Protecting Inputs and Outputs in Digital Electronics

[T3sl4co1l]

A nice gotcha with RS485 is the common mode range, particularly if there is a long cable in play with significant current flow, a 60M cable with a 6A DC supply doing remote battery charging for example had me because the screen formed the return and could put the data link outside the common mode range due the IR drop (How we laughed, the other end was 30M underwater).

If you're using RS485, do take a close look at the datasheet.  Notice the CM range is outside the supply rails -- receivers have internal divider resistors to sense the line.  If you are adding TVS diodes to such an input, don't make the mistake of using unidirectional, say 5.0V parts!  6-12V bidirectional is better.  Too low a clamp voltage and you ruin the CM range, making things worse!

Most receivers also boast a modest ESD rating themselves, so external protection isn't necessary in many cases.  When it is, it's usually because of surge, or excessive ground offset; in which case, a fully isolated receiver may be a better idea.

Tim

[dmills]

Yep, but this was shifting the ground voltage to be outside the +12 -> -7 range that RS485 does, we swore copiously.

Ended up that we could finesse it with a pair of L pads inside the plug, halved the signal level but put the signal within the common mode range, and at 9600bps level was not a problem.

Regards, Dan.

[tszaboo]

Yep, but this was shifting the ground voltage to be outside the +12 -> -7 range that RS485 does, we swore copiously.

Ended up that we could finesse it with a pair of L pads inside the plug, halved the signal level but put the signal within the common mode range, and at 9600bps level was not a problem.

Regards, Dan.

That means you had more than 7V voltage drop on the cable. Almost no matter how you look at it: that is a lot. It is not the electronics problem, but the electrical guy deciding on the cable.

[dmills]

Actually it was more of a need to re use an existing cable design, and the reworked electronics had a charger that actually worked properly.....
Sometimes when you are redoing an existing system you have to work within some constraints.

The fact they changed the max cable length from 20M to 100M did not exactly help matters.....

Regards, Dan.

[Gavin Melville]

I work with simply awful noise in industrial environments. If it's only 4kv transients on DC supply rails I'm having a good day.   In the end all interfacing is via light. Opto isolators work well, but my standard goto interface is 1mm low cost fibre, which between perfect isolation and low bandwidth keeps noise where it should be.   Is it expensive, of course. 

TVS type diodes need careful design. The turn on is just too slow, often msec.  Even small ones start clamping after 10usec.

I did get caught recently - LT makes a nice transient stopper, lt4363.  Good idea, but if there are so many transients it never turns on then it isn't much use. Perhaps better on well behaved noise.

[T3sl4co1l]

TVS type diodes need careful design. The turn on is just too slow, often msec.  Even small ones start clamping after 10usec.

Surely you mean thyristor type..?

Avalanche TVS diodes break down essentially instantaneously (sub 1ns), so that lead inductance dominates their performance.

TVS diodes in forward bias are either "normal" or "fast" recovery; I'm not sure what.  (I've never seen it in a datasheet, and I don't see any info over a selection of SPICE models, except for bog-standard zeners where they all are given the same token value "TT=50.1n").  In any case, forward and reverse recovery won't be more than a few microseconds, even for the highest voltage devices; you just can't make silicon worse than that.

So they're certainly good for surge, no matter how you use them.  As long as they're used right, because of that inductance thing, which is especially dominant for fast transients (EFT and ESD).

In contrast, thyristor type TVS turn on when the voltage is changing rapidly, or at a peak voltage (they still clamp like regular avalanche TVSs).  Turn-on takes some us, but clamping voltage is quite low, even at high currents.  They're good for clamping surges on signal lines and (non mains) AC power lines (since, they'll remain on if supplied by DC).

I did get caught recently - LT makes a nice transient stopper, lt4363.  Good idea, but if there are so many transients it never turns on then it isn't much use. Perhaps better on well behaved noise.

