Hi Mike, 

I will do my best to cover everything and also have a few questions for clarification. 

1. When the ESD8004 was initially damaged, what was the EOS event, was it an ESD strike?

2. You mentioned you are interested in using discrete ESD diodes and don't want to use TPD4E02B04. Do you want just the best part recommendation here? I had initially recommended TPD4E02B04 because it is the same package as the ESD8004. 

3. 2kV is a standard cutoff for HBM (human body model) ESD rating in a lot of devices, so I would recommend raising the kV rating to at least 4 or even 8 to see if you can recreate the failure. 

Best,

McKenzie

Hi Mike, 

Thank you for the clarification. See below for my response:

"I'm honestly not sure, all I know is the failures we've seen where if we remove the diode the symptom/failure goes away.  We're seeing where the voltage on that node (comparator input) with no battery attached was falling below 3.15V, meaning the diode was partially conducting.  We found some that were shorted so the voltage was 0V on that node (on the comparator input), and some that were about 3V meaning it was conducting slightly.  Out of the last batch of 6 units we had, we sent 2 out to OnSemi and they found that I mentioned and just said "damaged due to Electrical Overstress (EOS) event, and that they failed the curve trace test shorted.  So did that occur due to ESD on the charger when no battery is present, or repeated insertions/removal of the battery, I'm not sure.  The battery has a flat sold pin, and the charger has a pointed pogo pin, so there is certainly some bouncing that could occur when the mating happens."

To be completely honest, Electrical Overstress Event is a very typical response because it is hard to damage an ESD diode unless you are operating outside of abx max values or going over the ESD ratings. Was it a field failure or was the ESD8004 failure found in testing? Also, I may have already asked but is anything else being damaged in the circuit or is it just the ESD8004?

"We've had more failures since we last communicated, and replacing these components is a challenge which is why I was wondering about discrete components that could perhaps be easier to change, such as 0603/0402 sized packages.  Also, I'm not sure if there are different recommendations you'd have for simple I/O circuits like this one, vs the SMBus circuits, and is a "one size fits all" approach a good one to take?  I'm fine having individual components that are more tuned to specific ESD protections.  Been doing a lot of reading on white papers, was hoping to get some suggestions from the Experts in the market on appropiate solutions.  Does larger bodies allow for higher power capabilities, and therefore more robust solutions that can take more of a beating?  Additionally, I have to update this design to add another very similar circuit, so the 4 channel would need to be a 5 channel anyway."

I would approach it with "one size fits all". We do have devices that might be more targeted to certain applications or interfaces like CAN or LIN, but you can always still use them for a different application. Higher power capabilities really only come into play with TVS diodes like SMA, SMB, and SMC and that's because they are quite large packages, and yes, they are more robust when it comes to surge current. 

As long as you don't need low capacitance, a device that comes to mind is ESDS311. It is a 1-channel device in SOD323 and has both high ESD performance and surge performance. If that package is too big then we can look at other options. 

Best,

McKenzie

Hi Mike, 

See below:

"I'm honestly not sure, all I know is the failures we've seen where if we remove the diode the symptom/failure goes away.  We're seeing where the voltage on that node (comparator input) with no battery attached was falling below 3.15V, meaning the diode was partially conducting.  We found some that were shorted so the voltage was 0V on that node (on the comparator input), and some that were about 3V meaning it was conducting slightly.  Out of the last batch of 6 units we had, we sent 2 out to OnSemi and they found that I mentioned and just said "damaged due to Electrical Overstress (EOS) event, and that they failed the curve trace test shorted.  So did that occur due to ESD on the charger when no battery is present, or repeated insertions/removal of the battery, I'm not sure.  The battery has a flat sold pin, and the charger has a pointed pogo pin, so there is certainly some bouncing that could occur when the mating happens."

To be completely honest, Electrical Overstress Event is a very typical response because it is hard to damage an ESD diode unless you are operating outside of abx max values or going over the ESD ratings. Was it a field failure or was the ESD8004 failure found in testing? Also, I may have already asked but is anything else being damaged in the circuit or is it just the ESD8004?

For the most part, it seems to be just the ESD8004 failure. However, I have some reports of the MCU being shorted or DACs being shorted, however I'm trying to correlate that data to determine if it was with this ESD8004 device, or an older model that had 5.6V zeners instead.  I know today I did ran ESD tests, I didn't see the ESD8004 show any sign of "latching up", but at -8kV we did see a I2C DAC start holding the SDA/SCL lines low and lock up, then when we reset power this went away.  The DACs were hot under the thermal camera, and were normal when power cycling.  I removed the DACs to see if it was the DACs latching the lines, or the ESD8004 latching the lines, then I really couldn't get it to fail.  Eventually I did get a failure which may be the MCU, my SCL line is being held low, or lower, sometimes it resets during power cycle but then drops back to <1V.  It's odd, I was zapping the 26V DETECT signal the whole time, I saw no issues with that!

Do you happen to have a design file you can share? If you are shocking the 26V DETECT signal but seeing issues on the SCL line then maybe a layout issue where coupling is occurring. 

"We've had more failures since we last communicated, and replacing these components is a challenge which is why I was wondering about discrete components that could perhaps be easier to change, such as 0603/0402 sized packages.  Also, I'm not sure if there are different recommendations you'd have for simple I/O circuits like this one, vs the SMBus circuits, and is a "one size fits all" approach a good one to take?  I'm fine having individual components that are more tuned to specific ESD protections.  Been doing a lot of reading on white papers, was hoping to get some suggestions from the Experts in the market on appropriate solutions.  Does larger bodies allow for higher power capabilities, and therefore more robust solutions that can take more of a beating?  Additionally, I have to update this design to add another very similar circuit, so the 4 channel would need to be a 5 channel anyway."

I would approach it with "one size fits all". We do have devices that might be more targeted to certain applications or interfaces like CAN or LIN, but you can always still use them for a different application. Higher power capabilities really only come into play with TVS diodes like SMA, SMB, and SMC and that's because they are quite large packages, and yes, they are more robust when it comes to surge current. 

As long as you don't need low capacitance, a device that comes to mind is ESDS311. It is a 1-channel device in SOD323 and has both high ESD performance and surge performance. If that package is too big then we can look at other options. 

I'm fine with the package size for this application, and using 1 per line.  You see no issue with this on SMBus or I/O lines like I'm showing in my schematic?  Does it have snapback?  

Would there be any value in installing a 50 ohm resistor inline?  I know it may shift the SMBus "lows" a little, but would it be helpful to reduce potential current into an ESD device?  Or does that reduce the usefulness of an ESD diode? 

I appreciate the support, it's a bit uncomfortable specifying a new component when I cannot confirm why mine is failing in the first place!

No, I don't see any issue with this on the SMBus or I/O lines, but having separate ESD diodes for each might be beneficial. The ESDS311 has a shallow snapback but should not have any issues with latch-up. 

Adding a resistor before the ESD diode will reduce the amount of current the diode sees, but I don't think it would be necessary here since it doesn't seem like current is an isssue. 

Hopefully we can figure out a solution for you! 

Best,

McKenzie