Hi all,

I'm taping out a chip with a high-speed serializer on it in about a month.  In a meeting with the team designing the package, they requested I put a ground pad between the P and N pads of the differential high-speed digital output.  They asked for this because otherwise different lengths of bondwire would need to be used.  (the pads on the die or in-line, but the landing pads on the package side are staggered).

Anyway, this makes me feel a little uneasy since the two differential outputs will not be adjacent pads on the die.  I have two options:

1.  Put a ground between the P and N of the Tx outputs.  So the five pads would look like this.

 GND TX_N GND TX_P GND

2.  Alternatively, I could just remove the GND pad in the middle and leave the dimensions the same, so it would look like this:

  GND TX_N      TX_P GND

This way the package designer is happy but I don't have a ground between signals.  I've simulated both schemes and I don't really see much of a difference in the simulation.

Which approach (1 or 2) do you suggest?  Am I missing any gotchas?    

RobG wrote on Feb 9^th, 2015, 4:09pm:


Thanks Rob.  The package designer wasn't advocating this.  I have the room in my padring so I was trying to decide the best thing to do.  

We had a phone call this morning and the package designer agrees with you.  So #2 it is.  Thanks for the help.

Interestingly, I extracted the pads, put in bondwire models, and so on and couldn't get a significant difference in the two approaches from a performance standpoint.  My thinking is that it just doesn't matter that much, but avoiding having the parallel ground bondwire in the package cavity is probably the right thing.