Wide-band RF filtering for mains
Since Ostripper asked about mains filtration on the RF & Audio thread, here are some notes and ideas. I haven't built any of this yet - if you decide to, I'll be interested in any results - photos and listening reports. Be careful with this stuff, its at mains potential.
Pre-built mains filters are of some use in audio, but they're not really optimised for high-end sound. For audio source components, the draw is normally under 100W and so the current (assuming here 230V supply) is under half an amp. Few if any off the shelf mains filters are designed for such low currents.
Second, the commercial filters assume that we wish to stop conducted interference getting out as much as prevent muck getting in. The internal noise would normally only come from the rectifiers in a linear power supply (SMPSUs are a totally different matter). Once we've snubbed them (you have done that already right, or are using soft-recovery or schottky diodes?), with a linear PSU in an audio application our primary concern is against interference getting in.
Mains filter elements
A Mains filter consists of one or more common-mode chokes and a handful of capacitors arranged in fairly standard configurations. A common-mode choke is one in which the current flows equally through two mutually opposed windings. Thus current which comes in and goes out again through the choke sees negligible inductance. That's the current which goes into the transformer and powers our circuit - we don't want that impeded very much. But current that comes in and leaves by a different route - generally on ground interconnections to other units - sees the full inductance in series with it and hence gets attenuated. Its those common mode currents which can have an impact on the sound we hear as when unbalanced interconnects are used, they induce noise voltages which directly add to our wanted signals.
If you look at a catalogue of such filters, you'll notice they come in various current ratings, and often the lowest one is 3A. For audio source components, we're rarely drawing 700W so it'd be nice to have a lower rating. That's where building your own becomes attractive. As the current rating falls, so the common mode choke's inductance rises. A higher inductance ought to give us better filtering, all other things being equal.
In reality though, all other things aren't so equal - higher inductance usually means more windings and more wire area means higher self-capacitance. Thus while the rejection is better at lower frequencies for a lower current rated choke, the frequency where it effectively stops working entirely (due to the stray capacitance) reduces too. Have a look at the Murata PLA10 family of curves - on the right you'll notice all the chokes follow the same line. At those frequencies they look like capacitors to the interference, which consequently sails right through, proportional to frequency.
There's a fairly simple fix for this problem - use more than one common-mode choke in series. Some commercial blocks already do this. Commercial blocks though also usually incorporate Y capacitors. There are two places capacitors are connected - directly between line and neutral (called X) and between line/neutral and earth (called Y). Y capacitors are potentially a problem for audio because they connect the mains (both sides) to our local earth at RF frequencies. [That is assuming you do have an earth - if you're building a class II source then you won't, so having Y caps isn't a problem.] If you are using the earth, then you'll need to think carefully whether Y caps will actually help you.
Rolling your own filter
A DIY filter is certainly going to give the best potential results, if you know how to design and build it. My idea here is to build one which has three stages - each stage optimised for a different frequency band. My hand-drawn sketch shows a possible design.
The highest frequency band (say up to 100MHz or so) is best addressed with series air-core chokes (L1 & L2 in the sketch). These have the lowest self-capacitance (hence highest SRF - self resonant frequency). There's no need to have these chokes arranged in common mode configuration as they can't ever saturate. Air core chokes with high SRF can't have very high inductance - say single digit microhenries in practice. They'll be rather bulky as the windings need spacing out but they can (and should) be wound with thin wire as we actively want them to be lossy at HF - we're actually looking for low Q - the skin effect works in our favour here.
To that end, steel is better, rather than copper - hence my idea to use piano wire. Windings need spacing out by a mm or so, and of course as its at mains potential it will need insulation. At the very least fit some sleeving over it - for class II put them inside a plastic box too.
The second stage filter (L3) I think is best wound as a bifilar choke. The optimum core material and topology I am going to do some research into and when I've got some results, I'll post it up. Ideally it needs to be nice and lossy, so all the RF interference we don't want reaching our sensitive circuits is turned into heat. My initial suggestion is to go for a ferrite toroid in 4C65 material - preferably at least 20mm diameter to keep self-capacitance low.
The third and final stage filter - L4 - is just be an off the shelf common mode choke, wound with separated windings for best HF rejection. A suggested part number from Murata is PLA10AN1330R5D2 - rejection curve shown in the attachment.
You'll notice I've thrown in a few snubbers for good measure. There are different capacitor values as then they'll have effect across the whole frequency band. A large capacitor like the 470nF has a lowish SRF so needs paralleling with a lower value (1nF).

Pre-built mains filters are of some use in audio, but they're not really optimised for high-end sound. For audio source components, the draw is normally under 100W and so the current (assuming here 230V supply) is under half an amp. Few if any off the shelf mains filters are designed for such low currents.
