What's your biggest PCB mistake?

[slburris]

Since I'm just a hobbyist and don't make PCB's for a living, I have
yet to make a board that's perfect the first time.  Luckily most boards
I have been able to modify if they don't work.

So where have I gone wrong?  Let's start with the worst.  I've had two
board runs that have been little more than green coasters.

The first involved my first FPGA design where I didn't pay enough attention
to pins that were input only and needed to be outputs or I/O.  Too many mistakes
to fix that one.  Some pins didn't lead to vias that I could solder to, and
I just couldn't tack wires to pins on a VQFP100 packages well enough.

The other disaster was a 16 digit LED display for a clock where I incorrectly
made the symbol for the constant current driver chip.  Somehow I
had the pin mappings off by one pin.  On first power up, all the magic
smoke came out of those chips.  The second spin of the board worked...

Biggest fixable oops I've made was forgetting to put a clock oscillator
on a board.  "No problem, I'll just use the internal oscillator.....oh this chip
doesn't have one".  Had to mount that dead bug style with double sided foam
tape and wire the leads into the board.

Most mistakes involve mixing up TX and RX for RS232 connections, and wiring
up TDI and TDO backwards for JTAG connections.  I've learned to double check
those :-)

Scott

[Hypernova]

Neglecting thermal reliefs on polygon fills. The GND pins were a real bitch with solder.

[JohnS_AZ]

The board is 13" X 3", a 6-digit "universal LED display".
There are no traces from the common cathodes of the left most 2 LED displays to the driver.
There's no ground trace to the regulator.

The mistake? Not using a gerber viewer to double check the art before sending it off.

Luckily, this was a proto run of 8 boards (thus no silk screen or mask). Everything was fixed before the production run.

[Time]

I usually sit on a board for atleast 3-4 days before I send it off.  It always helps to go over your design again and again with fresh eyes.  Hell, even have a few beers and see if a slightly altered state brings out an error you might not of seen otherwise (though I would double check it the next day just to make sure it was real ).

I have made home revisions on boards plenty.  The first version of anything I make will have atleast one or 2 errors on it.  I think the worst mistake is probably just goofing a footprint.  I am a control freak so I never use footprint libraries other than my own which I have built up over time to have all the standard footprints with different clearances.  I like to customize the clearances when I know if I am going to be using an iron, hot air, or reflow oven.  If you take a razor blade you can cut/scrape small spots of the solder mask off points on traces and use those spots as soldering points.  If I need to cut a trace I do it with a metal engraver.  This tends to make a nice controlled trench that isn't too terribly ugly.

[TheDirty]

Most of my boards are home made, so it's never much of a problem for me, cuz I can just fix and re-etch a new board, but I've made a lot of boards and errors are pretty uncommon for me now.  I also make a lot of breakouts, so I can breadboard everything before hand and footprint errors would come out then.

The ones I have sent away for, the biggest error I've done was forgotten a diode to separate the backup battery for the RTC, so I had to scrape off some solder mask and cut a trace to solder I diode in.

Bottom right, the SOT-23 that's at a 45 degree angle.

[DJPhil]

I have this horrible habit of unnecessarily mirroring PCBs when printing them out for transfer. After the fourth time I printed out a reminder to myself with some strong language and taped it over the printer. It wasn't too much trouble when it happened because the designs were through hole and almost all passives and transistors, but I've got more than one board where I had to solder a dip on the solder side. It seemed safer than flipping the leads 180° (I was worried about breaking a pin, they were salvaged parts) and gave me more space on the top.

[VITO]

Didn't checked pin numbers of sot23 transistors in ORCAD versus datasheet, neither layout editor had checked.
Had to bent legs of 20 transistors solder them on back.
Since then i always make pin numbers of transistors and diodes visible in schematics.

[TheDirty]

I have this horrible habit of unnecessarily mirroring PCBs when printing them out for transfer.

I almost always have text in the copper.  It help, because I can see that as soon as I transfer the image.

[RayJones]

Neglecting thermal reliefs on polygon fills. The GND pins were a real bitch with solder.

+1, even more so if you cut and paste the PCB when making a panel of multiple boards and make the fatal mistake of "repour polygon"
What were perfectly connected pins to GND become islands in a sea of copper

[Neilm]

Since I'm just a hobbyist and don't make PCB's for a living, I have
yet to make a board that's perfect the first time.  Luckily most boards
I have been able to modify if they don't work.

I am a professional engineer and I have yet to make a board that is perfect first time.

And if I did I'm sure that the requirement would promptly change.

Mind you I have to design with extra complications, safety and EMC being two of them. The worst mistake I ever made was letting the mechanical engineer make a minor change to the board. He knows the layout package as he does lay them out and give me the board outline. However, I needed the board in a hurry and was out of the office for a couple of days so I told him to go and order some boards. When they arrived I found he had had to move a power track, but he didn't regenerate the copper pour so the main supply was shorted to 0V.

Neil

Yours

Neil

[Balaur]

I have this horrible habit of unnecessarily mirroring PCBs when printing them out for transfer. After the fourth time I printed out a reminder to myself with some strong language and taped it over the printer.

