I received some PCBs from a board house and they have connectivity between the ground and Vcc layers. I populated one and plugged it in before I realized this. It immediately tripped the fuse on the board. The DRC checks do not kick back anything. I paid for electrical testing on the board. Should I have expected the board house to catch the shorted planes? I have found an anomaly on the PCB view of the board that will not fill in. The yellow layer is ground and the magenta is Vcc and this looks to me like it could explain the short but nothing shows up on the board checks and I cannot fill the gap.
Did you mean to post some visual aid here? I think as a basic user you should have permission to do so.
You are dealing with not only a freshly constructed virgin unit (never before known to have given satisfactory performance), but also an unproven basic design (no unit like it has ever been known to operate acceptably). These facts complicate run-of-the-mill troubleshooting efforts, which typically assume there is a single point of failure in a unit known to have been working at some time in the past. Just to be ABSOLUTELY CERTAIN (and avoid potential dead-ends in the troubleshooting process) . . . .
- Can you test an unpopulated, virgin, undefiled, board for an ohmic short between VCC and GND? An ohmmeter check is adequate. If the bare board has the short your next step is to put it under a magnifier to find the point(s) of violation, remove the shorts from the virgin board with your draftsman’s knife, then confirm that you have solved the problem. (There may be more than one such error on the board - some of us are remarkably consistent when it comes to repeating our mistakes. Don’t ask how I know this.)
And if you have access to a milli-ohmmeter, the process can move much faster.
The next step, of course, is to look at the Gerber files and make a guess at who is responsible for the error.
If there doesn’t seem to be a problem with the bare board’s fabrication . . . . Can you confirm the VCC-to-GND short on your populated board with an ohmmeter?
(Some components will sink large amounts of current if they are installed incorrectly - like a 14-pin DIP IC rotated 180 degrees, so pin 1 is soldered to pad 8 - giving the appearance of a shorted power rail when power is applied but registers a few ohms of resistance on the ohmmeter. Don’t ask how I know this.)
If the bare board is free of power rail shorts but the populated board has zero-ohm shorts, you’re probably looking for a solder bridge. Sometimes they are right there under your nose, but sometimes they hide behind pins or even underneath the body of an IC.
Another option why one would see shorts on a board is if a part is inserted the wrong way round (diodes, tantals, …) if this is the case then there should be a difference in measured “ohms” depending on what direction the measurement current flows. (plus it could easily be voltage or current dependent)
And of course: wrongly made symbol/footprint pairs. Be aware that some parts exist with exchanged power supply pins (i got bitten by this just a few weeks ago. The opamp i used is available in 2 versions that differ in only one character in the order number but have vcc and gnd exchanged on the package)
As @dchisholm suggested, check a bare board for shorts. Check several if they’re available. If no shorts are found, look for components installed rotated or laid-out mirrored.
Some years ago a client received the first fully-populated board of a brand new design. The R&D VP was on-site with the corporate board of directors and he insisted it be powered up for show-and-tell, despite objections by engineers saying they hadn’t had time to do even basic checkout. The VP insisted, so they plugged it in while standing a respectful distance away. Apparently the resulting loud bang and the column of smoke did impress the BoD, but not the way the VP intended.
The next day, after the smoke cleared, the engineers found that the footprint for a power FET in the voltage regulator had been mirrored during layout, resulting in a short across a high-current supply through the FET’s body diode. They were able to use other boards from the prototype run for checkout by removing the FET and wiring it in dead-bug style.
Years ago, in a manufacturing environement, I had a populated board for qa-testing. It had a wierd short. Decided to drill right down to the cause. Removed every component one by one. Still shorted. Nothing visible between the copper traces either. It turned out to be a short inside the actual board material itself. Eventhough it was simple FR2 the same thing could happen to FR4 and such. Anything is possible.
I did not see any confirmation as to whether fredmellon had tested a bare board. But there is another empirical technique for locating shorts. I have used it successfully I think more than once. I guess it can also work on populated boards.
Fundamentally copper shorts tend to have low resistance so using an ordinary ohmmeter can be difficult. So the idea is to apply a limited current between the two shorted rails; then use a digital multimeter on its most sensitive DC Volts scale to probe small mV voltage drops in the copper.
