PCB Routing Solutions: Introduction

This is the first of a series of blog topics related to routing PCBs

My name is Charles Pfeil and my role at Mentor Graphics is one in which I always have a primary focus on routing technology and methods – although I do get involved in numerous other projects. I began in this industry working my summers in high school inspecting Ruby-Lithe designs for scratches and repairing them. Sounds rather antiquated? Indeed, this began in 1966 (some would say “way back” in 1966). I went to college to study architecture but soon realized that my passion for PCB design was so strong that I should pursue that instead.

After years of working as a job-shopper designing boards, I opened my own design service bureau, Sunshine Design. We started out with manual design and tape-ups. I still have some of those Mylar sheets with black tape-ups if anyone knows of a museum that might be interested them…I suddenly feel very old. In 1978, we made the change from manual design to CAD after getting two Racal-Redac Minis. During this transition, we lost all our customers who weren’t interested in CAD data. Fortunately, we were able to hook up with some customers who also used the Racal-Redac systems. At that point, CAD routing technology, both manual and automatic, became of great interest to me.

Since then I have worked at Racal-Redac, ASI, Cadence, Intergraph, VeriBest and Mentor Graphics. This industry has been very good to me and my career. I am grateful for all the wonderful people that I have met and worked with over these years.

CharlesIntroBlogPhoto

My goal for this blog is to provide meaningful discussion of routing problems and their solutions. Over the years the routing task has become much more complicated, not only due to evolving component and fabrication technology, but also the need to fulfill high-speed requirements and manage the signal integrity concerns. I look forward to hearing your ideas and methods and hope that the discussions we have will help better equip us to solve the routing challenges that we face.
Here are some topics I look forward to discussing with you:

  • Length of time needed to manually route a design
  • Difficulty of phase match tuning
  • The dense, tangled netlines between BGAs are hard to route
  • The need to center traces between vias in BGA fanout array
  • Fanout pattern complications and the difficulty of adding them

What challenges are you encountering? Are there any topics you would like me to add to my list?

Post Author

Posted March 21st, 2014, by

Post Tags

Post Comments

6 Comments

About Xpedition Enterprise Blog

Learn about the industry's most innovative PCB design flow, providing integration from system design definition to manufacturing execution.

We'll keep you updated on the latest design trends, and other interesting tech tidbits in the industry.
Xpedition Enterprise Blog

@mentorpcb tweets

  • Mentor Graphics Announces Xpedition System Designer for Comprehensive Multi-board systems development. http://t.co/uBJe9PVYfp
  • Only two more days until U2U Conference in Bangalore! Register here for the event happening October 31st: http://t.co/w0fDzGfCD7
  • Did you hear the news? Mentor Graphics Appoints Vice President for PCB Design, Read more here: http://t.co/7Yi7LXi2D5

Follow mentorpcb

Comments

6 comments on this post | ↓ Add Your Own

Commented on April 8, 2014 at 7:44 pm
By Brian Yu

Actually, I am not sure how much to match the length of time per frequency.

[…] This is the second post in this routing series. View the introduction post here. […]

Commented on April 11, 2014 at 7:12 pm
By Charles Pfeil

To Brian Yu,
I asked Patrick Carrier of Mentor to provide some help with your comment, “Match length requirements vary between different kind of signals. DDR3, for example, requires tighter matching per byte lane, and the amount of matching should typically be in the 10s of picoseconds because the bus margins are typically on the order of hundreds of picoseconds. That equates to matching ~100mils. If we look at high-speed differential signals, on the other hand, they require very close matching to keep their edges aligned. If they have 50ps edges, that means there should be less than 5ps of skew between their positive and negative sides, which means length-matching tighter than 30 mils. Actually, it is probably best to keep differential pairs matched to within 5 mils to maximize the amount of margin.”

Commented on April 25, 2014 at 4:01 am
By Chiggs

I would like to be able to specify certain pins as flexible or specify the pinout using constraints and for the placement algorithm to automatically swap nets around to optimise the routing.

Commented on July 10, 2014 at 11:07 am
By Jack

How did you get the 3D picture in the post above? Is that a “view” in Expedition?

Commented on July 21, 2014 at 11:38 am
By Charles Pfeil

To Jack:
The image was created using the Expedition PCB 3D Viewer.
Charles

Add Your Comment