I am extending my test results blog for the experimenting with polymer capacitors design challenge to include the test results for the 220uF capacitors. As the competition is officially over, I will keep this in my personal blog, but the results may be of interest to others.
Only two sets of 220uF capacitors have been tested. Panasonic Polymer capacitors from their SEPC range, and Nichicon electrolytic 220uF capacitors from their VY range. The Panasonic capacitors were rated at 16V and the electrolytic at 25V. The full details of the capacitors used are in the table below. They have not been removed from the amplifier boards, so there are no post testing capacitance and ESR measurements.
Other major differences are the endurance at the rated temperature, which is much higher for the polymer capacitor and the ESR, which is 10 times lower for the polymer capacitor, in comparison to the electrolytic. This increased performance in these aspects comes at a higher price though.
Fitting the Panasonic 220uF polymer capacitors was a bit of a squeeze. The lead pitch prevents them from sitting flat on the PCB and the physical spacing means that they are touching. This was a similar finding when installing the Rubycon 100uF capacitors. The Nichicon 220uF electrolytic capacitors are taller, but with a smaller diameter, and fitted easily into the space on the PCB.
The same set of tests were conducted as previously carried out on the lower valued capacitors. The two tables below summarise the results of the tests.
Plotted out as a comparison, it can be seen that there is virtually no difference between the supply rail voltage levels. Less fluctuation and AC ripple are seen on the positive voltage rail for the polymer capacitor, but this is then reversed on the negative rail, where the polymer capacitor shows increased fluctuation and AC ripple levels.
The plot below on the left, shows the increase in the DC fluctuation as the load on the amplifier is increased. The results are very similar across the load range except at the 4V input, equivalent to around 2.50A output. At this load, the polymer capacitor seems to out perform the electrolytic, having a lower DC fluctuation. In terms of frequency, the DC fluctuation is highest at lower frequencies. The performance across the frequency range tested is similar for both capacitors.
A similar pattern is seen with the AC ripple readings, but with these, there seems to be more consistency across the range of tests in comparison to the DC fluctuation readings.
Below are typical plots of the waveforms captured.
The first plot shows the input and output waveforms along with the DC supply rails. The load or this particular plot is around 2.50A. A reasonable level of fluctuation can be seen in the DC supply rails, but the output waveform is clean and shows no distortion.
As the load is increased to 3A, more fluctuation can be seen in the DC supply lines and distortion is starting to appear in the negative peak of the output waveform.
The next plot is an example of the AC ripple on the supply rails captured at 10kHz input signal to the amplifier and a 1.55A load on the output.
In previous tests with lower capacitance values, I have found that the polymer capacitors appear to operate at a higher temperature than the electrolytic equivalents. The two thermal images below reveal this same trend in the measurements.
In the 2.5A load tests, the polymer capacitors have a temperature around 5 Deg C above the electrolytic capacitors. It was whilst I was writing about the heatsink performance, I made a comment about not wanting the heatsinks too close to the other components due to radiated temperature affecting them. This got me thinking and I noted that the polymer capacitors are a complete metal can where as the electrolytic capacitors have a black / grey plastic covering. Bare metal will soak up more heat, so I decided to run an extra set of load tests with a heat shield between the capacitors and the amplifier.
The results are a bit mixed. The SP1 stamp marks the C8 capacitor on the negative rail. The temperature of this capacitor is even across the polymer and electrolytic devices at around 32 Deg C.
SP2 marks the C6 capacitor on the positive rail. On this device the polymer is running around 5 Deg C above the temperature of the electrolytic, the same result as the previous tests. I will look to repeat this test with a better heat shield and I will see if I can attach a thermocouple direct to the capacitors to improve the accuracy of the temperature measurements.
The gain was calculated as a ratio of the input and output voltages. Nominally, it is expected to be 2.0 based upon the resistor network installed. The gain is displayed in the table below, along with basic statistical values.
The polymer capacitor seems to slightly out perform the electrolytic capacitor. The spread of results is slightly tighter and the average is closer to the nominal value for the polymer capacitor.
Performance relative to capacitor size
In these next plots, I have compared the different value capacitors used to one another, to see how the actual capacitance value has affected the performance of the supply rails. This first set of data compares the different values of Panasonic polymer capacitors. The test data for the 10uF capacitance value is used as the baseline.
Average rail voltage for both positive and negative follows the same pattern. Unexpectedly, both the 47uF and 100uF capacitors have a lower average rail voltage than the 10uF, but the 220uF capacitor produced the highest average voltage. DC fluctuation was seen to be highest for the 100uF capacitor, the 47uF and 220uF capacitors has similar performance. For the AC ripple measurements, the 10uF capacitor produced the worst results. The results for the other capacitors were comparable, with none of them producing a significantly better result.
The plots above display the DC fluctuation for each capacitor as the load is increased. At lower loads, there is not much between the different capacitors. At the highest load, the largest 220uF capacitor is seen to reduce the fluctuation much better than the lower valued capacitors. Blips in the data are seen in the middle of both plots, with the 10uF capacitor showing the least fluctuation with the other capacitors more comparable to one another.
