High density layout standards ... placement and return path signal integrity

Thread Starter


Joined May 31, 2023
A client gave me a two layer TM4C123GXL board layout this week for design review wanting to push it into production in the 10K plus qty level. I called out a number of typical design violations and the client went crazy defending the design as meeting normal industry standards.

The layout to me is poorly placed and autorouted with traces hopping between layers several times and no attention to maintaining return path signal integrity. Eagle EDA had added over 700 microvias to stitch the badly chopped up ground planes together.

I took a long day and found a natural placement so 95% of the traces are on the top side, and a dozen or so short traces are on the edges of the bottom side to untangle ordering on several connectors. USB shortened to 8mm with series termination. Xtal shortened to 6mm Total trace length on the board is 25% of clients design. Bypass caps right against processor and other parts with 1/3 the original loop path. And of course client walked away for calling it for what it appears to be to me.

So I have to ask, is this what is now considered industry acceptable designs for high volume products?



Joined Apr 3, 2014
Indeed, it looks as if your client didn't bother reviewing the autorouter output. I see plenty of traces that can be shortened and other little autorouter "artifacts".
How much it needs to be fixed really depends on the signals on the board. If this is audio frequency, it's not likely to be an issue. It may even work up to a few MHz. Beyond that, your client should definitely look into remediating some of the issues. I agree with your signal return path integrity statement. That's an oft overlooked cause of EMI/EMC issues that gets worse with faster signal edge rates.


Joined Oct 28, 2019
So what was the purpose of you doing the design review?
If the client is looking for a scape goat for his (future) failed design then you are it.
First red flag is that they used an autorouter. No PCB layout engineer worth his salt would trust a design to autorouter. The human factor still remains (though AI seems to be replacing it)
Secondly, as someone has already pointed out, the track lengths are not optimised (kept short) May or may not be a problem, but not good design practice.
I'd be wary of continuing with this clients project as you may end up wearing the cost of the problems.
BTW to go from zero to 10k units is completely wrong.
You'd get a few PCBs made and check the operation. Its unusual that a design works right the first time. There could be interface or mechanical issues to still be solved. Then scale up from there.


Joined Oct 2, 2009
I don't see what going to 10k production has to do with the design.
If it is bad design, it is bad design, regardless of the quantity of boards to be made.

And yes, I don't use autorouter regardless of the complexity of the board.


Joined Mar 31, 2012
A client gave me a two layer TM4C123GXL board layout this week for design review wanting to push it into production in the 10K plus qty level. I called out a number of typical design violations and the client went crazy defending the design as meeting normal industry standards.

So I have to ask, is this what is now considered industry acceptable designs for high volume products?
I can't address what industry currently considers acceptable design practices for high volume production (assuming that 10k units can be considered high volume).

A lot of it depends on what kinds of signals are running around the board, where they are running around the board, and what kind of environment the board is expected to operate in.

A lot of it comes down to the old adage that if it's good enough, then it's good enough. So the question is really whether the current design is good enough.

Has the current board (or end product) already passed EMC testing? If so, then the performance might be good enough for their purposes. More to the point, it would make hesitancy to make changes to the design that would force another round of compliance testing a lot more understandable and defensible. Will all of the extra track length and vias reduce the yield and/or reliability enough to make a difference? How much of a difference is tolerable?

The fact that you made a lot of improvements in one long day would seem to indicate that the layout can be significantly improved for pretty low per-unit NRE (about a dime per board?). Especially if they already had to pay you for the work that you did, regardless of whether they use the results or not. If they let their ego get in the way of improving their product, at least to the degree that work already done will permit, then they deserve whatever happens to them.

The use of an autorouter likely indicates either that the required performance is low enough that improvements have to get over a pretty high bar to meet the justifiability threshold, or that whomever laid out the board doesn't have much of an understanding of what makes for good layout -- and why autorouters usually fall so short of it that it's often faster to just do the work by hand to begin with.

I used to let the auto place and route tool have a shot at it on the premise that it would give me a good starting point. Almost immediately I realized that trying to use anything that it produced was more hassle than it was worth. I then shifted and figured that its output would at least give me some ideas for my own layout and routing. This illusion didn't last long either. It was far more productive to spend some time studying the schematics and making sketches on a piece of paper, especially if I was the one that did the schematics because I tried to organize them with an eye toward eventual layout. Now, that's been twenty (twenty-five?) years ago and I'm sure the tools have improved, though from what I've gleaned over the years, I don't know by how much.


Joined Jan 19, 2021
So to join in on the banter here, indeed the board could have improvements as most any can, I limit myself to no more than an hour per layer for cleanup or you can clean a layer all day. As for placement and autorouting go, I am a firm believer that I can place as good as an autoplacement tool but I reserve the right to say it depends on the tool as for the Autorouter also. I use Xpedition for 12+ years and Board Station for 20+ before that, and have never used any auto placement feature, not saying it's not OK, I just don't use it. As for the Autorouter, as any seasoned designer will tell you, if the placements good, the routing should be really good too. Do I route by hand, Yes, and I also use the Autorouter interactively. Not having used any other tools, I can only say that boards without constraints these days may be doomed to fail in many applications, that said, not all will and as was mentioned, this board may work fine. As for the original design/company that did the board, if they deem it industry standard, I would question the industry. I work in Aerospace and that's not going to fly both figuratively and for real. If this was going in a toy or low cost consumer device, maybe it's OK as long as it meets all the applicable requirements, FCC, EMI, etc. If it fails in the field, they can deal with the returns and fall out, etc. I once asked an engineer where he would like the Mic Jack on his board. He looked at me with that look and asked what do you mean? I told him that if it stays designed as is, its going to radiate like crazy so he might as well put a Mic on it and make the interference useful. Not much was said after that but I did remove routes over splits and a boatload of other issues that would have failed testing, (which I found out later it had done badly). It worked great after but the point being, if you tell them it's not a good design and point out the whys, etc. and offer fixes, etc. and they say no thanks, walk away and feel good about what you did. On the other hand, if they say OK, fix it, you do your best to do just that and if you have done your homework, it will be much better. As for a out of the gate 10K production run, as was mentioned, good luck with that, somebody's going to be buying a lot of scrap boards or assembled scrap if they don't do the right thing. Even BIG companies, (The one with the I in the name as an example) don't do 10K releases unless they did a 50 or 60 unit run and test first. It's been a long time since I did work for them, so I don't know the numbers today, but if they can afford to pay the price, maybe they get lucky and maybe not. Good luck.