5

I've been facing a weird situation and unable to find the reason behind so far, therefore I frustratedly turned up for help.

I have a system comprised of 2 separated PCBs (A control box and an on-board controller), connected through a thick 10-pin cable. The system is using a 24V supply (65W) for sensors and motor, 5V for sensors and 3.3V for digital I/Os. It has been working fine until I updated both PCBs with TVS diodes for ESD protection. Things turned bad and I identified the culprit being the TVS diodes, only after burning up dozens of microprocessor chips.

The symptoms includes power-switch self-shorting (the switch contact piece will not re-bounce back when unpressed), motor driver dying (which is supplied with 24V for motor, 3.3V for signal, and GND), dropped voltage on one of the output channel of optocoupler (which display a 3.3V at first, but quickly drops down to around 1V. What came to mind is that the channel also has the shortest trace), and mainly - the micro-controller of the control box got heat up in seconds and permanently stayed in this status. My assumption is either 24V or 5V somehow got into the 3.3V trace and wreak havoc in the system. Or there could haven been a reverse current. However, during the whole process the 3A on-board fuse didn't blow.

The TSV diode layout is shown below: TSV diode layout!

The TSV diode I chose is MMBZ33VAL, and it has the Reverse Standoff Voltage at 26V. I know it might be inappropriate to use it on 3.3V and 5V application, but I believe it will only decrease the protection capability instead of compromising the system. The system will function normally (well, only as far as I can tell) If I removed all the TSV diodes on the new version boards. So must have made a mistake somewhere but I cannot identify it myself, so it would be appreciated if someone can point it out and explain why.

Thomas Lin
  • 51
  • 4
  • 1
    This is a good start for a question, but we need more information: a more complete schematic and a section of the PCB artwork would help. I cannot consider GND and DGND without knowing how they relate to each other in the circuit, for example. That being said, "properly connected" TVSs would not cause issues as you've described: either the PCB footprint is incorrect or there's something weird going on with your circuit that we cannot see (yet). I'll share an upvote as incentive for you to better define your problem. – Adam Lawrence Apr 19 '16 at 12:19
  • Do you need so many TVS diodes? I think of using them on lines that "go out to the world" where people may attach the wrong thing to your circuit. For internal connections (between boards.) I don't see the need. (But I might be wrong too... perhaps this is standard practice?) – George Herold Apr 19 '16 at 14:02
  • This may seem trivial but have you triple checked your SOT-23 physical footprint matches what you have on schematic? And also check the 24V supply output when you switch it on. – Barleyman Apr 19 '16 at 13:54
  • @GeorgeHerold I believe TVS diodes are for ESD suppression, which might be induced by human contact or plug/cable connection. The B system indicated in the picture can be used stand-alone if provided adequate power/signals, and it can be connected to A system (the control box) through a 5m cable. And beside from the cable connection pins, I/Os that are close to the exterior (e.g. LED/membrane keyboard...) is also protected with TVS diodes. – Thomas Lin Apr 20 '16 at 10:23
  • @Barleyman Yes, that is the 2nd thing I did during the debugging procedure. Since nothing is apparently abnormal, it took me quite a while before identifying the culprit. – Thomas Lin Apr 20 '16 at 10:25
  • @AdamLawrence Thanks for the comment. As for the grounds: in B system the GND and DGND is not connected (it is designed if the user uses different grounds) until they reach A system (internally connected). The DGND pin in A system is isolated and different from the DGND in B system. I didn't show the whole schematics or PCB layout mainly because the systems work if these TVS diodes are removed. the TVS diodes are placed neighbor to the input/output pins. I've checked the PCB layout with the software which turned out to be no error/difference to the schematics. – Thomas Lin Apr 20 '16 at 10:48
  • I wonder if the TVS chips are being cooked by the isolated reference planes. You may want to add (at least) a resistor between the ground planes to ensure they're at roughly the same potential. Unless there's an actual isolation requirement, I'd just connect the nets together. – Barleyman Apr 20 '16 at 12:26
  • @Barleyman Well, the fact is presently I haven't been using the User Input I/Os, which corresponds to the A system DGND. So I think they can just be considered as floating pins with nothing attached. Shouldn't be the source of the problem I think. – Thomas Lin Apr 21 '16 at 02:26
  • Any chance of having _one_ tvs in short? That could cause a good set of magic side effects. – Gee Bee Apr 21 '16 at 11:25
  • @ThomasLin Unlikely an issue but (unless you figured this out already), just for giggles, try removing the TVS that connects GND to DGND (upper right sch board A). And looks like there's one on board B too. Let me know if the problem still happens. If what Barleyman said and your grounds are floating at different potentials, maybe there's a conduction path there. Unlikely, but just try it. – Joel Wigton May 17 '16 at 20:12

1 Answers1

5

The idea is correct.