Hmm, with response on the order of 5us, it would only be just barely helpful with IEC 61000-4-5 surge (8/20us), but quite suitable for telecoms surge (10/1000us) and load dump.  Which is likely the intended application.  It's useless for anything faster, including ESD and EFT.

Its internal design likely uses a charge pump, so that the turn-on charging / recovery time is rather slow (ms?).  It would be a poor choice for extremely intensive environments.

Have you considered using a Big Fuckin' Inductor(tm)?  If you've got that much density of surges, it sounds like you're inside a switching converter!  Might as well harness what you can...  An inductor is big and pricey, but won't ever fail.

Tim

[Rerouter]

Another question, I will have to expose 12 bidirectional I/O to the outside world, 3.3v in, and 1.8-12.5v out, wondering if there are any good approaches for this, each pins direction will need to be independent and should ideally survive short to ground or short to 32v, (while being capable of 30mA)

[T3sl4co1l]

Not too hard, but sounds interesting if you need particular capabilities (why is it 1.8-12.5, is it analog output? adjustable digital '1'?), or a lot of them while packing it into a tight space!

Got any budget for the design?

Tim

[Rerouter]

Digital 1 at that voltage. 0 is ground, in essence I am making a diagnostic tool, which will involve talking to devices in the 1.8-5v families while being able to kick out 9-12.5v high voltage programming pulses on 6 of those pins to cover all use cases, budget is ~ $60 for the base hardware for that function, lets say $110 cap, excluding voltage regulators, as it will be used by others in the feild, it will have to survive but not necessarily work when shorted to 32v or ground on any given I/o, each pin should have 12-35mA sink/source to be safe

So essentially level shifting and protection

[pascal_sweden]

I am still puzzled about the ESD protection for non-edge programming/debugging/expansion connectors.

From my perspective, every interface on the PCB board should be protected for ESD protection.
Also all the non-edge connectors that don't reside on the edge but in the middle of the PCB board.

Such as programming and debugging connectors (JTAG, ISP).

And peripheral expansion connectors (SPI, I2C).

And connectors that directly interface with the main processor (Feature connectors).

Eventually the PCB board will be mounted in a box, where only the edge connectors come out.

But even then, one needs to open the box in the field once in a while, to connect a cable to the non-edge connector on the inside, and can touch the connector with his finger during that operation (meaning that ESD risk is also here present).

Maybe these non-edge internal connectors don't need the same level of ESD protection as the edge connectors, but at least some basic protection would be helpful.

Would a TVS diode work on all of these connectors (JTAG, ISP, SPI, I2C, Feature connector)?

Or would that interfer with the JTAG/ISP/SPI/I2C signals/bandwidth/protocol?

What solution would work best to have good ESD protection, and at the same time don't interfer with the signals/bandwidth/protocol?

I am looking for a solution that provides ESD protection to non-edge connectors, whereby the programming/debugging/expansion features still work the same as before without ESD protection.

Could any expert out here please advice? I am working towards a dead-line on a project, and need this input asap.

N.B: I have heard stories from some people, that ESD protection on a non-edge connector in the middle of a PCB board, is not very wise, as that would mean that the spike is disposed through the ground plane, and has to travel all the way from the inside of the PCB board to the outside. Basically meaning that it would harm more with actual ESD protection, than without, in the event of a spike. Is that really true? I would like to have a better understanding about this explanation with spikes being disposed through the ground plane.

[T3sl4co1l]

From my perspective, every interface on the PCB board should be protected for ESD protection.
Also all the non-edge connectors that don't reside on the edge but in the middle of the PCB board.

Meh. ESD is a risk item.

Getting field failures from ESD strikes in the middle of the board?  Tell the end users not to stick their grubby ape-fingers on it without basic ESD precautions!

Once it's in the box, protected from access by errant appendages, the protection components are no longer necessary.

PCs haven't seemed to have much trouble with this -- amateur builders are comfortable following ESD precautions.  (Well... whether or not they're necessarily even doing the right thing, but it helps that modern semiconductors are fairly robust, too.)