You want the current to be high enough to help find the short but low enough so that it will not burn the smallest trace or plated through hole it might be passing through. In my work I have used a test current of about 1 Amp.
The current can be controlled in one of several ways.
- Some good lab DC sources have a nice adjustable constant current limit.
- Another way to limit the current is to use an electronic constant current load in series with either the + or - connection to the DC source.
- Finally you could do this with a power resistor. If the resistor heats significantly the resistance might change; that will change the current and make it more difficult to track small differences in I*R voltage drop.
Finally…I have tried it…I do not recommend intentionally trying to burn away the short with high current.
I hate to say this, but you need to determine if there is a design problem or a fab problem.
Print the GERBERS! Those are what was fabbed. Print each layer large on a separate sheet and use a color highlighter to trace Vcc. Don’t just assume the layout is correct. Have a coworker or buddy look too. You may have to manually trace Gnd as well.
Also, the fab house pcb test would not identify a design error.
To find a physical short, solder a ps Gnd lead to 1 corner. Use a 100 ohm series R, with a few volts from the ps and a DVM to measure the voltage across the 100 ohms. Probe the Vcc line at various places to find the highest voltage.
The short will be a very low R compared to 100 ohms, so read the Dvm carefully.
Have you made any progress toward lengthening those shorts, or otherwise solving this problem?
Some folks would like to have the assurance that KiCAD did NOT create any short circuits in the process of generating Gerber files.
Good point. I had not thought of that…
When you are dealing with a shorted pcb, a resistance measurement is typically trying to resolve low values such as low milliohms. A two wire ohm meter has difficulty to do this because of the variable contributions of test lead wire and contact resistance. Even without that, the meter itself might be incapable of resolving anything better than 100 milliohms. For measuring low values, it is best to use a 4 wire method with 2 wires in the current path and 2 more for probing voltage. That way, you are not measuring the test lead and contact resistance. For this purpose we do not need absolute accuracy but we do want fine resolution. The maximum safe current will depend upon the board but I have used 1 Amp without difficulty. I think that even inexpensive DMMs can resolve 100 uV on the voltage scale. With 1 Amp of current you now have 0.1 milliohm resolution. This is much better than you can do with a 2 wire ohm meter alone. It allows you to detect the direction of current flow even in wide traces and perhaps copper planes. Of course if your voltmeter can resolve 10 uV, or if you can safely apply more current then your resolution improves even more.
… and there is always the time tested and proven method of attaching a really big power supply, then cranking it up! The smoke will quickly identify the location of the short.
The logic is that the board is no good so there is nothing to lose.
Stinks up the lab… I have tried it. I did fix a board by drilling out the misplaced via. I admire the simplicity of “classic” ExpressPCB but it is a basic tool which is relatively stupid. It did not prevent me from shorting copper plane of V+ to ground. The other thing is that you might burn the board in the wrong place…
If the board is truly written-off as scrap then any, and all, destructive approaches are permissible. But in some organizations, and under some circumstances, a board with a shorted trace is considered usable after manual rework. I wouldn’t have a lot of confidence in the long-term reliability of a board after clearing a short with a thermo-nuclear current pulse.
Board manufacturers should catch such shorted planes. If the shorts are not visible on the outer layers, there is a possibility of shorts happening within PCBs due to trapped moisture during the manufacturing processes itself. In such cases, an isolated in-circuit test is performed. Most manufacturers ignore such type of testing, though.
My thinking too. If it’s destroyed anyway, look for the smoke and smell. But there is a method a little bit non-destructive: Use a thermal camera. I have one from Flir that I can plug into my phone. It’s a really nice tool for all sorts of fault finding. Not more expensive than a decent DMM.
most manufacturers board tests are “open” tests, they probe all pads on each net to make sure they connect, a “short” test is much more time consuming, as you need to check every net against every other net after the open test succeeds,
If you do not have a low ohm multimeter, the easiest non destructive is a multimeter that can measure mV and a power supply with a current limit, drive say 1 amp into it, and measure along the net to find where it reads lowest (assuming short to ground and common ground between meter and power supply clip)
Of course I have designed all sorts of errors, but I have not had shorted boards so often. It seems to me that the thermal camera would work when there is a reasonable amount of current flow, and when the short has resistance which is noticeably higher than other portions of the overall shorted circuit. I guess that happens sometimes…maybe most of the time?
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.