Top ten (10) stupid pcb layout mistakes

Ouch, hurts. Did you have to say that just now? I've got one of those with two SOT23 in a fender bender. Beats me why the layouter's system (PADS) didn't let the DRC sirens wail.

Or assuming a radius of zero for wires :-)

I'm not even sure that it's an issue, but if it is it's entirely my fault, because I made the footprint for the connector.

I have a board with a circle on it, and an LED that's distinctly outside the circle. And, I have a 3D model of the same board with an LED whose corner is occupying the same space as the connector. I _think_ that the LED isn't going to be physically bumping into the connector, but rather sitting under a flange -- which isn't acceptable, but will at least let me get a prototype done.

I made that mistake on this board, too, but fortunately and entirely by chance there's room to form the wires the way they need to go.

I did this in spite of being proud to take a system's view of things, and of having been burnt by mechanical engineers neglecting cabling. So it's not like I have many excuses to hide behind.

Yeah, it's easy to forget scope grounds. I like to use a few small holes, and insert a 2-56 screw from below as needed, nice place to clip on a probe ground.

We should get some of those surface-mount omega-shaped test points. Anybody use them?

Making sure that board-board ribbon cables are correct can be tricky. Yesterday at a design review I tore up a strip of paper, marked the connector pinouts, and folded it 3D to simulate the board positions. It was OK.

Once we had the wire pairs swapped, so 1-2 connected to 2-1 etc. Once we had to do that on purpose.

I use some of these things from Keystone.

George H.

I've seen LEDs numbered that way. Not sure why LED pins have numbers, but there it is.

I put some on my test fixture. Wish I had used more. It's nice having the loop. Hook connectors can move around on the loop without falling off.

these look interesting:

Wun Tam a guy who did software was tasked to do board layout for an in-ho use replacement for a no-longer-available product needed to fill out a proj ect because nobody else had room on their heads for that particular hat. He 'd seen a few boards in his life and the circuit was "trivially simple" (hi s words) and one criterion was to keep the board as small as possible, so h e ran leads to two potentiometers so that rotation was "reversed"- that is, turning to the right raised the "on" threshold of the controlled devices, the opposite of what users expected from similar products. Picture the volu me knobs on a radio going the wrong way.

Took me an hour to convince him to redo it.

Software people drive me crazy sometimes.

Mark L. Fergerson

But we can just reprogram the head of the user. It's just one more line of code!

:-)

I had a case where a major PCB fab house "improved a perceived ground loop condition" via direct Gerber edits and without communicating that to me. They ended up doing a free rush job with the "old" Gerbers.

After a small design change, a forgotten via connected Vcc and Gnd. The considerable amount of PCBs was saved by drilling out the via.

PADS checks for stuff like that, so we don't have plane shorts any more.

I just finished testing a sensor for an industrial application. It needed to be cheap and have protected power and I/O - it has a few energy-handling parts that are big and also indispensable. The space available limited it to 0.875 inches in diameter (1/3 less area than I originally told them I needed - in a square board), and the cost limited it to a two-sided pcb with parts on only one side. There is an excitation line running to 5 sensing elements, and it took me 3 days and

When I tested it I got about 5x the expected noise in the measurement. When I examined the excitation line, I found that it made a nice arc, directly around the processor. D'oh.

ChesterW

I just finished a prototype, the two processors on the board keept fighting for the control of the UART TX line (I got one of them swapped, TX to TX a nd RX to RX). Learned that on my next schematics, I will assign one a maste r controller and write more specific text that just RX and TX

Luckily we could rearrange the UART on one of the microcontrollers :-)

Cheers

Klaus

Just one character -- put a '-' in the right place, and you're done.

(Unless it's assembly -- then you need a "neg" instruction somewhere).

People sometimes make mistakes like that, but you need an autorouter to do things really wrong.

I don't know how he did it but at my PPoE, the librarian made a SO14 opamp footprint with the power pins swapped. No one caught it until we got the boards back. It was a very small company and a large board (~8x11"). The owner wasn't pleased.

Haven't done that in ages. DRC checks such things.

Test points are also used for bed-of-nails testing, where ground may not be needed or is readily available.