, TVS dioes etc - what about when the device is unplugged - and somebody touches the pins? Surely, all that protection against ESD no longer works, as there is no longer a ground or power connection?
Hmm...
True, ESD is not as effective with no ground. Many ICs seem to hold up okay once installed on a board tho, whether the ground plane is connected or not.
jtw11:
Just a thought - if one provides I/O protection on a devices pins to the outside world, TVS dioes etc - what about when the device is unplugged - and somebody touches the pins? Surely, all that protection against ESD no longer works, as there is no longer a ground or power connection?Hmm...
The fundamental protection is the input-protection diodes which route low-energy discharges to the power rails.
The power rails are connected together via decoupling capacitors (which have far far more capacitance than the
human body).
This means that discharges simply flow to the capacitors, whether they are charged to 5V or 0V.
However the input protection diodes are not able to handle high energy events, most CMOS chips are
rated for an idealised electrical model of a human body charged to a few thousand volts and with fingers
that behave like resistors of several kohms. They are not able to handle severe electrostatic discharge
(nylon carpet in winter, or touching a pin with a hand-held metal object - much higher peak current)
Consequently adding extra protection externally is sometimes done - adding (schottky) diodes in parallel with
the on-chip protection diodes is one good technique.
Protection circuitry is sometimes added to protect against low-impedance low-voltage events (such
as 12V being connected to a pin). Supplies able to provide large currents will overload the input
protection diodes (either melting them or sending the whole chip into "CMOS latchup mode"), so
you often see a series resistor (to limit current) followed by a zener (to limit voltage) for such protection.
You are right, the latter circuit doesn't protect against electrostatic discharge when the chip is unpowered.
Thanks for the replies thus far... so, good - I'm not completely mad...
Knowing that then, how - if at all possible, does one go about protecting a board with a connector in excess of 100 pins from somebodies highly charged fingers with their Nylon carpet?
Re CMOS latch up, need to read up on that - heard about it, but that's it...
EDIT - Good old Dave Jones... EEVblog #16 - CMOS SCR Latchup Tutorial - YouTube
EDIT, again...
I think I get it now, if somebody were to be charged and then touch the device even if it were not connected - the current limiting series resistors would prevent immediate damage, and then when it's connected to ground - the same current limiting resistors prevent excess current.
...but then, what about things like h-bridge outputs? Clearly, we don't want series R there? Maybe I don't get it fully.
h-bridge outputs typically have diodes in the design to alleviate back emf generated in the motor when current is turned off. Those can be some pretty hefty spikes. If not running and the motor is connected, the diode/motor combination will also serve as protection.
Still, sliding your feet along the carpet in the winter and then touching electronics is asking for trouble.
Still, sliding your feet along the carpet in the winter and then touching electronics is asking for trouble.
I guess that means - there's no real way to fully protect against ESD damage with a device that's not connected to a power supply?
If however, all pins on the connector have the neccesary clamping diodes, and series R where suitable - the device should be okay.
Using the example of high side FETs then, how would one protect these pins? Of course with clamping diodes, but if it's a high power output - series resistance is not possible?
I guess that means - there's no real way to fully protect against ESD damage with a device that's not connected to a power supply?
Yes you have to protect all the input pins, powered or not the protection will still work if it of the correct type.
An unopowered chip just looks like a bunch of diodes so with input series resistors, capacitors and clamping diodes you stop the static discharge. Clamping diodes to the rails still work because of those vast discharged decoupling capacitors on the supply rails absorbing the charge.
http://www.thebox.myzen.co.uk/Tutorial/Protection.html
Latch up is only a problem with unpowered chips and a powered input. Latch up occurs when when the power is restored to the chip. If the chip latches up then you have to remove the powered inputs as well as the power to the chip and then restore them all.
Ah yes of course, the board is full of decoupling caps all over the place - so there's a lot to charge there!
Latch up of that sort shouldn't be a problem, as everything connected to my device is powered by outputs of the device - so until my device sends power to the switches etc, no input voltage can be present.
I understand plugging chips in to powered prototyping circuits etc is a surefire way to latch up a CMOS chip however?
