This is a little extension of my tests on the current amplifier I have been working on. The amplifier was previously used in the Experimenting with Polymer Capacitors design challenge and during the testing I noticed that the polymer capacitors seemed to be running at a slightly higher temperature in comparison to the electrolytic capacitors I was comparing them to.
A suggestion by jw0752 was to take temperature measurements of the capacitors with them electrically isolated from the amplifier supply circuit, this should then determine if the heat differential was being created from the internal impedance of the capacitors or from the heat radiated from the surrounding components.
As the PCBs are now actually in quite a state, due to the continual de-soldering and soldering of the different capacitors, isolating them from the supply rails was relatively easy. Temperature measurements were therefore successfully carried out and showed the Polymer capacitors still running a little warmer then the electrolytics.
However, with the opportunity of the board in this state, I could not resist running the test measurements and then comparing them to the 220uF Polymer capacitor test results. The table below shows the results of those measurements.
Put into graphic format;
The average voltage for both the positive and negative rails is slightly higher without the de-coupling capacitors installed. For the positive rail, both DC fluctuation and AC ripple are significantly lower when the de-coupling capacitor is installed. This would be expected. The same trend is not seen for the negative rail, where only the AC ripple is seen to be lower with a de-coupling capacitor installed.
Looking ore closely at the DC Fluctuation;
As the load is increased in the lefthand plot, the fluctuation remains relatively constant up to the 3V input, around 1.8A output. After this the 220uF capacitor starts to have an effect and shows a lower fluctuation level. At 5V input, the negative rail shows the same level of fluctuation with or without the capacitor.
The righthand plot shows the fluctuation as the frequency is increased. The results are a little sporadic, with the fluctuation on the negative rail generally lowest without a de-coupling capacitor installed. At the highest frequency of 20kHz, the fluctuation is reduced with the 220uF capacitor installed, but this is the only result that really matches expectations.
The same set of test are then completed whilst measuring AC ripple.
The effects on load change are not as apparent as they were for the DC fluctuation, with the majority of the readings similar with or without a capacitor installed. At the highest load, about 3A output, the 220uF capacitor is starting to have an effect, more on the positive rail than the negative.
The plot for the frequency changes, shows better performance with a 220uF capacitor installed at the 10kHz and 20kHz frequencies, with no real differences seen at the lower frequencies.
Comparing the gain of the amplifier, it can be seen that a slightly lower average gain is seen when a de-coupling capacitor is installed. However, there is much less deviation in gain across the range of tests, when the capacitor is installed. The minimum gain seen is the same for both sets of tests and in actual fact the same 4V Input tests shows the lowest gain of 1.965 with or without a capacitor. With a capacitor installed, the maximum gain only ever reaches 2.020 for the 2V input test, which is different when no de-coupling capacitor is fitted than has the highest gain of 2.068 at the 5V input test, which represents the maximum load on the amplifier.
A further set of temperature measurements has been carried out to cater for the different current output on the amplifier. I run these tests with both amplifier boards in operation, one with the polymer capacitors in and the other with the electrolytics. The load resistors measure 1.004275 Ohms and 1.001264 Ohms, but one amplifier always puts out about 2.50A and the other 2.35A. This means that the amplifier runs slightly cooler on one board. I can only attribute the difference, to slightly different gain in the two amplifiers as swapping over the load resistors makes little difference.
In the first set of tests, the polymer capacitors were installed in the board operating at lower load but they operated about 4 Deg C higher than the electrolytic capacitors. For some reason, I didn't think to swap the capacitors over between the two boards to verify if the temperature went up, but I have now done this to provide a little extra data.
The picture on the left shows the original temperature measurements, the picture on the right, is with the capacitors swapped over between the boards. The table below summarises the results.
It can be seen that for both boards, the polymer capacitors run at a higher temperature than the electrolytics, irrespective of the amplifier temperature. It should also be remembered that the test boards are open to the air, mounted inside a case, the temperature differentials would likely be even higher. Having said that, they aren't at a level that are beyond the specification of the capacitors.
I still am not too happy with the performance of the thermal imaging camera. You can see in both pictures, that the spot temperature on the amplifier is actually higher than the scale to the side of the picture. I am not sure what the cause of this is. There are different the of surfaces on the components and to be technically correct, the emissivity of them will be different, which could cause slight in accuracies, but I am not sure that this is why the scale would be recorded as lower.
If I can find some way to attach thermocouples to the relevant devices, and find enough thermocouples, I may look to make a comparative measurement using a different methodology. It would also be an ideal test for the DAQ970A, that I have been luck enough to be selected to roadtest.