gongii
5th October 2009 - 07:00 PM
I read an interesting article in Semiconductor International today about copper being used for wirebonding. Interestingly for frequencies >100 MHz, the path of least inductance makes the current more confined to the circumference of the wire (skin effect). Even more surprising, for two wires next to each other, the current would be further preferentially confined to the edges between them. It seems this would aggravate any pre-existing current density issues.
Enthalpy
2nd November 2009 - 11:28 PM
Skin depth is something like 6mm at 50Hz for copper and diminishes as the square root of frequency. So:
- In bonding wires, you have skin effect beginning at about 100MHz, yes. But then, do you achieve any significant current density in a bonding wire? And even worse, at 100MHz? And anyway, only DC currents have an electromigration effect.
- On the chip itself, this would need too high frequencies.
- Confined to the edges between them: this depends on the relative direction of the currents.