The Evolution of ACE™ (Automated Circuit Extraction)

 Dr. Mike Heimlich, Microwave & RF Marketing, AWR
 

The concept of “automated” circuit extraction (ACE) arose from a question asked at a tradeshow a few years ago that was "if you place two MLIN elements on a schematic, does the harmonic balance (HB) simulator see the coupling?"

Interesting. The defacto approach to HB simulation, and all circuit simulation for that matter, is to view each MLIN as an S-matrix.  And, for a long time, that S-matrix has been computed by taking the MLIN circuit (or rather its layout) into an EM simulator. But what the user actually was asking for and wanting was an HB simulation that takes the coupling into consideration but at the speed of circuit simulation and not at the slower pace of full-on EM extraction. 

ACE was the answer as it gave the customer a layout-driven design experience that cut through the point tool approach of EM and the resulting S-parameter hand-stitching back into the schematic.  

From another viewpoint, it was a convergence of a whole bunch of similar problems: 

• RFIC designers trying to design on eight layers of CMOS model with lossy silicon below 
• high-speed PCB designers tackling 12-layer phone boards without a ground 
• handset PA/FEM designers trying to connect four die with hand-routed inductors on a few layers of FR4 

At first blush, these problems seem different but in actuality ACE is beneficial to all. AWR’s ACE borrows from the playbook of analog parasitic extraction and makes it uniquely microwave. The advantage is that you have "on-demand" EM, how you need it and when you need it. It's ludicrous to do a full-chip EM after you lay out the first stage of a three-stage PA, but it's not so crazy to use ACE at this stage to get a feel for what your metal is doing and to capture some of the couplings that aren't in the microstrip on your schematic. You can hand-wind a spiral on a PCB and get an approximate inductance and parasitic capacitance while you are figuring out space requirements, and then get the two decimal place accuracy with EM later. 

The usefulness of ACE isn't so much a technology as it is a frequency-range and design style. If you want to forestall managing your metal until later in the design flow (after a lot of circuit simulation), then ACE is perfect.  This mostly occurs in RF-style design below 5 or 6 GHz where you are doing most of your matching with passive elements or transistors themselves. It could be 900MHz ICs or 5.8GHz WiFi router boards. Really it comes down to an RF vs. microwave design style. If you want to capture your metal on the schematic, as microwave designers do, then ACE is good for getting the extra couplings as they are developing. But you would use ACE a lot more with a RF-style design. And of course, you always want to do a final, full-up EM with AXIEM!

ACE always...why? If you only EM everything you are typically waiting too long into the design flow and spending too much time with an EM solver to look at sensitivities of simple couplings. ACE enables you to insert those couplings into your design as models and then play with them along the lines of their "natural parameterization."  What is "natural parameterization?" For our coupling case, for example, if you have two broadside-coupled lines,  the "natural" way to represent that as a "parameterization" is by the distance separating the edges of the parallel lines. As we make the distance smaller, the coupling goes up and as we make it larger, the coupling goes down.  Now, in an EM solver, this means an awful lot of huffing and puffing: move a line, do an EM, repeat, rinse, and dry.  In the ACE world, once ACE finds the couplings for you, you have a model that you can then tie to a parameter and study that parameter by TUNING! 

For more information on the capabilities of ACE, check out the videos on AWR.TV, or download the ACE White Paper.