10

I have a DB25 I/O connector, thru-hole. The pins connect to an SMT MCU, which I want to protect from ESD, specifically IEC 61000-4-2. I want to use SMT Zener diodes to protect the pins.

I am considering various layouts. I imagine the optimal layout would have the diodes between the DB25 and the MCU. In this way, an ESD event can be shunted to ground before it gets to the MCU

MCU <-> Diodes <-> DB25

However, I would like to take advantage of the thru-holes in the DB25 to simplify routing and reduce the number of vias that I would need. However, in doing so, the diodes will end up on the "other side" of the DB25.

MCU <-> DB25 <-> Diodes

Is this a bad idea? I'm slightly concerned about whether a sufficiently fast ESD strike could "split up" and reach the MCU before the diodes begin fully conducting.

If this is the case, would it be mitigated if the MCU <-> DB25 traces were run on bottom-layer, while DB25 <-> Diodes traces were on the top layer? Would the added vias between the MCU and DB25 encourage the ESD current to go through the diode instead?

ajs410
  • 8,381
  • 5
  • 35
  • 42

2 Answers2

12

ESD is difficult to deal with, and solutions are more black magic than science. That being said, what you want is for the impedance to ground to be smaller than the impedance to the chip you're protecting. There are several ways to do this, and the most practical solution will probably involve several of these things at once.

  1. Placement and routing of traces is a good start. As you noted, MCU <-> Diodes <-> DB25 is probably the best, although MCU <-> DB25 <-> Diodes can work. To make it work, the traces to the diodes should be thick and short. The traces to the MCU should be long-ish and thin. But, IMHO, just doing this is not enough for a commercial product.

  2. Put some sort of of resistor or ferrite bead between the DB25/Diodes and the MCU. I prefer resistors for this because their impedance is more predictable at high frequencies, but a bead could work too. A resistor of around 10 to 50 ohms is good, depending on the nature of the signals you're running. This resistor/bead will increase the impedance to the MCU, guiding the ESD to ground through a different way.

  3. Put a capacitor in parallel with the diodes. A value of 3 nF is ideal for ESD protection. But depending on your signal you might have to use a smaller or larger one., or none at all. The largest you can get away with will also reduce your EMI problems. The basic function of the cap is to quickly absorb the ESD shock and re-emit it more slowly and with a smaller voltage. If the cap is large enough then the diode is not required. This cap also forms an RC filter with #2 above and prevents EMI from going in or out of the box.

  4. Connect the shield of the DB25 to chassis ground, and make sure your chassis make a good shield.

Recently I had an issue with a USB device that would crash whenever an ESD zap happened within 8 feet of the box. In the end I had to connect the USB shell to Chassis, add 33 ohm resistors to the USB data lines, add caps, and diodes. Until I did all that I still experienced failures. If I left off one of those, any one, it would fail. Now it runs solid, even with 1 inch long sparks right to the chassis.

davidcary
  • 17,426
  • 11
  • 66
  • 115
  • 2
    You had to add 33 ohms to the USB data lines? *And* caps and diodes? Wouldn't that do terrible things to the USB eye diagram? – ajs410 Apr 01 '11 at 20:35
  • 2
    It wasn't bad. This was USB 1.0, not version 2 or 3. So the data rate wasn't bad. If I remember correctly, the caps were only 22 pF and the diodes were <1 pF. While I was skeptical about 33 ohms, I saw the same resistors used in a TI MSP430 demo board schematic. In the end, it works beautifully. –  Apr 01 '11 at 21:08
  • Just curious. How did you generate this ESD "zap"? I mean, is there some predictable, consistent way to generate a zap? – Earlz Apr 02 '11 at 02:58
  • 2
    In the past I've used several methods to generate a zap. By far the most reliable method was a static gun. This is an expensive piece of gear made for this task. I have also used push button piezoelectric BBQ grill lighters. Not as predictable, but less than US$10. But the zap's I was making a couple of weeks ago was simply normal ESD created due to the super low humidity here in Colorado. Again, not as predictable but super abundant. –  Apr 02 '11 at 14:05
  • I'm curious, did you have a solid inner ground plane in your application? – ajs410 Apr 04 '11 at 16:27
  • @ajs410 It is rare that I have to design a PCB without at least 1 gnd plane. Most of the time I have 3 non-overlapping planes: analog, digital, and chassis. Plus at least 1 power plane, sometimes more. –  Apr 04 '11 at 16:45
3

To start with I would use special ESD suppression diodes instead of common zener diodes; they're faster and withstand the high voltage better.

Your concerns about the relative placement are justified. The current may indeed split up and reach both protection diode and the controller. Therefore alway place the diode between the connector and the controller, and don't place them on a stub trace, because you will create the same problem. Place the ESD diode on the trace itself.

Make sure the distance and resistance to a ground plane are as short as possible. The larger the ground area the larger its capacity, and the lower the remaining voltage.
Don't count too much on earth, that's too far away; a discharge can zap all you CMOS before it reaches earth.

If possible try to make some kind of "lightning rod", this can be a bare trace ending at 0.1mm from an also bare ground trace, so that an ESD discharge can spark over the gap.
For one project we had a small slit in the enclosure which exposed the PCB to a distance of 2.5mm to the outside of the enclosure, at a place which would be touched by the user (button). So I feared that ESD discharges might pass through the slit. I removed all copper nearby, and placed an 0603 resistor, with one end connected to ground, the other under the slit. The idea was that, if we can't avoid a discharge, we at least know where it passes, so the resistor should work as a lightning rod. A resistor instead of a 0\$\Omega\$ jumper reduces the discharge current, which otherwise would couple to nearby traces and induce excessive voltages there. ESD tests results were fine.

stevenvh
  • 145,145
  • 21
  • 455
  • 667
  • I have heard of the "lightning rod" referred to as a spark gap, a piece of copper with no mask on it. I have read that this is a very good technique for many-kV, especially in combination with something a little "faster" to catch the few-kV stuff. The bare copper can also add some parasitic capacitance, which may or may not be a concern depending on application (for me, it's not) – ajs410 Dec 03 '11 at 01:44
  • @ajs410 - If the traces to the spark gap are pointy (as they should be) capacitance will be in the femtofarad range, I can't think of many applications where this would cause trouble. – stevenvh Dec 03 '11 at 08:22