Board "debug" (glitches, weird behaviors, ...)

Hello Sylvain,

Why do I now get hungry again? The same happens when I read Italian schematics and turn to the page "alimentazione".

It is hard to take a stab at it from miles away. Considering that you have a switch mode supply on board and probably some hefty RAM bus traffic look at the grounds. Are they all well tied into chassis?

Look at the USB signals with a scope. Do they appear normal or are there spikes riding on them? Look at the 5V on the USB if it is used by the remote device. You can do the same with the LAN port, serial port and the signals going to the LCD. Also, look at the 5V, 12V and other voltages you might have on the peripherals. For that, switch the scope to AC and crank it up to 10mV/div. Make sure it's a good scope that doesn't blow its input if you accidentally turn the supplies on or off while it is connected.

If you suspect higher frequency noise get some clamp-on toroids, the biggest you can find. Place them over the individual cables and see if there is a noticeable change.

Then let us know what you find ;-)

Regards, Joerg

Hello Sylvain,

But alimentation sounds nicer. There are a lot of technical phrases that sound kind of pretty in your language. Just had to do a project with LCD and my favorite is retro eclairage for back lighting.

That is a huge concern. My clients would have me dunked in water if I ever did that. Maybe you should try to make a wide and short ground connection there for a test.

Any analog scope would be fine as long as its bandwidth is over 200MHz and the inputs are safe WRT to overload. Can't really go wrong with a good name brand.

A couple 100mV is a lot. It can cause clock jitter. Considering that many CPU boards run at whatever the chips are guaranteed to deliver this might push it over the edge. Missed data bits etc.

If the USB signals look ok then it is most likely that the data is being corrupted elsewhere, quite likely on the CPU board.

You can chase that stuff with a logic analyzer but that is a lot of work. I'd first clean up the ground situation and the noise on the supply. Much easier because all you need is a roll of copper tape, some caps and a large soldering iron (not the usual 30W pencil type). Be careful with copper tape, it can cause nasty cuts that hurt for hours.

Regards, Joerg

Hi Joerg,

Joerg wrote:

Damn : "Power supply" not "Alimentation" ... someday, I'll remember ;)

Yes, I know but I've run out of idea myself ...

All the ground pins of the CPU module are heavily (thru multiple big vias) to the baseboard ground plane. But the cpu module connector where USB, serial, jtag and lcd connection are, has not a single ground pin ...

I'll try to get access to the university scope, they're certainly better than the one I have access to for now.

What I've notices is a 300kHz spike until now but it's there wether the cpu is under load or not (iow whether the lcd and serial are noisy or not). It looks like it's due to the PoE module as it's the only things having a 300 kHz frequency. But it's only a couple of 100mVpp.

The usb signals looked clear but it's hard to look at the very packet where it crash ...

Thanks for the tips.

Sylvain

I read in sci.electronics.design that Joerg wrote (in ) about 'Board "debug" (glitches, weird behaviors, ...)', on Wed, 5 Oct 2005:

Loudspeakers are almost always pregnant (enceinte). (;-)

Visiting a castle in France, the guide told us we were in "l'enceinte superieure". I asked him if she was going to be mere superieure, but he didn't get it. Must be my French accent, or lack thereof.

Paul Burke