Embedded CPU power supply ESD filter

[coldkryten]

I'm having trouble with a PC104 on a mobile robot. It is cursed with a very dry climate and carpeted floors, so massive static discharge a constant problem - it sees +/- 40V transients on the 5V supply quite frequently due to coming in contact with metal or being handled by people.

The several on-board microcontrollers handle this just fine, however the PC104 CPU tends to reboot when these transients occur. This board is very sensitive and will sometimes reboot when running off my bench supply if there is static discharge anywhere nearby.

The robot takes power at 12-15V from batteries and I use a TI PTN78020 power module to provide the 5V supply. This has capacitor/ferrite filters on both the input and output. The 5V supply then passes through a small filter with a common-mode choke, a couple large supply caps and some smaller ceramics, and a zener TVS before hitting the PC104. This seems to take a few volts off the peaks of the transients but doesn't make much of a difference; the spikes are on the order of 10nS, so I guess the TVS doesn't clamp quickly enough?

I don't have space to add a commercial PC104 power supply to the stack, and I don't have much background in filtering. Any suggestions on what I could do to mitigate this issue? Thanks!

 

[zappedagain]

Did you filter your ground lead to the PC104 too? Common mode spikes go through that wire too regardless of the name. Just watch that you don't introduce ground bounce.

What value are your caps? You need some pretty small ceramic ones (0.1uF, 0.001uF, 0.001uF) to nip the 10nS rise time; the parasitics on the larger value caps will keep them from doing anything that fast. Just watch your voltage rating, but at 40V you should have no problem with ceramic capacitors.

John D

 

[coldkryten]

The ground passes through one of the coils on the choke. The choke is two 400uH coils.

There are two 1200uF electrolytics, the small ceramics are 1uF and 0.1uF. There are two of each between 5V and GND supply from the supply, and there is a 1uF and 0.1uF from each of 5V and GND to a separate connection to chassis ground.

The above setup really doesn't make much of a dent in the spikes looking directly at the PC104 inputs.

Cheers,
Mike

 

[itsmoked]

You're sure this is a static problem and not switching noise from your motor controls?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[coldkryten]

Quite sure. The motors are very clean and reasonably well isolated. The issue is repeatable and generally occurs when there is an obvious static discharge involving the robot, even when not moving.

 

[geekEE]

Are you certain that the problem is coming in on the power lines and that's not just a secondary effect of the discharge entering through some other path? Also, how long is the path between the power supply and the CPU? Your ESD capacitors should be close to the CPU or the inductance of the power leads may be a problem. Have you tried an isolated power module?

Glenn

 

[zappedagain]

Do you have access to the reset pin on the PC104 micro? It sounds like it is a high impedance circuit so it is susceptible to the nearby EMI events. Perhaps a pull-up or pull-down resistor may make it a bit more robust.

These problems always have three sections - the source, the path, and the receptor. You can't do anything about the source (ESD events), you are already working on the path (filtering), so reducing the receptor susceptibility is another option.

John D

 

[coldkryten]

I have no idea if the issue is due to the power lines or if just shows up on it. The run from the power supply to the filter is about 5", and the run from the filter to the PC104 power inputs in about 3", both runs have 2 conductors for each of 5V and GND.

I hadn't thought about the reset pin - I assumed there is an internal pull-up somewhere but I'll try adding one and see if that makes a difference.

As the CPU currently shares a common supply with the rest of the 5V electronics, an isolated supply would involve a new board revision and isolating the serial port, but may be the right solution. I've only used isolated supplies before for ADC's or audio - would one of the small modules like the TDK CC-E series likely work well for this, or should I be looking at the heavy duty industrial packages?

Thanks for all the input,
Mike

 

[zappedagain]

For an isolated supply how about running everything off a battery? 5D cells and a 7805 should give you enough to run a quick test if it supports your current requirements.

Your comment that the 5V supply is shared made me wonder if the problem isn't somewhere else. You've seen the 40V spike on the 5V line; is another spike showing up on a PC104 I/O line?

John

 

[GonzaloEE]

I'd consider using TVS diodes and short, thick GND wires between boards.

Cheers,

 

[itsmoked]

These type problems are often NOT related to the power supply but coupling to the I/O logic. There is enough system capacitance to allow Schottky events that cause all sorts of undefined results. Are your power lines twisted? Do you provide paired lines or a metal chassis for I/O points?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[coldkryten]

Thanks to everyone for pointing me away from the power supply and towards I/O lines.

I'd left a couple unterminated wires running from the board to a PS/2 connector for a keyboard. After disconnecting this bundle, several large shocks wouldn't cause a reboot, reconnecting the wires and right back to rebooting. I'll have to do some more testing next week to confirm, but this looks pretty promising.

Cheers,
Mike

 

[itsmoked]

Great! That's the kind of thing I'd expect. Thanks for the feedback.

Keith Cress
kcress -

http://www.flaminsystems.com