EDIT - Good old Dave Jones... https://www.youtube.com/watch?v=S0TZMivVzVk
His videos are always useful, too bad at around 9:30 into the video, when he was mentioning use of "external" clamping diodes on I/O pins, he failed to mention [mainly of use to noobees] that CMOS chips already have clamping diodes installed inside.
So, a better instructional video would have been to discuss how well those internal diodes work in conjunction with external series-Rs.
that CMOS chips already have clamping diodes installed inside.
True, but these often can only handle tiny amounts of current compared to external diodes can they not? Therefore smaller series R can be run when using external diodes?
That's why I said a better instructional video would have been helpful.
It really makes little sense to add, say 200 external clamping diodes and 100 series-Rs, to protect the pins on a mega2560 board [as if anyone would actually ever do it] when just series-Rs would be adequate 95% of the time [and which hardly anyone uses anyways].
Unfortunately, there is little hard evidence about what works. There is reputedly an Atmel appnote that says the clamping diodes can handle a couple of mA. OTOH, in an ancient technological age, about 10-years ago, I specifically tried to burn out the clamping diodes on a PIC16F876/whatever, and applied upwards to about 70-mA without damage, IIRC. For that age and time, I figured that was adequate, but today's more-highly integrated chips may be less robust.
Eg, 5V + (70mA * 330 ohms) ~ 28V, which would likely protect the chip in most circuits - back then.
jtw11:
Thanks for the replies thus far... so, good - I'm not completely mad...Knowing that then, how - if at all possible, does one go about protecting a board with a connector in excess of 100 pins from somebodies highly charged fingers with their Nylon carpet?
Put it in a bag...?
There is reputedly an Atmel appnote that says the clamping diodes can handle a couple of mA.
Not reputedly - it is this one, I have posted it before.
Atmel AVR182 Zero Cross Detector.pdf (95.1 KB)
Well yes... This is true! 
So, what about the protection of non MCU pins, such as MOSFET source/drain pins? If you're controlling multiple amps, clearly the resistor solution is no good anymore?
So I guess here, external protection diodes that can handle a voltage transient with a total energy with the devices SOA, without any series R?
For big switched loads, probably the only good protection is a fuse on the power supply.
For microprocessors, short of a series-R on each pin, probably best to use a board with a socketed processor rather than an smt chip, then you can plug in a new CPU chip when you blow it up. Preferable to tossing the entire board.
CrossRoads:
There is reputedly an Atmel appnote that says the clamping diodes can handle a couple of mA.
Not reputedly - it is this one, I have posted it before.
eh I have personally witnessed those thing slowly die out with about a ma going to them when I misread what zeiners I was using, do not trust
CrossRoads:
True, ESD is not as effective with no ground. Many ICs seem to hold up okay once installed on a board tho, whether the ground plane is connected or not.
thats half of it, its very easy to build up a charge on a board, then you come along at a much lower potential and wack! I have tons of data showing that is actually a bit more destructive than coming in with a positive charge and hitting a pin
eh I have personally witnessed those thing slowly die out with about a ma going to them when I misread what zeiners I was using, do not trust
So, (a) was the zener connected directly to the I/O pin, or (b) was there a series-R between the zener and the pin? I assume it was (a). ???
I gave up trying to read zener info from the glass, too difficult because of the glass...
I should make myself a zener tester which spits out the voltage on a readout.
oric_dan:
eh I have personally witnessed those thing slowly die out with about a ma going to them when I misread what zeiners I was using, do not trust
So, (a) was the zener connected directly to the I/O pin, or (b) was there a series-R between the zener and the pin? I assume it was (a). ???
24 volt dc, came in hit a large (33k) resistor then went to the cathode of the zener (which ended up being a 24 volt) then to the input pin, in a nutshell the input pin was exposed to direct 24 volt DC, and they started failing within a week of fairly uniform cycling
I was sniffing 8 bits from a PLC to trigger reactions in a C# program
Amazing - 24V via 33k feeding to the internal clamping diode means current into the diode was < 1mA. Those diodes are almost useless.