I am currently building a self contained power supply unit to power my pre and power amps. I have two banks of 4 large caps per channel and would be grateful for some advice on the following questions.......
1. I am going to add some high frequency filter caps (100uf and 100nf) after the main filter caps and before the power amp modules. Does the type of cap have a large effect on sound? If so what is best?
2. Where would they be best located? In the psu box, or in the amp enclosure?
It might be asking a bit much, but could any answers please have an explaination?
Thank you in advance.
1. I am going to add some high frequency filter caps (100uf and 100nf) after the main filter caps and before the power amp modules. Does the type of cap have a large effect on sound? If so what is best?
2. Where would they be best located? In the psu box, or in the amp enclosure?
It might be asking a bit much, but could any answers please have an explaination?
Thank you in advance.
That's a very nice can of worms you've opened, there! <grin>
1. You didn't actually ask whether or not you should add them. But the answer is probably "No". The explanations are here: http://www.diyaudio.com/forums/power-supplies/106648-paralleling-film-caps-electrolytic-caps.html . Actually, the 100uF might be OK, in the PSU. But the small caps should be right at the point of load, i.e. very close to the power supply pins of the individual components that amplify, with their grounded ends probably connected to the load ground (or maybe back to the star ground). And if they are to be outside of an amp enclosure, just leave them out. If you "must" use them that way, use lossy ones, or add a very small resistance in series with them (for film caps, for example).
2. See 1, above.
That could be a slightly different situation than I thought. If you want to install actual low-pass filters, that probably would not be a bad idea. But it's a little more difficult to do than I thought it would be, once you take into account the inductance and resistance of the wires, and the source and load impedances.
Just a cap or parallel caps to ground might be problematic (see previously linked thread). But you could use an actual low-pass filter topology, such as a simple LC filter, with a series inductor (or resistance) followed by a capacitor to ground. I don't know your maximum DC current requirement. But there are some good high-current inductors with very low DC resistance, that might work well. For example, go to Mouser Electronics - Electronic Component Distributor and do a part-number search for 542-23, which will show all of the Bourns/J.W.Miller 2300-series toroidal inductors. They range from 10 uH/20 Amps max/0.005 Ohms max to 1000 uH/2.4 Amps max/0.3 Ohms max, for $2.78 for qty 1 or $2.08 each for qty 10. They are 1.28 inches diameter max and 0.65 inches thick max.
To also filter out stray RF (Radio Frequency), it would be best to locate the filter very close to the amp module.
The f(-3dB) cutoff frequency of the LC filter would be f(-3dB) = 1/(2*Pi*sqrt(LC)). So, for example, if you picked L=10uH and f(-3dB)=30 kHz, then C would be 2.814 uF. If you pick f(-3dB) and L, then C = 1/(L*(2*Pi*f)^2), where ^2 denotes "squared", i.e. (2*Pi*f)^2 = 2*Pi*f*2*Pi*f.
Using L=10uH and two standard C values, 2.2 uF would give a cutoff frequency of about 67.9 kHz and 3.3 uF would give a cutoff frequency of about 27.7 kHz.
I just used 10uH as an example because in the Bourns/J.W.Miller 2300 series, it is the value with the highest max current and the lowest DC resistance. But you could use higher L values and correspondingly-lower C values, to get similar cutoff frequencies, e.g. 100uH and 0.22uF would give an f(-3dB) of about 33.9 kHz. But note that the 100uH inductor in the 2300-series has a max current of 7.0 Amps and a DC Resistance of 0.037 Ohms.
Unfortunately, if you use a capacitor that has a very low ESR (equivalent series resistance), such as a nice film cap, or a C0G/NPO ceramic, you will tend to get a large resonant peak ay around the cutoff frequency. So if you use a film cap, you would probably want to put it in series with a small-ish resistor, to flatten the peak. However, that will also affect the roll-off of the attenuation versus frequency, basically making it roll off at a higher cut-off frequency for higher resistances. And there might be a smaller peak, but below the cutoff frequency, which you probably don't want to fall in the audio range.
I have modelled it in LT-Spice, just now. It looks like a good bet would be to use 10 uH followed by the series combination to ground of a 10 Ohm resistor (1/4 Watt should be OK) and a 1 uF film capacitor. But it somewhat depends on the length of the wires and the equivalent series resistance of your power supply.
I used 25 nH per inch and 1 mOhm per inch, to model the power and ground wires from and back to the power supply, and between 0 and 1 Ohm to model the equivalent series resistance of the supply itself. The values I gave above for the filter components look pretty good for 24 inches of wiring, giving a -3dB (0.5 power) frequency of about 150 kHz, going down to about -16.5 dB (0.022 power) at 1 MHz. If you add a second identical filter stage in series, and make the resistors both 22 Ohms instead of 10 Ohms, then you get -3dB at around 120 kHz and better than -25dB dB (.0032 power) above about 1 MHz.
I think those were modeled with 0.5 Ohms as the equivalent series resistance of the power supply. It would be best to model it with your actual wire length (or the actual parasitics of the wire), and a range of power supply equivalent resistances (or the actual equivalent, if you can measure it). Or, maybe you could just try it with a few different resistance values for the filter(s) and see if there's any noticeable difference.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.