Would a TVS diode work on all of these connectors (JTAG, ISP, SPI, I2C, Feature connector)?

There's certainly nothing wrong with it, and I've done it before, for all of the above.  The hardest to protect are: anything with really oddball voltage or current requirements (usually only for specialized test equipment, like scope inputs or electrometers, where the protection can be customized anyway), or anything at stupendously high speed (USB2+, PCIe, etc.), which technically shouldn't need ESD protection anyway because of the excellent connector design (USB, IEEE-1394, eSATA, HDMI..), or precautions in use (PCIe is usually for internal board-to-board connections, not hotplugging; though it can be extended on cables, or used for hot-plugging in server situations).

Or would that interfer with the JTAG/ISP/SPI/I2C signals/bandwidth/protocol?

All of those are slow enough not to worry, and reside within their supply voltage range -- you can use a boring dumb zener, or a diode pair, to handle that no problem.

The capacitance might even help out with RFI, which is a related concern.

What solution would work best to have good ESD protection, and at the same time don't interfer with the signals/bandwidth/protocol?

I am looking for a solution that provides ESD protection to non-edge connectors, whereby the programming/debugging/expansion features still work the same as before without ESD protection.

Could any expert out here please advice? I am working towards a dead-line on a project, and need this input asap.

N.B: I have heard stories from some people, that ESD protection on a non-edge connector in the middle of a PCB board, is not very wise, as that would mean that the spike is disposed through the ground plane, and has to travel all the way from the inside of the PCB board to the outside. Basically meaning that it would harm more with actual ESD protection, than without, in the event of a spike. Is that really true? I would like to have a better understanding about this explanation with spikes being disposed through the ground plane.

The point of a ground plane is to shunt currents and fields around the components.  As long as the fields couple well with the traces themselves, it's fine: it acts like a Faraday cage.  ESD striking the middle of a board will couple into modest length traces, enough to upset logic values, or trigger the ESD protection circuits of ICs.  A poorly designed board might behave quite badly, like the cheap old hardware that was made with two layers of traces going every which way, and no ground plane: it's no better than a jumble of wires in air, and pretty much every single one will get a substantial fraction of a nearby ESD pulse coupled into it (hundreds of volts worth).

Still, I can't see why adding TVS diodes wouldn't help, assuming the board is designed to at least work on its own, and isn't an intentionally designed pathological case, like say, looping the TVS diode's return path completely around the board before hitting the main ground connection...

So, is this an add-on thing to an existing board?  Or a new design?  If you'd like some review or consultation, I'd be happy to take a look.

Tim

[wraper]

There's certainly nothing wrong with it, and I've done it before, for all of the above.  The hardest to protect are: anything with really oddball voltage or current requirements (usually only for specialized test equipment, like scope inputs or electrometers, where the protection can be customized anyway), or anything at stupendously high speed (USB2+, PCIe, etc.), which technically shouldn't need ESD protection anyway because of the excellent connector design (USB, IEEE-1394, eSATA, HDMI..)

I repaired many hundreds of Raspberry pi2, and there are tons of them with one HDMI line blown out. Likely not because of ESD as such but because of hot plugging and Y caps in SMPS. And probably crappy Chinese HDMI cable which often have connector shell not connected to anything. Low speed USB devices like mouses and keyboards don't require any shield at all, and often there is none.

[T3sl4co1l]

I repaired many hundreds of Raspberry pi2,

Wait, who the hell would ever pay to have something so cheap repaired?

and there are tons of them with one HDMI line blown out. Likely not because of ESD as such but because of hot plugging and Y caps in SMPS. And probably crappy Chinese HDMI cable which often have connector shell not connected to anything. Low speed USB devices like mouses and keyboards don't require any shield at all, and often there is none.

Good points!

Tim

[wraper]

I repaired many hundreds of Raspberry pi2,

Wait, who the hell would ever pay to have something so cheap repaired?

I bought them for cheap and repairs are fast to do (with my throughput optimized process  ). I can do around 50-60 pieces per day in average. So it pays off well enough.

