[noonespecial]

I also had done more revisions of the board and spent way more money than I initially planned for... even though in hindsight, having only three revisions for a fully functioning, moderately complex design, with no previous experience working on something like this, probably is not that bad.

Are you kidding? That falls into the realm of superhuman. I've seen 20 year veterans blow a 100 grand and get half that far. I think even Woz would do a double-take.

[bsder]

> Are you kidding? That falls into the realm of superhuman. I've seen 20 year veterans blow a 100 grand and get half that far. I think even Woz would do a double-take.

A high density BGA through an OSH Park board with a homebrew reflow oven and only 3 revs and a stainless stencil on only the last pass? Yeah, that's really quite good.

You practically can't avoid making a mistake on a PCB of that complexity. There are simply too many details buried on too many datasheets and one of them is going to drop.

If you have the time, the best thing I have found for catching mistakes is to have a design review. Even if the person doesn't understand electronics, the fact that you have to walk through every part and explain it to somebody means that you will catch some really stupid things.

However, if you're good and the board isn't too expensive, it's often cheaper just to fab the board. You're time is worth something, and you can be designing something else while the PCB is being fabbed.

[laobeylu]

That's actually sort of what I did. I didn't explain it to someone else, but went through it in my head, talking to myself explaining the design. Explaining it to someone else must be even better!

[noonespecial]

Or just a rubber duck.

https://en.wikipedia.org/wiki/Rubber_duck_debugging

Personally I use a cat. Its Steve-Jobs-like attitude seems ideally suited for the task.

[dbcurtis]

Yes, great work! I think of PCB design like golf -- low scores are better, but nobody hits a hole-in-one every time they tee up. 3 revs is shooting par for even simple designs, much less a complex RF design like this.

Having someone else look over the design is always useful. A healthy distrust of datasheets is always useful. But at some point, you have to ask yourself you have reduced the risk of errors enough that a couple of rework wires with get you through debug, at which point just go fab it. Blue wires tack-soldered on are a fact of life in prototyping. It is easy to obsess over PCB designs, but "fail fast" applies here, too.