Design Article
Debugging for antenna issues in copper processes
John Ferguson and Vigen Boyajyan, Mentor Graphics Corp.
5/4/2011 5:54 AM EDT
Antenna repair
As can be inferred from the description above, the antenna effect applies to individual transistors. The failure mechanism, however, comes from the history of metal and via layers, ultimately culminating in a specific interconnect.
There are several approaches for correcting potential antenna failures.
One common approach is connecting diodes to drain charge from metal regions that may cause a failure, thereby protecting the more sensitive transistor gate oxides. For a diode to be most effective, it must be electrically close to the transistor or transistors that may fail (Figure 3). In addition, there must be enough diode strength to pull all the charge off the metals.
Another common approach is the addition of “metal jumpers.” This approach is essentially a routing change. Because the charge build-up is a direct result of the area of the metal or the area of higher-level vias on a metal line, and because the damage applies to individual transistors, it is possible to reduce the amount of charge with a direct electrical path to a transistor by cutting the metal line and jumping it up to a higher level metal (Figure 4).
Figure 4: In example A, gate A gets charge from section 1/7th of the previous blue metal line. In example B, gate A gets all of the charge of 5/7th of the previous blue metal line.
As can be seen in Figure 4, the location of the metal jumper is not arbitrary. The user must know which transistors are in failure, how much they fail by, and how much charge damage comes from the metal line or via layer in question to determine where a cut can be applied to alleviate the impact of the charge and allow the transistor to pass.
A final, less common, correction approach is the inclusion of additional “dummy” transistors. The charge from a metal line is shared across all transistors that connect to it. By connecting more transistors to the offending metal line, the charge (and associated gate-oxide damage) can be distributed more evenly, lowering the chance of impact on any individual transistor. Of course, those transistors with a less-resistive electrical path to the metal may experience a more immediate impact. Also, each transistor has its own path, and therefore its own unique damage history across multiple metal layers. As a result, this approach requires detailed knowledge of the paths from failing transistors to the offending metal, the accumulated oxide damage history of each transistor, and the total number and area of other associated gates.
Next: Improving antenna repair
As can be inferred from the description above, the antenna effect applies to individual transistors. The failure mechanism, however, comes from the history of metal and via layers, ultimately culminating in a specific interconnect.
There are several approaches for correcting potential antenna failures.
One common approach is connecting diodes to drain charge from metal regions that may cause a failure, thereby protecting the more sensitive transistor gate oxides. For a diode to be most effective, it must be electrically close to the transistor or transistors that may fail (Figure 3). In addition, there must be enough diode strength to pull all the charge off the metals.
Figure 3: Adding diode protection
Another common approach is the addition of “metal jumpers.” This approach is essentially a routing change. Because the charge build-up is a direct result of the area of the metal or the area of higher-level vias on a metal line, and because the damage applies to individual transistors, it is possible to reduce the amount of charge with a direct electrical path to a transistor by cutting the metal line and jumping it up to a higher level metal (Figure 4).
Figure 4: In example A, gate A gets charge from section 1/7th of the previous blue metal line. In example B, gate A gets all of the charge of 5/7th of the previous blue metal line.
As can be seen in Figure 4, the location of the metal jumper is not arbitrary. The user must know which transistors are in failure, how much they fail by, and how much charge damage comes from the metal line or via layer in question to determine where a cut can be applied to alleviate the impact of the charge and allow the transistor to pass.
A final, less common, correction approach is the inclusion of additional “dummy” transistors. The charge from a metal line is shared across all transistors that connect to it. By connecting more transistors to the offending metal line, the charge (and associated gate-oxide damage) can be distributed more evenly, lowering the chance of impact on any individual transistor. Of course, those transistors with a less-resistive electrical path to the metal may experience a more immediate impact. Also, each transistor has its own path, and therefore its own unique damage history across multiple metal layers. As a result, this approach requires detailed knowledge of the paths from failing transistors to the offending metal, the accumulated oxide damage history of each transistor, and the total number and area of other associated gates.
Next: Improving antenna repair
|
|
|
|
|
Navigate to related information
docdivakar
5/12/2011 1:06 PM EDT
Thanks for a good article. I was aware of the charge buildup and the tricks to take care of it at layout. The article clearly explains the differences between Al & Cu on charge buildup which also implies the Cu circuits are inherently more expensive to fabricate because of the additional number of vias needed (not to mention other costs like material cost, dual damascene & process-related differences).
Dr. MP Divakar
Sign in to Reply
John.Ferguson_#4
5/17/2011 4:16 PM EDT
Thank you Dr Divakar,
I don't have any data to validate my thinking, but I would expect that the charge that accumulates to a lower level of metal due to the etching of vias in copper should be less than the charge used to actually etch out the interconnect lines themselves as is done in aluminum. This would be worth some further investigation.
That said, you remind me of a question I've had. To help address issues associated with copper voiding, via doubling has become a common practice in nanometer IC design. But we know that more vias means greater susceptibility to antenna failure in those same copper processes. I'm curious if anybody has done any analysis on the trade-offs?
Sign in to Reply
docdivakar
6/27/2011 2:39 PM EDT
@John.Ferguson_#4: thank you for the follow up. I am not aware of any tradeoffs on the number of via's vs. susceptibility to antenna failure (I am not a process guy but do talk to couple of friends who are and hence I am dangerous with half-knowledge!)
MP Divakar
Sign in to Reply