Second, the commercial filters assume that we wish to stop conducted interference getting out as much as prevent muck getting in. The internal noise would normally only come from the rectifiers in a linear power supply (SMPSUs are a totally different matter). Once we've snubbed them (you have done that already right, or are using soft-recovery or schottky diodes?), with a linear PSU in an audio application our primary concern is against interference getting in.
Mains filter elements
A Mains filter consists of one or more common-mode chokes and a handful of capacitors arranged in fairly standard configurations. A common-mode choke is one in which the current flows equally through two mutually opposed windings. Thus current which comes in and goes out again through the choke sees negligible inductance. That's the current which goes into the transformer and powers our circuit - we don't want that impeded very much. But current that comes in and leaves by a different route - generally on ground interconnections to other units - sees the full inductance in series with it and hence gets attenuated. Its those common mode currents which can have an impact on the sound we hear as when unbalanced interconnects are used, they induce noise voltages which directly add to our wanted signals.
If you look at a catalogue of such filters, you'll notice they come in various current ratings, and often the lowest one is 3A. For audio source components, we're rarely drawing 700W so it'd be nice to have a lower rating. That's where building your own becomes attractive. As the current rating falls, so the common mode choke's inductance rises. A higher inductance ought to give us better filtering, all other things being equal.
In reality though, all other things aren't so equal - higher inductance usually means more windings and more wire area means higher self-capacitance. Thus while the rejection is better at lower frequencies for a lower current rated choke, the frequency where it effectively stops working entirely (due to the stray capacitance) reduces too. Have a look at the Murata PLA10 family of curves - on the right you'll notice all the chokes follow the same line. At those frequencies they look like capacitors to the interference, which consequently sails right through, proportional to frequency.
There's a fairly simple fix for this problem - use more than one common-mode choke in series. Some commercial blocks already do this. Commercial blocks though also usually incorporate Y capacitors. There are two places capacitors are connected - directly between line and neutral (called X) and between line/neutral and earth (called Y). Y capacitors are potentially a problem for audio because they connect the mains (both sides) to our local earth at RF frequencies. [That is assuming you do have an earth - if you're building a class II source then you won't, so having Y caps isn't a problem.] If you are using the earth, then you'll need to think carefully whether Y caps will actually help you.
Rolling your own filter
A DIY filter is certainly going to give the best potential results, if you know how to design and build it. My idea here is to build one which has three stages - each stage optimised for a different frequency band. My hand-drawn sketch shows a possible design.
The highest frequency band (say up to 100MHz or so) is best addressed with series air-core chokes (L1 & L2 in the sketch). These have the lowest self-capacitance (hence highest SRF - self resonant frequency). There's no need to have these chokes arranged in common mode configuration as they can't ever saturate. Air core chokes with high SRF can't have very high inductance - say single digit microhenries in practice. They'll be rather bulky as the windings need spacing out but they can (and should) be wound with thin wire as we actively want them to be lossy at HF - we're actually looking for low Q - the skin effect works in our favour here.

The second stage filter (L3) I think is best wound as a bifilar choke. The optimum core material and topology I am going to do some research into and when I've got some results, I'll post it up. Ideally it needs to be nice and lossy, so all the RF interference we don't want reaching our sensitive circuits is turned into heat. My initial suggestion is to go for a ferrite toroid in 4C65 material - preferably at least 20mm diameter to keep self-capacitance low.
The third and final stage filter - L4 - is just be an off the shelf common mode choke, wound with separated windings for best HF rejection. A suggested part number from Murata is PLA10AN1330R5D2 - rejection curve shown in the attachment.
You'll notice I've thrown in a few snubbers for good measure. There are different capacitor values as then they'll have effect across the whole frequency band. A large capacitor like the 470nF has a lowish SRF so needs paralleling with a lower value (1nF).
Total Comments 3
Comments
-
Good idea. You should really have a look at the Felix Project over on Audio Circle.
Felix project
It's a well developed project with a history.
One thing to look out for is that the noise has a very low impedance path to ground. Otherwise it can't get out! You end up with more noise than you started with.Posted 12th February 2011 at 01:42 PM by Pano -
Thanks Pano, I wasn't aware of that project, I will definitely check it out.
<later> Ok, have read through all 20 pages. They just use a single CMC - although at one point Occam mentioned cascading chokes. So it seems I'm doing something different here in cascading different CMCs for different frequency bands. I liked your post testing that other mains filter btw, good to get some real measurements in there. No-one so far has done measurements on the Felix, I'd be interested to see how well it does above 1MHz.Posted 13th February 2011 at 01:15 AM by abraxalito
Updated 13th February 2011 at 03:12 AM by abraxalito -
Posted 13th February 2011 at 03:24 PM by Pano