With the load kept the same, but the frequency varied, a different response can be seen. At lower frequencies, the 47uF seems to have the better performance, but the fluctuation is higher for all the capacitors at the lowest frequency. In the middle of the plot, a mixed set of results are again observed. At the higher frequencies, more consistency is starting to be seen, with the 10uF capacitor showing the most fluctuation and the other capacitors lower but comparable to one another.
The AC ripple seen during load changes, follows a reasonably uniform data plot with an expected increase in AC ripple as the load is increased. Significant differences are seen at the lowest load, with the 47uF and 220uF capacitors both displaying a much lower ripple value. At the 4V input, about 2.50A output, the 47uF capacitor shows a break from the trend, with a much higher AC ripple than any of the other capacitors.
With frequency change, the 47uF capacitor shows a constant AC ripple across the data plot. The other capacitors have a much higher AC ripple at low frequency that reduces as the frequency is increased. At the higher frequency the AC ripple creeps up for the 10uF and 47uF capacitors. The 100uF and 220uF capacitors have a slightly lower AC ripple at the highest frequency.
The table below shows the differences in amplifier gain for the various sizes of polymer capacitors.
The 10uF produces an average gain, closest to the nominal value of 2.0. However, a much higher spread of gain is seen for this capacitor in comparison to the others. At the other end, the 220uF capacitor statistically shows the least amount of spread, but there is no real significant difference to the 100uF capacitor.
The next set of data represents the comparison of the different values of electrolytic capacitors, again, the 10uF capacitor forms the baseline for the comparison.
The averaging of the results produces sporadic behaviour. In terms of voltage rail values, the 10uF capacitor produces the best results overall. On the negative rail, the 47uF, 100uF and 220uF all recorded lower voltage levels than the 10uF, which would not be expected. The DC fluctuation levels are also odd and not consistent with expectations. For both rails the 100uF capacitor was found to reduce the DC fluctuation, but the 220uF capacitor showed higher levels. When measuring the AC ripple, it was the 47uF capacitor that had the lowest ripple values, but both the 100uF and 220uF showed higher levels of AC ripple.
Looking more closely at the individual test results, a general pattern emerges of increasing DC fluctuation as load is increased. In the majority of cases, the 10uF and 220uF capacitors saw higher levels of DC fluctuation as the load was increased. At lower loads, the DC Fluctuation was more consistent across all the capacitors.
Looking at the DC fluctuation with frequency variation, the usual behaviour is seen with a higher DC fluctuation at the lowest frequency for all the capacitor that then drops off as the frequency increases. What is intriguing is that as the frequency is increased, the 10uF capacitor shows the lowest levels of DC fluctuation, with the other three capacitors having results similar to one another. The electrolytic 10uF capacitor was the first capacitor tested, although the tests were repeated for this after I found that the polarity was reversed on C8. None the less, it would seem that there could be errors made in the test methods for this capacitor.
If I had the time, it would probably be wise to re-run the tests on the 10uF capacitor to verify the results. The PCBs are starting to suffer now though with all the soldering carried out on them.
Moving on to the AC ripple measurements, the results for the load changes seem to be opposite to expectations. Although the AC ripple does increase as the load increases, it seems to increase more for the higher value capacitors, with the 10uF consistency showing lower levels of AC ripple.
The frequency variation tests, seem a little more predictable. Again, higher levels of AC ripple are seen at the lower frequency on all the capacitors, although the 10uF is lower than the 100uF and 220uF capacitors. As the frequency is increased the levels of AC ripple drop more on the three larger capacitors, leaving the 10uF showing much higher levels. At the 20kHz level, the 100uF and 220uF capacitors start to produce AC ripple levels lower then the 47uF capacitor.
Comparison of the amplifier gain with electrolytic capacitors installed does not follow the same pattern as the polymer capacitors. The 220uF and 10uF capacitors produce similar results to one another. The average gain is the same and the spread of the results is similar, with the 220uF capacitor showing a slightly lower spread.
The 47uF and 100uF capacitors showed a lower spread of results than the other two, although the average for the 100uF capacitor was similar. The average gain for the 47uF has a negative deviation, instead of the positive deviation shown by the other capacitors.
The test results for the polymer capacitors seem to show better consistency and are closer to the expectations for the results. The 220uF capacitor tended to show a better performance than the lower value polymer capacitors. However, temperature measurements showed that it still appeared to operate at a higher temperature than an equivalent electrolytic capacitor. An attempt was made to shield the capacitor from the heat generated by the amplifier to ascertain if the metal can of the polymer capacitor was sensitive to the heat radiated by the amplifier, but the results were mixed and inconclusive.
The results for the electrolytic capacitors were mixed and the 220uF did not come out as a clear winner. This may be due to experimental errors and / or damage to the PCBs affecting the results that were not present when the smaller capacitors were tested.
Ideally, I feel as though I should re-run some of the tests, but I also want to move on and experiment with parallel / serial operation of the amplifiers.