My first experience with computerized printed-circuit-board (PCB) design was in 1978 when I acquired two Racal-Redac Mini systems for my service bureau, Computer Circuits. What memories remain from that adventure? I remember we taped the button down on the light pen so that we could perform the “clicks” by moving the pen just far or near enough to the screen. I also remember how tired my arm got from holding the pen up to the screen all day. We certainly welcomed the next-generation Maxi system that used a mouse.
I also remember that the auto router was pretty useless. Why? First, it added too many vias due to the strictly X/Y routing layer bias and grid requirements. Second, the routing meandered too much, wasting space. Third, the quality of the routing was poor; it often took more time to clean it up than to just manually route it from the start.
Thirty-six years later, auto-routing technology has evolved in many ways. However, we still confront most of the same problems of the past: vias, meandering, and quality. I’ve been a part of the auto-routing evolution, working with the routing teams at ASI, Cadence, Intergraph, Mentor Graphics, Redac, and VeriBest. Most auto-routing improvements supported the continually changing component and board technology along with fulfilling signal-integrity requirements. Considerable effort has been put into advancing auto-router capabilities. Yet, after all these years, the vast majority of PCB designers still don’t use auto routers.
Is successful auto routing of PCBs really all that difficult? Apparently, yes. In the ’70s and ’80s, auto routing was also limited by compute power and memory. Imagine, the PDP 11/34 used for the Redac Mini and Maxi had a 16-bit processor and contained only 64 kbytes of memory.