But check the followings:

  • the TVS for 26V is not a proper choice for the 24V input. Remove the TVS from the 24V input (you shall put there a tvs with a voltage over 30V as a protection). Same applies to all 24V inputs.
  • consider how TVS works: it will go short once the breakdown voltage is exceeded. Ensure that there is a component where voltage can drop, otherwise it will cause a heavy ground bounce in your circuit. For example, an input pin can look like: pin - series 330 ohm resistor - TVS to gnd, which allows the resistor to dissipate the excessive power caused by an over-voltage
  • note that TVS very likely can not trip a fuse, definitely not a 3A fuse (that requires like 6A to trip). TVS are for protecting short ESD style pulses, if you want to trip a fuse, use a varistor as a slower (but more rugged) device.
  • I think that in your case, the TVS will be tripped because its maximum breakdown voltage is 26V, and it may trip at 24V. This causes a serious ground bounce, causing the port protecting diodes of the microcontroller to open, and the reverse current makes the micro heating up.

I found it is a good idea to have a multi-level of protection, and use the TVS for ESD protection (instead of very narrow overvoltage protection). The data sheet says its maximum breakdown voltage is 26V, so actually it can start conducting at 24V or at 23V as well.

So it is a good idea to:

  • use some other mechanism as a rough primary protection (e.g. varistor, spark gaps, PCB inductors - refer to security panels and PIR sensors of a good example on this)
  • have some series resistance, and use the TVS as a secondary protection. Note that the series resistance together with the capacitance of the TVS will cause a low pass filter, so in some high-speed applications (e.g. USB) you can not afford having a series resistor.
    • the TVS shall be a little higher voltage than the allowed max. TVS have max trip and min trip voltages - the min-trip must be above the allowed max normal input.
    • note that even though you do have a TVS, there is a chance of an overvoltage because the TVS tripping point is not really precise. Therefore use another series resistor of say, 100-470 ohm, before connecting to the microcontroller.
    • the port diodes inside the microcontroller will clip the voltage in a reasonable range. Thanks to the series resistor, even though these diodes can survive less than 20mA, everything is protected and still in the safe area.
Gee Bee
  • 1,036
  • 6
  • 9
  • Thanks for replying. I do agree with your comments, but I'm not sure if there has been a mistaken on **Reverse Standoff Voltage** and **Breakdown Voltage**. I did some testing on the TVS diode, and the result does in accord with the datasheet, that I could measure about 3mA of current conducting through (with 1K resistor) under **37V**. The **Breakdown voltage** of the TVS diode is 33V, so I think it shouldn't have tripped off the diode under 24V normal operation. Anyway I'll still add some serial resistors nevertheless, it just that I think there is something else bugging the board. – Thomas Lin Apr 20 '16 at 07:40
  • I might have been mistaken by breakdown vs reverse stand off. Get your 24V power supply. Connect an amper-meter in series, and connect the TVS diode to it (just like if it would protect the input of your board). Try this with a few TVS diodes to minimize false results due to manufacturing margins. You may see: a) no current flows or only microamps - this shows that TVS is _not_ conducting, b) see some milliamps - this does not look good, consider TVS at a higher voltage, c) it will trigger the shortcircuit protection of your PSU immediately - then you need to pick a TVS at a higher voltage. – Gee Bee Apr 20 '16 at 12:39
  • I redo the test and this time measures the current with multimeter instead of the display on output of the power supply. I did it to 2 diodes, with 2 channel each, and there is no current flow below 32V. I could get a reading of about hundreds of microamps around 33V. Anyway, thanks for trying to help :) – Thomas Lin Apr 21 '16 at 03:27