Hi - I've often noticed small round solder dots on traces on PCBs. They aren't a part of a component's footprint, and they don't appear to be vias as they don't have holes going through them. For example, in this picture of an XBox's motherboard:
they're very visible (for example the circled "D0" points). Can somebody tell me what these are for? My guess has always been that they're a sort of test pad - but why would test pads be needed on a mass produced PCB? I mean my understanding is that these days when a board like this goes bad it is replaced, not repaired. My only guess is that they're for testing and/or programming in the factory. Can anybody shed some light on this for me? Thanks,
Are they in sets of 3 widely placed across the board? If so, they are fiducials. These are alignment references that the auto placement machines use to get the solder mask and component placement precisely aligned with both the position and rotation of the board. The ones labeled D0 appear to be vias that have filled with solder from wave soldering. They are not fiducials, because traces run through them.
I'm pretty sure those are testpads. Testpads are not only for troubleshooting, but also can be used to flag boards as bad prior to function or final acceptance tests. They can also be used to look for shorts between nets on bare boards.
A few years ago I worked for a small design group within a large company we are all familiar with. Our group designed Intel architecture SBC's for CompactPCI chassis. The test guys always wanted us to have test points on every single net (even both sides of a series 10-Ohm resistor). The layout guy we had hated testpoints and thought they were stupid (this is because bring-up and debug and testing were not part of his job). Anything that made routing a board take longer was stupid, in his view. We had to watch him carefully. ;-)
Usually the test guys would let us put the test pads on only one side of nets with small series resistors. And then there would always be some nets in densely-routed areas where there just wasn't any way to place all the test pads without creating long stubs. For high-speed nets, I would usually veto long stubs.
Access to points on the board which are not otherwise brought out can improve testability. For example, on an analog circuit, you might want to probe an internally generated voltage to make sure it is within specs rather than trying to infer that it is from the outside. You might want access to an internal node on a digital circuit because it otherwise would take too long to test. The board can be put in a test jig with probes (eg. pogo pins) and tested without the cost and board area of connectors or the routing required to get different nodes to the connector.
This kind of thing is probably more the rule than the exception in complex mass-produced boards.
Some things don't lend themselves to easy testing- DIPswitches, for example. If the board is 100% electrically tested and you use a good quality DIP switch, and inspect you might be willing to assume that things are okay if the board tests okay with the switches in one position. However, there are many, many faults that this will not catch, and will only show up by switching a particular switch from one position to another. There may be faults that could show up only if combinations of switches are put in a particular set of positions. It's clearly impractical to test all combinations of, say, 2
8-position DIPswitches, so brute force is out. Last design I did with those d*mn things (customer spec'd BTW), I tested each switch in both positions by forcing the operator to go through a programmed test sequence flipping one at a time. More than good enough for the application and it caught an almost invisible QFP-80 pin solder short in my prototype.
Best regards, Spehro Pefhany
"it\'s the network..." "The Journey is the reward"
firstname.lastname@example.org Info for manufacturers: http://www.trexon.com