I've come across an issue that i'm a little stumped by. Of course, one can protect an IC against negative transients using TVS diodes, active clamping or whatever other means - however, when using diodes for example - one will never be able to clamp the positive rail exactly at 0V, it may end up at negative potential depending on the clamping voltage of the TVS.
So, with that in mind - to what degree can ICs be expected to survive negative transients in terms of time, and voltage? 1nS, 1uS, 1mS etc? -1V, -15V, -4V etc?
Thanks for your reply. This is my question however - there are plenty of solutions to get the number close to 0, but at what point can damage be expected?
"To what degree can ICs be expected to survive negative transients in terms of time, and voltage? "
Consider the 100 top selling ICs of this year. There is a variety of survival traits. Let me speculate on typical microprocessors that run at 3.3 volts :
VSS = 0.000 volts, VCC = -2.0 volts
60 picosecond pulse : no damage noticed in 168 hours of rigorous testing by thirty methods and sixty criteria.
10 nanosecond pulse : after testing by three teams of experts for 1000 hours, a problem was noticed in which one Flash cell threshold voltage had shifted by 0.02 volts.
10 microsecond pulse : one bond pad resistance has increased from 0.02 ohms to 0.03 ohms.
10 millisecond pulse : four pins have open circuits
10 second pulse : an explosive vaporization ejects hot aluminum for twenty foot radius. Ceramic shrapnel injures an intern without eye protection. Latch-up on 44 pins drew 66 amps for 35 milliseconds, rapidly dropping to 1 mA.
You could also use a zener to clamp below the rail. If you have 5 volts rail to rail, you can use a 4.5 volt zener from the opposite rail.
May you explain this a little furthur, can't be cathode to the positive rail as 4.5v would just conduct at normal 5v operation, and cathode to negative would just conduct forward.
This isn't an input pin I'm protecting, but rather the input to a voltage regulator.
The answer is (should be) in the absolute maximum ratings of datasheets.
Ranges from -60V to -0.3V for the devices on my board, albiet not listed with a pulse time - seems the majority of Schottky diodes are around 350mV forward voltage, which is still a little too high.
For example, if I can only clamp to -0.3V - then surely this will still degrade electrolytic caps for example over time?
I'm just a little stumped, I can see that clamping to low voltages is fine for the majority of ICs, but for things like caps which don't have absolute minimum values listed - I don't know where is safe.
caps have a limited life time, I'm not sure that so low reverse values will have an impact on it . We have a caps wizard here, if he sees your post, he'll tell us
So, you are not talking about a minimum voltage, as being closest to zero.
You are talking about the maximum reverse voltage (being a negative voltage) ?
Which would be a problem for many types of capacitors, but not for "Ceramic capacitors". Are we on the same page now?
This isn't an input pin I'm protecting, but rather the input to a voltage regulator.
With a voltage regulator you have another (perhaps better) option... A series diode. When reversed, you'll have essentially zero voltage, and perhaps more importantly essentially zero current.
The only downside is that you will increase your regulator-dropout voltage by the diode-drop. In most voltage regulator applications you have more than one volt to spare and dropout voltage isn't an issue.
If you can handle increasing the dropout voltage by 2 diode-drops you can use a bridge rectifier and the thing will work normally with the voltage reversed!!!! (That's assuming the ground/negative is isolated so it can be "reversed" also.)
You are talking about the maximum reverse voltage (being a negative voltage) ?
Yes, not continuous - but rather as a transient, continuous reverse is already dealt with, with active circuitry as I'm flowing in excess of 100A, so a diode bridge isn't an option here.
In terms of the series diode - thanks - yes, I had thought of the series diode idea this morning and that's solved the problem for the regulator powering only control ICs. However, the small negative voltages are still coupled to devices such as my h-bridges - which I'd rather it wern't.
I guess it's a case then of clamping with a small Schottky and if the current draw is low enough, some series R to prevent excess current. However, that dosen't protect bulk electroylytic capacitors from such events.
alnath:
I think there is no universal answer, each IC is different.
The answer is (should be) in the absolute maximum ratings of datasheets .
For CMOS generally you can take Vdd to -1V, at that point all the protection diodes start conducting
(they start at about 0.5V and there are two diodes in series for each pad). Hence a DIP CD4000 series
chip becoming hot if plugged into socket backwards.
This action itself is clamping and only high currents are going to cause issues I think (such as CMOS
latchup or burnt-out protection diodes).
Of course if other pins are held at +ve voltage you'll already be taking protection diodes way beyond their
ratings...
Datasheets for logic chips generally show details of all protection circuitry.
Thanks - I've already got reverse protection such as that, albiet much more advanced that outputs impeding shutdown signals etc - so I'm fully protected from the transients that may be input to the device on the powerbus, however - what I'm dealing with here, is the transients that this device may potential create in itself.
The power bus is protected, filtered etc by my device, and then that power is used to power bridges and other devices - but, when one of these devices suddenly switches off - then the PCB trace inductance etc will create some sort of transient, and couple it to the other devices taking power from that source too.
So it's almost like I need a much smaller clamping device on the input to all my devices, and the transient energy created from the PCB tracks etc cannot be that high?
