@ leforban - It sounds like some kind of RLC spike if they only burn out when you disconnect the power, especially if you’re hot-unplugging it (power is still on while you’re disconnecting the power leads from the PCB). I think at this point you’ll have to use an oscilloscope to monitor what’s going on. When you disconnect the power, does the 2.5k pullup resistor remain connected to the IO pins? This, I think, is what is needed, so it can limit any kind of current arcing that might occur if you’re hot-unplugging. You also generally want to be sure that your protection components are as close to the IO pins as practical and that the traces are as wide and short as feasible to minimize inductance.
I ha ve poor DSO on parrallel port running on windows 98 (an old velleman product) and I can not see any spike when I unplugged the power supply of the one wire. The hardware is as follows:
I have a one wire box where I provide 12V. Then a buck boost generate 5V (10microF between 12V and GND and 100microF between 5V and GND. The 5V is then connected to the 2.5k resistor to the one wire net. Before connection to the EMX board,the one wire network passes through the poly-fuse the DS9503 and the BAS40. May be I can print the schematic tomorrow.
Hmmm, to be honest, the hot-unplug theory was a bit of a stretch anyway. You normally see RLC spikes when you hot plug (power something up), instead of the other way around.
But I wonder, how do you know that your IO pins are burning out when you remove power? What if they’re burning out the next time power is plugged back in?
Here’s the schematic of the power supply for the “one wire” box.
The one wire network is approximately 100 meter long and there’s 24 DS2401. In the one wire box I have also 6 DS2408 that switch on and switch off leds. I know that pins are burning out when leds are not switch off after starting the platform. Each time it happens to me it was correlated with a manual unplugged of the power supply of the one wire box while the emx board is still under power supply and connected to the one wire network.
Hmmm, looking at the DS9503 datasheet says its Trigger Voltage is 10V typical, 11V max. That means the DS9503 will only begin protecting your I/O pins once the voltage is 10V or higher. The pins would have probably fried long before then. You should look at a voltage suppressor that triggers at a much lower level, say 3.3V or 5V.
Also, 100 meters of wire is A LOT of inductance. So I think you should make sure to take RLC effects into account. Earlier I thought the hot-unplug theory was a bit of a long shot, but now that I know there’s this much inductance, I think it’s actually plausible.
As a starting point, here’s a calculator that gives you inductance based on wire length and diameter: http://www.consultrsr.com/resources/eis/induct5.htm
Interrupting a charged inductor of that size will generate a large kickback voltage which might be what’s killing your pins: http://www.coilgun.info/theoryinductors/inductivekickback.htm
The TVS arrays that Ianlee mentioned are good options, for either 3.3V or 5V tolerant pins depending on what you choose. If you shop for different models, make sure you read the datasheet since the terminology can vary. What one datasheet calls the “clamping voltage” is the “reverse breakdown voltage” of another datasheet which is the “trigger voltage” of yet another part from a different manufacturer . . . .
OK I read your post and calculate with the tool you mention the inductance involved by the cable length. This gives: 254941nH. Therefore I think I need a an other protection to avoid kickback current. Do you think that the TVS diode mentionned by IanLee can be efficient in my case? I will choose the part on Radiosopare. Does anyone have a ref to propose? Constraints are: EMX One wire input (should be 5V tolerant) and a one wire bus whose power supply is at 4.8-4.9V. The pullup resistor is not determined yet but should be between 1 and 4.7k.
Trying to look for the correct device in Radiospares but there’s a tons of devices and sometimes not stored in a correct manner. Anyway, if I understand well you advices, since my one wire network work at 4.8-4.9 volt, I need a 5V transient diode. The fact is that there’s not a unique 5V TVS diode, there’s also other characteristics and I am lost in all the terminology. I tought it was simpler to choose the proper one. What about this one? http://docs-europe.electrocomponents.com/webdocs/0dac/0900766b80dac46a.pdf ESD5Z3.3T1.
At 4.8 - 4.9V the diode is near the breakdown voltage but the leakage is small <1mA. If I choose the 5.0V model, the Breakdown voltage is 6.2 and the IO will be already out of order…
Am I missing something?
As I have no clue on the signal wave form, i do’nt know if ESD is postive or negative, therefore do I need to have 2 diodes on opposite side each other?
ESD strikes can be positive or negative, so if you’re primarily concerned about protecting against ESD, you should pick a TVS that can protect against both polarities. If you’re primarily concerned about protecting against an RLC spike, those are generally positive only. You’re on the right track with the breakdown voltage. You need to figure out the maximum permitted voltage for EMX pins and choose your TVS trigger voltage accordingly. A TVS will list a Typical breakdown voltage, but you should also pay attention to the Min and Max values, due to manufacturing tolerances. You want to make sure your pins are protected under all conditions, despite tolerance stackup, and that your TVS will not start clipping digital signals during normal operation.
A TVS comes in discrete and array packages. If board space is important and you need to protect multiple pins, a TVS array should work well.
Whitepaper on TVS protection against ESD: http://www.littelfuse.com/industries/~/media/Files/Littelfuse/Technical%20Resources/Documents/Application%20Notes/Content/Littelfuse-Selecting_an_Appropriate_ESD_Device.pdf
I’m going to find out tonight if I can successfully hand (hot air) solder this…
Geeze that thing is tiny. In about the same space as an 0603 resistor it essentially places 8 diodes. We’ll see if I’m able to keep all the pins separated during soldering…
Looks like it is going to be fun! 
@ ianlee74 - stencil makes it easy 
For sure! I was pissed when I saw the DHL sticker on the door when I got home Friday because they wouldn’t deliver without a signature.
Unfortunately, no stencil this time. 
@ ianlee74 - i wonder if you could even try without solder and just rely on the hasl finish…
Interesting thought. I’d feel better knowing there’s additional solder but if I run into issues, I may try just putting on some flux and see what happens.
@ ianlee74 - look forward to the results, im sure you will get it sorted 
Now 20p in the cuss jar
It ended up being simple as pie. I’m loving the toothpick technique now that I’ve figured out the correct amount of paste that needs to be applied. I just basically wet the tiny pads and put a dab on the ground pad and added some air and it looks great. Here’s the completed product. Sorry, I don’t have a way at the moment to zoom in for a good pic
I had a look on the LPC2478 datasheet. In limiting values I 've found Vi equals to 6V on 5V tolerant IO pins under conditions that the 3.3V is still alive and Vdd+0.5 for other IO pins. I am using EMX.IO75 which is said to be general purpose digital I/O. I hope the TVS diode that I have chosen will be the solution!!!