| Electronics > Beginners |
| is it worth chaining mains filters? |
| << < (2/2) |
| tkamiya:
Say one filter reduces "noise" by 60dB. Having two will reduce the same noise by 63dB. I would think it's a kind of pointless gain. Also, those specs are to be taken with grain of salt. In reality, it is nearly impossible to achieve even close to the kind of number that shows up in spec sheet. Once inside the case, it will get re-radiated also. You'll have to go through your practice with fine tooth comb. Grounding, seam in case, wire routing, etc, all become very critical. In my former life, I was part of a team that designed a micro computer driven equipment that will interface with large LARGE factory machines. I think we had 3 layers of filters. At input, in the case, and in the power supply module. For acceptance testing, we ran a pulse generator at 1KV and overlayed it on top of mains input. Pretty much everything went THROUGH. From there, we brought in scopes and bypassed where we suspect was picking up noise. We got it done but not by adding more of the same. By the time more layers are necessary and effective, you are talking about serious targeted approach. You got to know the nature of the noise. I would just put one good filter and leave it at that. |
| T3sl4co1l:
Where does 63dB come from..? The specs are at 50 ohms (and some give odd impedances, like, 1Ω/100Ω, which can be more representative of some situations -- or, at least, hints at what you might expect at very different impedances), and yeah, in a typical application you'll have something very different from that at most frequencies (SMPS being a typical example, where the source impedance is essentially the capacitance of the switching transformer, no resistance there). If the pulses were IEC 61000-4-5 surge or similar, the frequency content of that is pretty low (a peak in the, uh.. 50kHz range?), and the amplitude quite high, so that you may expect the common mode choke to saturate and let pretty much everything through, slightly delayed. Or let it through regardless because the load side has a high common mode impedance (= higher cutoff frequency). This and other reasons are why the line-to-ground voltage rating is higher (2.5kV?) than line-to-line. You need to be able to ride through something like that. Tim |
| tkamiya:
Did I screw that up? If one filter reduces whatever by 60dB (ficticious number), having it done twice and measured that in power will be -3dB, so 63dB? |
| T3sl4co1l:
60dB is one millionth power, or one thousandth voltage or current; if the filters stack perfectly (they won't!*), the output is one millionth of one millionth the power at the input, or 120dB. *We can figure this because, judging by the internal diagram, the filter's port impedance will not be a constant 50 ohms. A given filter is measured into 50 ohms, but two filters measured together will only give the product when they both see 50 ohms, which means one or both (if the filters are identical, then both of course) need to be constant-resistance filters. :) Which are perfectly possible to create -- but I don't think you'll ever see one in a stock mains filter. In practice, you can expect more, if maybe not actually double (in dB's), and give or take peaks and dips and shifting of cutoffs. You can expect closer to ideal if you add some damping inbetween, to get the impedance back closer to 50 ohms -- at the midpoint between filters, an R+C to ground (R = 47R say, and C of some nF's, comparable to what Y caps are on there already) from each line will help with this. :-+ Tim |
| Calvin:
Hi, filters complex output and input impedance interact. So the chance is high that peaks develop where attenuation is wanted. You should verify the result before trusting blindly. regards Calvin |
| Navigation |
| Message Index |
| Previous page |