I am working on a not exactly simple design at the moment. For the last several days I have thought to myself, “I’ve got this, it’s as good as DONE!” (<<< yelling at myself should be allowed)
I keep seeing that it seems I am learning that there are better methods to handle the challenges in the PCB design of my project.
The last two problems I ran into:
Silkscreen was on copper.
Dumbass me forgot to make room for the mounting holes.
And, I guess:
Finding a good method to make imperial and metric parts work together.
What I did in college, was to print a “test” paper copy of everything before it got printed for submission. The brain can’t see the mistakes while made on the monitor, but it can see them when it is “different” and on paper.
At the moment, I’m using the 3D veiwer to accomplish the same effect. And, it is working quite well; much better then I expected.
I’ve worked with a couple of desings that required a second spin of the board, and these were done by people that were experts and had done it for 20+ years. At what point is “giving it a spin” to have it “in hand” to check for errors worth it?
Is there a “check list” that you go through, to make a “best effort” for the first spin?
Depends on how fast you need the project to proceed really, how many prototypes/spins you can afford, etc.pp.
So far I only did one spin each of my layouts and usually find 3-4 errors right away, no matter what.
Most of them are fixable though and as it’s prototypes anyway, the alternative would have been to build those circuits up on proto-boards… so even if I wind up with SOT233 turned 120 deg on the footprint or having to run 2-3 wires across a board I’m still good
A lot depends on how sure you are that the functional spec is final. If you suspect that the user might make changes, getting the pcb out early and accepting minor errors wins
[quote=“Sprig, post:1, topic:7040”]
The last two problems I ran into:[/quote]
The examples you cite are common errors. Organizations that have their feces amalgamated typically have business processes requiring some kind of review, walk-through, or peer inspection of your work. The effectiveness of these systems depends on how rigorously they are enforced and how conscientious the staff is about performing them.
This particular problem continues to afflict me after years of practice. It’s been a few years since I actually submitted a board with silk on copper but I have spent hours checking and re-checking Gerbers to make sure none of the silk lands on bare copper.
The KiCAD DRC does not check for silkscreen violations - something I consider a very basic shortcoming. Every time you update a footprint, or re-import a netlist with “Exchange Footprints” in effect, the silk reverts to the size and location specified in the footprint library. Of course (as Edsel Murphy observed decades ago) the locations specified in the library fall squarely on top of adjacent components’ pads. That’s one reason the silkscreen editing is one of the last tasks I do on a board.
(Some of the board fabs actually check for this fault, and either remove silk from bare copper - perhaps without telling you - or put the job on HOLD and alert you.)
Another common mistake, especially for beginners. In exchange for a serving of my favorite malt beverage, you can be assured that neither your supervisor nor your co-workers will hear of this incident from me. A somewhat related error - pertaining to the board’s mechanical layout - is creating an outline that is the mirror image of what you intend.
I think this has been mentioned in another thread in the last few months. My superannuated brain doesn’t comprehend the basic problem.
I commonly use 1:1 check prints to test for physical fit. Sometimes it takes some fumbling to make a printer produce a reasonably accurate 1:1 scale. (@Joan_Sparky is going to chime in here and tell me to use 3-D modeling for this. The traditional aphorism warns you about old dogs and new tricks. Besides, I see too many threads on here from people who have difficulty getting the 3-D features to behave properly.)
I also make liberal use of the Gerber viewer. (I like the geda program “gerbv” better than KiCAD’s “GerbView”). I make several passes across the board, usually displaying two particular layers at a time. It took some experimentation and practice to develop a set of color assignments, and layer loading order, that makes problems show up in a way that grabs my visual attention.
Many incarnations ago an old machinist told me that the first time is just to get a feel for the job and whether it can be done, the second time is for practice, and the third time is to confirm that you got it right the second time. Then you are ready to do it “for the record”.
This might come from the “problem” that tracks always snap to the center of pads. (you might get short tracks inside pads. or a short 45° a bit outside the pad. Some people do not like that.)
In stable there currently is no real way around that. In nightly one can use the shift key while laying a track to disable snapping.)
OK, I consider those little jogs and wiggles an aesthetic irritation at worst, not a real problem. Beauty is in the eye of the beholder. With practice, you learn how to minimize the visual effect of the metric/imperial adjustments even without that shift-key trick.
With the autorouter, it gets a little wonky when the two parts are on different grids (as they have to be as the two parts have different scales). This can create all sorts of oddball track pieces; or the “shotgun pattern” effect I mentioned earlier.
It also mess me with as I don’t quite have a feel for the clearances of metric parts. This affects how many trials it takes to get zones to fill the way I want.
If they can spare the time, let a colleague take a close look at your design, even if this is not standard practice for your company. From my experience you’ll catch quite a few mistakes when the co-worker starts asking questions like “Is this intentional?” or “Why did you do that this way?”
