Continuing the crusade against mains interference, this month we take a look at mains filters.

Because mains interference is a headache for all types of sensitive equipment, whether it's computers, telecoms, instruments or audio, 'wave' or 'lin' filters are available as standard components for cleaning up the juice. There are dozens of manufacturers, and many styles are available. Figure 1a shows a 'beancan' filter, made to fit inside the gear. The beancan bolts down to the chassis, and the mains is simply routed through the filter. Some beancans have 'Faston' (Lucar) blades, for crimp connections, others have solder lugs or turrets.

The filter in figure 1b has been incorporated into an IEC Euroconnector socket, the familiar rectangular 3 pin connector. It has the same Lucar blades, so an exchange can be made without soldering. However, the filtered version is longer - it projects further into the equipment. Assuming there's enough space though, this style of filter can be installed in most studio equipment in a few minutes, using a spanner and a screwdriver alone.

If you're too lazy to do this, figure 1c shows a lead with an in-line filter, which can be simply plugged in for an instant clean-up, while figure 1d, the box filter, is a variation on the same theme.

Mains filters come in a variety of current ratings, from 1 amp up to 25 or more amps. Tied in with this is the most important parameter, the attenuation curve. The idea of the curve is to tell us how effective the filter is across the range of frequencies, and presumably this data is a little embarrassing for these curves are rarely presented in any detail in catalogues.

If you idly imagined that an RF filter just blocks out RF nasties, end of story, then you're in for a shock when you look at figures 2 & 3. Real RF filters have erratic attenuation curves, the performance being especially subject to the mode of interference (We'll look at this shortly) and the source impedance of the mains at high (RF) frequencies. At 50Hz, the mains impedance is of course very low (otherwise it would be hopeless for driving 3kW electric heaters), but above 100Hz, the impedance rises, and then varies quite erratically between tens and thousands of ohms. The exact impedance vs. frequency curve is unique to every mains outlet (figure 4). Overall, this means the attenuation provided by a filtered can both vary dramatically with frequency, and depends on how lucky you are. Later, we'll see that this behaviour isn't as bad as it sounds.

Figures 2 and 3 display typical attenuation with frequency for a pair of filters, for two modes of interference, called symmetrical and asymmetric. The filters have different current ratings (1¼ Amps and 25 Amps respectively), but are otherwise from the same maker, and are built to the same pattern.

These curves illustrate a general rule of thumb: the lower current ratings have more attenuation across the band, yet over a specific portion of the frequency range, a higher current filter can excel. In this instance, the 25 Amp beastie outperforms the 1½ Amp filter on symmetrical interference, but only below 900kHz, and above 13mHz. The upshot is that we go for the lowest current rating we can get away with (it's cheaper for a start), but that for particularly difficult RFI (Radio Frequency Interference), it's worthwhile checking out the curves for higher power filters - higher than you strictly need, just in case they work out better over the most bothersome band of frequencies. Once again, I've identified the frequencies of some real-life interference sources at the bottom of Figure 2.

Sym or assym?

Puzzled by symmetrical and asymmetric interference? These are jargon terms for differential ('balanced') and common-mode ('unbalanced') inter-ference/noise sources respectively. For example, imagine the mains live and neutral are like a balanced input on a console. Looking now at Figure 5a, the interference source has been drawn to appear in line with the mains live. This could be an on/off switch, or a dodgy mains adaptor. The ear-splitting interference current returns through the neutral wire, so in effect, the interference source is in series with the live and neutral inputs. So equal, yet opposite currents flow up the live and down the neutral. This is a differential current flow, the same behaviour as in a balanced mic circuit. Interference currents are called symmetrical when they behave like this.

Most interference is asymmetric, particularly when it originates from any distance. Looking at Figure 5b, the interference currents on live and neutral are more or less equal, hence the description 'Common mode'. They have the same polarity ('direction'), and the noise current returns back down the mains earth. This common mode interference is in principle the same as the nasties that a balanced line or mic input is tuned to reject.

As you can see from Figures 2 and 3, big or small filters are generally better at attenuating asymmetric noises (Figure 3). The attenuation of symmetrical noise, created when the equipment's own power, or some adjacent gear is switched, is poor to non-existent. So don't expect filters to sort out the arc noises arising from your own dubious mains connections and worn out switches!

