For wedge bonding on the bare die pads, is there a maximum number of grams-force that should be used? We've been dealing with some shorts when wedge-bonding our parts without about 27 g... they seem to mostly be on the HV_clamped pads

I have not seen any specific information about wedge bonding on these parts. I do know that a number of people attempting wedge bonding have reported shorts. The top oxide is relatively thin and the ground and power lines are close to the pad. I am not certain whether failures are due to excessive force or to missing the pad center and shorting to one of the vddio/vssio power rails. What diameter wire are you using? Is this a manual, semi-automatic, or automatic wire bonder?

Thanks for the reply Tim. We use 1mil Al wire on an automatic bonder. We had previously dealt with our wire bonder slightly missing the pads, so we had thought the imprecise bond placement was causing the shorts. But, we made some adjustments to our bonder to make things more precise and ended up using it in manual mode. The bonds all looked well placed but still found the same shorts.

Can't think of anything specific to the HV clamp pads. Are they all shorting to ground? Are all the HV pads doing this, or just some subset, or is it not possible to tell without tediously testing after every single bond?

They were shorting to ground yes. I ended up removing the wires on the circled pads in order to resolve these shorts (in the sense that our power supply no longer becomes current limited), which is why I think it might have to do with the HV clamp pads. It's very possible still that there are other pads which are shorted in other ways but I have yet to find those

Thanks for all the detailed feedback. Some more questions: You circled the

vssio

pad on the bottom but not the one on the top. Did you wirebond the one on the top? I assume that you can't tell if

vssio

shorts to its own bus, and that you have circled it only because

vssio

and

vddio

are the pins that are shorting together? Did you wirebond either of the domains

vdda1

or

vdda2

? I'm guessing not.

Yes, we wire bonded all the other pads, except for a few io_pads that were unused by our project. I ended up removing the vssio wire on the bottom on accident just because it was close to the vdda wire I was removing

I'm curious to know if you can measure a short from

vdda1

or

vdda2

to

vssio

. If the

vdda*

domains are not connected to any internal circuitry, then they could short to ground with no adverse effect (i.e., no current draw) because there's nothing connected to them inside the padframe. Trying to gather enough data to make an educated guess about the failure mechanism going on.

Will do

I did find out yesterday that the fill generation code was using the wrong layer name for the glass cut, and as a result, there is unexpectedly a bunch of metal fill shapes on metal 4 directly underneath the pad. That undoubtedly reduces the amount of pressure required to cause a short. If you told me that any of your pads were shorting, and that they were shorting at random to vddio, vssio, or both, I'd have a higher confidence that that's the culprit. Otherwise, I don't have a good understanding of why the high-voltage clamp pads would be more likely to short than others. You did say "*mostly* on the HV_clamped pads". . . Are you seeing other shorts as well? The GPIO pads might be immune to shorting because the metal 3 shapes directly under the pad are tied to the pad net, but all of the power and ground pads have lines of both vddio and vssio running directly under the pad on metal 3 for the full length of the pad. I do know that we get good reliability from ball bonding, which is a bit lighter on the pressure than wedge bonding.

Thanks for checking for us Tim. In the past, we had an analog pad that was shorted to vddio that we had fixed by rebonding. I believe then that one possibility is that the shorts are then not actually occurring on the HV_clamped (the vddio) pads. Instead, it is a vddio-vssio short happening on any one of the pads where the applied pressure was too high. When I had said that removing the wires from the io power pads fixed the short, it is likely I was just seeing the fact that no power was being supplied to the vddio domain... we will investigate more into ball bonding

I am not going to try to guess the physics at that level of detail. : )

fair enough! we'll report back if we find a reliable way to bond

I don't have the document from the packaging company specifying the parameters of the wirebonding job. Jeff DiCorpo should have that information.