[pascal_sweden]

N.B: I have heard stories from some people, that ESD protection on a non-edge connector in the middle of a PCB board, is not very wise, as that would mean that the spike is disposed through the ground plane, and has to travel all the way from the inside of the PCB board to the outside. Basically meaning that it would harm more with actual ESD protection, than without, in the event of a spike. Is that really true? I would like to have a better understanding about this explanation with spikes being disposed through the ground plane.

The picture in attachment illustrates the issue with ESD protection for non-edge connectors.

Isn't there any way to solve this problem? I would like to have ESD protection for non-edge connectors, while at the same time not be confronted with the problem as illustrated in the picture.

[pascal_sweden]

How to avoid the traveling of a glitch through the entire ground plane when using ESD protection for non-edge connectors in the middle of a PCB board?

Any pointers from experienced PCB board designers?

It's hard to believe for me that there really is no solution for this.

[T3sl4co1l]

Have you read this thread?  You've received many answers already...

Tim

[pascal_sweden]

The point of a ground plane is to shunt currents and fields around the components.  As long as the fields couple well with the traces themselves, it's fine: it acts like a Faraday cage.  ESD striking the middle of a board will couple into modest length traces, enough to upset logic values, or trigger the ESD protection circuits of ICs.  A poorly designed board might behave quite badly, like the cheap old hardware that was made with two layers of traces going every which way, and no ground plane: it's no better than a jumble of wires in air, and pretty much every single one will get a substantial fraction of a nearby ESD pulse coupled into it (hundreds of volts worth).

Still, I can't see why adding TVS diodes wouldn't help, assuming the board is designed to at least work on its own, and isn't an intentionally designed pathological case, like say, looping the TVS diode's return path completely around the board before hitting the main ground connection...

This makes me conclude that ESD protection can be added on non-edge connectors, as long as the ground plane architecture meets a good standard?

[T3sl4co1l]

Fine with me!

[pascal_sweden]

ESD protection on non-edge connectors it will be

JTAG, ISP, SPI, I2C, Feature connector that brings out processor pins

[pascal_sweden]

Feedback which I got from my electronic design partner in India:

"Faraday cage is a concept of EMI not ESD. This cage is created by ground stitch as shown in the picture. It'll help other portion components not to get affected from EMI from it's neighboring portion. 

For ESD what best we can do make it 4 layer board and add a ground layer in inner layer bottom of top layer and stitch vias as much as possible so that return path will be less I think so."

See attached picture.

[pascal_sweden]

I was searching the Internet today, and found out about this Industry Council regarding ESD.

Never heard about them before. But it seems a very good initiative and good source, and I therefore wanted to share here

Industry Council on ESD Target Levels:
http://www.esdindustrycouncil.org/ic/en/about

There is a lot of good and free documents there:
http://www.esdindustrycouncil.org/ic/en/documents

Especially White Paper 3 - Part I and Part II:
http://www.esdindustrycouncil.org/ic/en/documents/7-white-paper-3-system-level-esd-part-i-common-misconceptions-and-recommended-basic-approaches

http://www.esdindustrycouncil.org/ic/en/documents/36-white-paper-3-system-level-esd-part-ii-effective-esd-robust-designs

In that White Paper 3 - Part II, they also refer to a good source about EMC from the Ford Motor Company:

"In addition to protection strategies for the individual pins of an ECU, the board level design (especially the layout of ground planes) has significant influence on the system level ESD and EMC properties. Ford has published a good introductory guideline document entitled “EMC Design Guide for Printed Circuit Boards” which is available from www.fordemc.com."

Overview of EMC documents from Ford:
http://www.fordemc.com/docs/Downloads.htm

EMC Design Guide from Ford:
http://www.fordemc.com/docs/download/EMC%20Design%20Guide%20for%20PCB.pdf

[T3sl4co1l]

ESD is EMI.  It's considerably more powerful than most conventional sources!

Tim