OK I know tradition says I start at #10, but I'm going right for number one. (the oops I just made.)
1.) You've got the mirror image, wrong side out.(it would be less embarrassing, if I hadn't made this mistake before. I can show you a demo unit that has the back panel silk screen on the inside.)
I hoping to hear of other mistakes. (I'll make a check list, or something, for the next one.)
George H.
I suspect that it's just 1a, but on my first PCB layout I laid the IC out as the mirror image.
Overlapping components is a good one -- I'll be getting a PCB in the mail that pretty likely has that.
Leaving out room for wires on connectors that plug into the board.
-- Tim Wescott Control systems, embedded software and circuit design I'm looking for work! See my website if you're interested http://www.wescottdesign.com
I've done that, and so have experienced layout draftsmen that I've known. It happens.
The mistake (someone elses) that I remember was the one that put the 100mA current going through a big PROM onto a crucial length of narrow ground ret urn, which gave a 10mV voltage drop between a divided-down voltage referenc e and the very precise analog circuit that turned the divided voltage into the current through the electron microscope lens.
Soldering a bit of heavy copper wire onto the track was a pretty effective fix.
A bit of local "star" grounding worked even better.
-- Bill Sloman, Sydney
We rarely make a serious mistake any more, because we review boards multiple times and we have a gigantic checklist.
I did have a mistake on a recent board: the pinout of a 4-pin optocoupler somehow got scrambled in the part library, even though we've used that part before.
U20 and U3 are PV optocouplers whose output voltage is reversed.
That sort of error is insidious, when a pinout gets scrambled in the library.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
On a design of mine from mid-last year:
We had a fine pitch surface mount component, 0.5mm pad pitch, 0.15mm gap between the Cu pads. It was supposed to have soldermask-defined pads but somehow ended up with no soldermask at all.
Mfg yield was less than stellar. The CM was alerted to the problem and took "extra care" which resulted in adequate yield. These are expensive boards and in the short term we can rework them until they are all functional.
Moral: look at the soldermask layers during the design review. (I mean, go to the trouble of overlaying the Cu with the soldermask to look for this sort of issue.)
This was actually an interesting board design:
- microvolt level analog signals
- photon counting gizmo
- 50A, 4 phase DC/DC converter
- about a dozen different voltage rails.
- Multiple 25Gb/s differential signals.
- PCI express used for chip to chip comms.
- largest BGA device has over 2k balls.
- cooling solution used a custom heatpipe assembly.
Worked first time, and with the exception of that soldermask issue, wouldn't have needed any PCB fixes before production.
Allan
That gap is only 6 mil. I seem to recall most PCB fab houses can't print solder mask that thin. Am I mistaken? I suppose I am. You use the term "solder mask defined pads" which reminds me of BGAs. They come in *very* small pitch 0.4 and smaller. They are often SMD so I guess
0.15 mm solder mask is doable by the right people.I have a board with a 0.5 mm pitch QFP. I just measured the solder mask and it is 0.046 mm. I'll have to look at it to see if anything printed. They don't seem to have any trouble assembling these boards.
It's late and I'm beat today. I put in a raised flower bed. Maybe I'm getting silkscreen mixed up with solder mask. I do see the gap in solder mask around the pad is 4 mil on each side. That seems a bit large. I'm not sure where that number came from on this board. I want to say 2 mil is sufficient. That would be the mask to copper registration.
-- Rick C
I haven't done so many boards, so I don't recall making stupid layout mistakes other than one where I left a piece of power plane isolated from the rest and thereby cut the ground connection for an IC. Easy enough to fix with a jumper wire and it was a production run of just six boards.
But I did screw up the XYRS file on a board so all the chips on one side were assembled rotated. I had pin one marked, but they blindly followed the XYRS file without checking orientation. Not a big deal to fix, but I ended up with another fab house who checks everything. They know engineers aren't as good as they'd like to think. lol
-- Rick C
You're on the ball there. That thin strip of soldermask between the pads would have been too small to resolve with the type of soldermask we were using, so the CAD guy just removed it altogether to fix the design rule violation (and didn't tell anyone - gah!).
This was actually a power device with 55A mosfets in it, rather like a DrMOS. (It's a worry when power devices have the same PCB design rules as the BGAs.) In this case, the device manufacturer recommended soldermask-defined pads, meaning that the soldermask would go over the edge of the copper. This would have resulted in a strip of soldermask that wasn't too narrow for the process we were using. Had we followed the manufacturer's recommendations, we probably would have been ok.
Regards, Allan
-- Boris
Assigning 16 pin package to 14 pin chip.
-- Boris
On Wed, 25 May 2016 20:36:33 -0700, John Larkin Gave us:
"pinout scrambled in the library" sounds like 'sab-a-toogee'. Or a really dumb dude who seems to think editing an established part contained in an in-house library without saving it under a new moniker is an OK practice. He owes everybody a beer.
On 26 May 2016 03:48:05 GMT, Allan Herriman Gave us:
Better "moral". Use gold over nickel for stuff that small.
On Thu, 26 May 2016 08:11:12 -0400, DecadentLinuxUserNumeroUno Gave us: snip
Also looks like you are still using tin/lead process.
Must one of those "exempt" customers.
On Thu, 26 May 2016 08:03:34 -0400, Boris Mohar Gave us:
This is where the term "dead bug" positioning came from, not the proto guys.
Flip the chip over to make it also a mirror. Only drawback is you then have to connect to all the pins with hook-up jumpers. Unless you take a chance and 180 the entire lead frame as well. That is only good for the in-house test builds, not production.
On Thu, 26 May 2016 08:07:39 -0400, Boris Mohar Gave us:
Commonly known as a shift register of the brain.
It also smells...
Yeah, I suffer from no good reviews. I'll have someone look over my mechanical drawings, but no one else here really does circuit stuff.
George H.
Oh boy. When we had a layout guy he would go and "fix" things without telling me. The pcb would come back and I'd ask, "What happened to the extra, ground return line?" "Oh, I thought that was redundant, so I took it out."
Arghhhh!
George H.
The bad pinout is
3 41 2
which couldn't have happened in a new part creation. NO part has that pinout!
We did upgrade from a really old version of PADS to the current release, and a lot of things had to be done about libraries and such. And we've had three different layout people in the last three years.
That kluge is sort of ugly, but it's minor in the grand scheme of things. The board has a 1400 volt power supply, a couple of 1200 volt regulators, and makes two programmable 1500 volt p-p pulses, and the only oops was the opto pinout.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Den torsdag den 26. maj 2016 kl. 18.13.32 UTC+2 skrev John Larkin:
a minifit connector does
-Lasse
-Lasse
On Thu, 26 May 2016 09:23:47 -0700 (PDT), Lasse Langwadt Christensen Gave us:
He should have said that no "chip" type package has that pinout.
Connectors can get all kinds of odd configurations.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.