Hehe, happened to me also.
Somewhat managed to not realize that during the whole UV and etching phase (it was for a home project). I also drill holes, the whole shebang, nice work, perfect traces, nicely made design. Very happy about myself, I started to assemble the components (without looking at the computer, since I knew the design perfectly). And you know that slight sensation that something is amiss? And then I've realized! 
The worst is that I've tried a second time and got it wrong again: printed correctly but placed the mask on the wrong side. 

The worst design I did at work is when I've gone power hungry with POLs. I designed a board with no less than 10 POLs. Totally over designed. Only one or two clock domains actually needed a high amperage, I could safely use LDOs for the remaining ones. To top that, I also forget to add the required decoupling capacitors, making every single POL oscillate. I think that the noise generated by that think when first powered up acts as a WOW signal for alien civilizations.

[TopherTheME]

I once got a footprint wrong on a micro and used 0.5mm pin pitch instead of 1mm. No fixing that one...

[Psi]

I have this horrible habit of unnecessarily mirroring PCBs when printing them out for transfer

heh, i feel your pain.
That's probably the most annoying thing ive done when making pcbs myself using press-n-peel blue and laser printer.

I either forget to mirror a side or get the top/bottom layers 90/180deg out of alignment when making two layer boards.
The boards ive got made professionally have been pretty good so far.  I tend to make one myself first using pressnpeel to check it all works before sending off the gerber files.

[tnt]

When creating a footprint for a QFN  boost controller, I didn't scroll all the way down the PDF page and I missed the "(bottom view)" remark ... so I reversed the whole foot print. Obviously had to redo the boards ... (4 layers aren't cheap for a hobbyist)

[mikeselectricstuff]

A couple of lucky near-misses :
1) didn't spot that the DIP was 14 pin but the SO was 16, fortunately it was bodgeable - even reflowed OK!


2) didn't notice that QFN was .45mm pitch instead of .5mm, but close enough...


My advice - always "sleep on a layout" before sending, and always view the gerbers in GC-Pervue or similar.

[Psi]

It's a good idea to always print out the pcb layout 1:1 and physically position all the IC's on the piece of paper to check everything fits correct.
I normally check a few pins on each IC are going where they're supposed to at this point as well. It picks up any chips around the wrong way or mirrored.

[House91320]

Forgot to upload the g fiels for the top layer.

[wardenclyffe]

It's a good idea to always print out the pcb layout 1:1 and physically position all the IC's on the piece of paper to check everything fits correct.
I normally check a few pins on each IC are going where they're supposed to at this point as well. It picks up any chips around the wrong way or mirrored.

That's a great idea. I wish I would've thought of that before I put my 7805 on backwards. On that PCB though it was a minor problem compared to forgetting to include a ground plane.

[allanw]

Once I included a ground plane but didn't set the rules to actually connect any ground traces to it... oops. Even worse than no ground plane at all because it'd act like an antenna.

Always do a gerber viewer sanity check folks.

[tyblu]

I screwed up many footprints on my first few boards; even recently got a simple SOT23 transistor footprint flipped the wrong way around. Another beginner mistake was assuming I could top-solder through-hole components that were really impossible (row of CAT5 magnetics jacks, in this case).

[Jon Chandler]

www.FreeDFM.com is a great free service provided by Advanced Circuits.  Upload your Gerbers, they run a sanity check and email results back.  It prevents really stupid things like having a mirrored bottom layer!  That kind of ruins your whole day if not caught before the boards are made.

[eliocor]

during the creation of the stackpad for a mini-USB SMD connector, I inverted the pin numbering!
Luckily it was only a prototype, but I had to mount it à la "deadbug", including a generous amount of epoxy glue...

[Chris_R]

My biggest PCB mistake was actually on my first PCB.  I made a trace too thin for the power that it was carrying causing it to fuse.  The burnt out trace then shorted out 2 pins of a Stepper driver chip causing it to blow a very large hole in the middle of it and create a lot of smoke.
I still have a picture of it somewhere!

[williefleete]

im not really big on making PCB's however the ones i did make either had flipped traces etc and at one stage having to mount a PCB secured 7805 heatsink tab up with the package front touching the board because i got the input and output of the regulator topsy turvy
these days i either use vero board (yay) or "kiwipatch" boards which are basically solderable versions of a breadboard, same layout but more permanent which i can get free from brightsparks

[tesla500]

On my first ever PCB designed on computer (with mspaint!!), I forgot to mirror the printout before etching it, as well as putting TO-220 packages way too close together to fit. I was going to solder the single IC in upside down and bend the TO-220s to make them fit, but I ended redoing it properly.

The only big blunder that comes to mind recently was sending my A!tium files directly to PCBcart, instead of gerbers, after reading they accept them. "So much easier than making gerbers", I thought. They produced everything nicely, except my ground planes didn't exist! At the time I had no money or time to have them remade, so this was the result: Surprisingly the poor grounding doesn't seem to cause any image quality issues.