> They mention a host of things that the autorouters are bad at as evidence that autorouters are useless, instead of recognizing the opportunity to improve autorouters to make their jobs easier.
Sounds like a job for DL! No, really, I'm serious.
If you can feed in enough existing layouts, and give the inputs enough information in terms of what signals are differential pairs, what is the intended impedance, what is high-speed, etc. it might be possible to have a system learn how to do this.
I'd think there's enough gerbers and such laying around the Internet from reference designs to have a big enough data set.
Free business idea for anyone who wants to disrupt the EDA market...
Sounds like a job for DL! No, really, I'm serious.
If you can feed in enough existing layouts, and give the inputs enough information in terms of what signals are differential pairs, what is the intended impedance, what is high-speed, etc. it might be possible to have a system learn how to do this.
I'd think there's enough gerbers and such laying around the Internet from reference designs to have a big enough data set.
Free business idea for anyone who wants to disrupt the EDA market...