In the previous post I described another failure in my Razer Mako speaker system. I found a defective electrolyte capacitor and I said it started to “short out”, which isn’t correct. What happens is almost the opposite, namely that the internal resistance in the capacitor starts to rise, creating heat dissipation (which almost burned me) which ultimately destroys the capacitor. Meanwhile, before it actually gets destroyed, it becomes less efficient at doing its job of smoothing out variations in the voltage applied to it, which I saw as increased ripple on the corresponding power line.
If you want to estimate the ESR in a capacitor, and if you have a signal generator and a scope handy, it’s extremely easy to do. All you need is a series resistor of a hundred or so ohms. You connect the signal generator in series with this resistor, then the capacitor you want to test between the resistor and ground. Hook up the scope probe over the capacitor leads.
Incidentally, you can do this measurement in circuit as well, i.e. without removing the capacitor from the board. Since a good capacitor will show up as a short circuit at the appropriate frequencies, the test won’t be influenced by other components in the circuit. If you keep the generator voltage low enough (100 mV is low enough), you won’t switch on any semiconductor junctions in the circuit either. You’ll still need the 220 ohm series resistor between the generator and the device you’re testing, of course1.
Set the signal generator to square wave and a fairly low output voltage, around 100 mV should be fine. At this voltage, polarity doesn’t matter. Set the frequency high enough that the capacitor with the capacitance it has ought to be effectively a short circuit. In other words, the lower the capacitance, the higher the frequency. In my case, I went for 10 kHz, but 100 kHz gave about the same results.
Finally, look at the trace. Does it look like a short circuit? If not, the ESR is probably too high. To be sure, do the same measurement with a known good capacitor and you’ll immediately see the difference.
To make the comparison more clear, I hooked up the capacitor I replaced in the Mako (yellow trace) and a new capacitor (blue trace) at the same time. The difference is obvious, to put it mildly. The defective capacitor is 330 µF, while the known good capacitor was 220 µF, which shouldn’t make much of a difference.
The blue trace shows what is effectively a short circuit, which is what you’d expect from a 220 µF capacitor at 10 kHz of input signal. The yellow trace is far from a short circuit, due to the internal resistance.
If I disconnect the known good capacitor, we get the input signal from the generator as a reference (blue trace):
As you can see from the slope of the yellow trace, the capacitor is still working as a capacitor, but has a pretty large internal series resistance (ESR), which explains why it’s ineffective and also why it’s heating up.
From the measured voltages of the traces (see Vpp below the traces), you can calculate the ESR value, but I didn’t bother. The pictures tell the story well enough.