Improvements: At both ends

What happens if our mains lead picks up the local taxi-cabs? At VHF frequencies, (10mHz and above), the typical beancan filter's attenuation is falling off. But we can soup up the performance of any mains filter by placing it inside the equipment, and keeping the mains input wires as short as possible before they pass through the filter. These conditions are most easily met by using a combined Euroconnector filter (Figure 1b). Their attenuation is generally healthier than beancans above 100MHz, although less good at the bottom end (500kHz).

Threading a handful of ferrite beads or ferrite rings over the incoming mains cable or even better, over the individual conductors is also worth trying out - at least it's inexpensive.

Because short lengths of wire make good resonant aerials for RF energy at VHF, but at spot frequencies only, we can often make a dramatic reduction in VHF feedthrough by altering the length of the mains lead or leads. Shorter leads are best, though in theory, lengthening can work just as well, provided you actually manage to detune the resonance. Usually, detuning means chopping off a few inches at a time, and obviously such incremental adjustments are a shade impractical if the cable is to be lengthened... Chopping down is done while metering the audio FX, and watching for the null point. If you got too far, the RF level will come back up, as the wire becomes tuned to a definite fraction of the original resonant frequency. For example, Taxis broadcast on 87mHz. Their wavelength, is 87,000,000Hz/30,000,000 = 2.9 metres. So mains cables with lengths of 5.8m (2λ), 2.9m (λ), 1.8m (5/8λ), 1.45m (1/2λ) and 0.73m (1/4λ) are all potentially good at hyping up the local taxis' RF, making breakthrough more likely. Yet by adding or subtracting a small percentage of these critical lengths, the problem can be made to go away - or at least go to another frequency, hopefully a less bothersome one!

At lower audio frequencies, (below 1mHz), common asymmetric noise currents on the mains can reach the audio circuitry through the large stray capacitance across the mains transformer's windings. To prevent this, quality conscious makers fit an electrostatic or interwinding shield (Figure 6). Built inside the transformer, the shield is not something we can fit as an afterthought. However, standard transformers are available with electrostatic shields (for example, the Avel-Lindberg range, available from RS components), and it's sometimes feasible and worthwhile to replace an existing unshielded transformer with one of these.

line filters as supporters

You can use a mains filter for suppression. The way this differs from filtration is that the filter is placed close up to the source of the interference; a noisy fridge or a dimmer pack, for example.

Figure 7 displays the contents of a standard line filter. There's a twin, bi-filar wound inductor (the wires run in opposite directions to prevent the high level mains current from saturating the inductors), L1/L2, and a triad of capacitors; C1-3. The resistor bleeds off any residual voltage on the capacitors for safety. Most importantly, the diagram shows the position of the capacitors for conventional filtration. They're on the output side of the filter, closest to the equipment's power supply. For suppression, the opposite is in theory correct (Figure 7b), where we reverse the filter, so the capacitors are on the input side.

Don't be afraid to test out the filter in reverse. For suppressing broadband hash from thyristor control equipment and switch-mode power supplies, it's routine to connect the filter back to front, because this direction works best (Figure 7c).

Safety

Looking closely at Figure 7a, C1 and C3 connect mains live and neutral to earth. If the earth connection is left out or broken at X, the filter's earth terminal hangs off the live and neutral, and floats at around half the mains voltage, ie. 120 volts. Meanwhile, filter bodies are sometimes earthed to the earth terminal. Leaving off earth wires can therefore give rise to a varied catalogue of potential safety misdemeanours. However, legislation now restricts the size of C2 and C3 in filters of European or US origin, to limit the leakage curent through C2 to less than 200uA. That's barely enough to cause a slight tingle, and certainly nothing worse. If in doubt, look to see if your filter is approved to BS613. Tingle or no tingle, it's still a good idea to go about earthing the filter correctly, otherwise you'll ruin its performance. The upshot is that whenever you fit a filter to unearthed oriental gear running from a 2 core mains lead, a 3 core cable must be fitted, the green/yellow wire going directly to the filter's earth terminal. Some unearthed oriental machinery has exposed metalwork. This is naughty, but two wrongs don't make a right: unless you can earth your kit, keep the filter's metal body away from this metalwork. Evostick or self-adhesive pads are the easiest way to mount the filter onto an area of plastic.

Filtering the incoming mains

The first step and most economical way to cut out mains borne nasties is to filter the mains overall, at the point where it enters the studio. Alas, the majority of purpose-built set-ups run off the dreaded ring-main, the bête noire that feeds interference to the maximum number of power points with the least wire. We'll go into the interplay between wiring and filters next month.