I have been looking into how to best guard rings and this is what I think up to now:
- in principle it is best to connect the P+ and N+ rings to their independent package pins
- however, when the maximum pin count is limited and/or when the amount of analog blocks is large, compromises have to be made
- the 'victim' block is isolated by P+ and N+ rings and these rings are connected to the analog blocks respective GND and VDD connections
Here are my questions:
- for the noise source block, a P+ ring is easily implemented (e.g. just down bond the guard ring connection); assuming that no dedicated pins are available any more for N+ ring connection, what should this be connected to? Or would it be best to leave out the N+ ring at the noise source all-together?
- to save pins, has anyone ever considered RC filtering the VDD line from another block to the N+ ring, so that no additional package pins are required, but that the N+ supply is still clean (cleaned up by the RC filtering)? This seems feasible, since the rings don't really carry current, so the RC time constant can be made sufficiently low without occupying much die area
Looking forward to anyone's thoughts and/or experience in this!
... assuming that no dedicated pins are available any more for N+ ring connection, what should this be connected to? Or would it be best to leave out the N+ ring at the noise source all-together?
- to save pins, has anyone ever considered RC filtering the VDD line from another block to the N+ ring, so that no additional package pins are required, but that the N+ supply is still clean (cleaned up by the RC filtering)? This seems feasible, since the rings don't really carry current, so the RC time constant can be made sufficiently low without occupying much die area
Looking forward to anyone's thoughts and/or experience in this!
IMHO filtering the supply voltage of guardRings doesn't make sense. Their purpose is to
- catch straying electrons (N+ guardRings) or holes (P+ guardRings) to prevent them from triggering parasitic action (BJT or thyristor action)
- shield from noise
For both intended purposes, they should be connected directly to their appropriate voltage levels.
Thanks for sharing your thoughts on this.
When connecting an N+ ring at the noise source to VDD, would this then just be the VDD of this particular noise source or something else?
I'd suggest the VDD connection with the least noise level, the "purest" VDD available. No further filtering. IMHO, an additional filter resistor just would attenuate the shielding function.
I have found that the guard ring connections need to be low impedance, in which case RC filtering is out of the question (this only increases the impedance).
Also, I have seen an UMC appnote where P+ N+ guard rings are used, but both are connected to GND.
Connecting N+ guard rings to GND wouldn't help much for their main designation: to suppress parasitic action possibly leading to latch-up. It would, however, still help to screen against noise.