A few weeks ago I was asked to look at a telecom PCB fresh out of production with a small problem–the inventory EEPROM would not load the card revision information, and a new EEPROM did not solve the problem.
I hunted up a card extender, blew off the dust, plugged it into one of our test racks, and set the system control card to repeatedly query the Eprom. I suspected the problem was one of the glue logic devices controlling the EEPROM, but rather than recommend a shotgun replacement of all the glue logic I decided to take a closer look first.
Good thing, shotgunning would not have worked.
First clue – the scope showed the EEPROM serial data pin was stuck at a logic high (Vdd power supply) voltage in spite of the control and clock signals to the Eprom looking normal. Since the Eprom had already been replaced, was there a short from this pin to Vdd?
Yes, the EEPROM output did measure about 0.1 ohms to Vdd on a DVM. This 0.1 ohms is the normal reading of the probe wires, which sets the low-resistance measurement limit of a DVM, so obviously there was a ‘dead short’ (approaching 0 ohms) somewhere along the length of the trace that was connecting the output pin to a few other logic device inputs and tri-state outputs. Maybe one of them was shorted to Vdd inside a logic device?
Unlikely, since a defective device input or output when shorted to Vdd or ground usually has a least a few ohms of measured resistance. But just to make sure, I lifted all the IC pins that connected to the trace, including the new Eprom output. This is very easy to do with surface mount devices, wick the solder off and pry the pin off the pad slightly with sharp tweezers while hot under the soldering iron. Careful, don’t overheat and lift the copper pad off the board!
With all the pins lifted off the trace I could still measure a dead short between the trace to Vdd. This exonerated all the attached devices, the short was hidden on the pcb somewhere along the branching run of that trace. Visual inspection under a microscope showed nothing obvious. Perhaps it was a solder bridge lurking where an external portion of the trace ran underneath an IC package.Or perhaps was invisible inside the pcb.
The short could be anywhere along the branches of the trace, and the limitations of the DVM in ohmmeter mode made it useless for resolving below 100 milliohms. Now it was time to get out the heavy artillery. Not a current probe, you cannot clamp a current probe around a pcb trace. But with the following procedure the physical location of a dead short is relatively easy to locate.
The trick is to apply a controlled current from a current-limited lab power supply through the short, in this case between the shorted trace and Vdd, and look where the current is going. The current flow will be directly through the short and the path leading to that short, but will not flow along any other path of the isolated trace (all logic device pins still lifted off their pads).Of course we cannot actually see where the current flows, but we can measure the result of the current flow to get that information.
The copper trace is not a perfect conductor, it has many milliohms of resistance depending on its thickness, width, and length. Forcing 1 amp through the short will produce 1 millivolt drop for every milliohm of trace resistance (Ohm’s Law from Electronics 101 ).
Even the cheapest modern DVM from the hardware store can resolve down to 0.1 millivolt, plenty good enough for this. The current level should be adjusted for as low as possible to still get meaningful voltage readings, too much current can burn open a very thin trace.On the other hand, if trying to locate a short that is between the power planes, 2 or 3 amps may be needed due to the much lower resistance of power planes.
Even when traces are buried on internal layers, measurement points are where the trace is brought to outer layers through vias. It helps if you know (from the pcb design artwork) or can deduce the internal paths of the trace layout,
Inject the current – find convenient places to solder wires to the shorted trace and Vdd. It doesn’t really matter where the current starts, it will eventually end up at the short. A good place to solder a wire to Vdd is at one end of a decoupling capacitor. Solder a second Vdd wire to another point nearby. Another wire soldered to a via or component of the shorted trace, and you’re in business.
Preset a lab power supply to about 0.4 volts open circuit, then short the power supply output and set the current limit to 200 mA. You can always increase the current later if needed. 0.4 volts is a safe starting point in the event that you overlooked a device and did not lift its pin from the trace; 0.4 volts will not bias input clamp diodes into conduction should the short suddenly disappear. Then connect the power supply to the trace and one of the Vdd test wires soldered onto the pcb.
Set the DVM to the millivolt range and connect one probe to the other Vdd wire, this avoids measuring voltage dropped along the Vdd wire from the supply. Touch the other DVM probe to the wire that is soldered to the trace. You will read several millivolts, this is the voltage dropped along the trace as the current flows to the short.
Now move the trace probe to the next access point (via or IC pad) and note the measured voltage drop. If the drop reduces, you have moved the probe closer to the short. If the drop does not change, there is no current flowing along the branch whose two points you have just measured.
Figure 1 below depicts a typical pcb trace routing, portions of the trace can be on inner or external layers, vias bring inner layer trace portions to the external layers for connection to surface mounted device pads. For clarity most devices are not shown. Wires to the current-limited lab power supply and DVM can be soldered to either vias or components that are connected to the shorted trace and Vdd.
Figure 1: Typical trace route and via measurement points
With one DVM probe connected to Vdd through the connected wire, move the other probe around to the vias of the shorted trace. Note the voltage readings at each measurement point.
Starting at the resistor 60 mV is measured where the supply wire is attached. Via A measures 50 mV. Thus 10 mV is dropped along the trace between the resistor and via A, indicating that current is flowing through this portion of the trace on its way to the short.
Via B measures 30 mV. If the trace is already on a surface layer and via B is not part of the trace, pin 1 of the IC is a good point to measure instead. Hmm, the trace runs underneath the IC on it’s way to via C. Could there be a solder bridge between the trace and where it runs between the IC pins? Via C measures the same 30 mV as via B or IC pin 1, obviously no voltage is dropped along this portion. Thus no current flows here, so no short under the IC.
Via D is 10 mV (getting closer!), but via E is even lower at 2 mV. The current must be flowing to via E. The 10 mV at via D happens because the trace forms a resistive voltage divider at the junction between the branch to via D and the section between via B and via E. The same effect causes vias B and C to measure 30 mV.
Via F is the same 2 mV as via E. No current is flowing along the trace between via E and via F, so the short is not at or beyond via F.
That leaves via E as the prime suspect. The current is draining into via E and no farther.
In my case, I had the luxury of an X-ray machine to see what might be lurking inside the pcb. Figure 2 below confirms the culprit–deep inside the pcb there is a small conductive particle touching the copper barrel of the via and shorting it to the internal Vdd plane layer near the upper center of the image. The thick object running to the left is a wire that I soldered to the via to confirm it was the location of the short prior to making the X-ray image. The cure was to use a small drill bit in a pin vise to drill out the via barrel and jumper the trace connections around the via using AWG 30 Kynar insulated wire.
Even without an X-ray machine, knowing the physical location of the short and probable cause enables one can drill out vias or cut external layer traces to isolate the short.
Figure 2: X-ray image of a pcb internal short. A small conductive particle upper center protrudes from an internal plane layer and contacts the barrel of the via at center. The wide object running to the left is a wire that was soldered to the via for final confirmation of the short location prior to making this X-ray image. (Courtesy of ICBS / EO Networks Division)
Glen Chenier is a design consultant based in Allen TX. His article has also been published on EETimes.