Line filtering

Tim Williams wrote: (snip)

(snip)

They are there to put an impedance in series with a signal that is in common with both sides of the line. For instance, assume you have a rectifier and capacitor down stream of the filter, and the DC supply feeds a switching mechanism that produces rapid voltage swings that feed capacitive current into ground (through a transformer inter winding shield, perhaps). That current completes a loop through ground, back to the power lines and through the common mode inductor back to the rectifiers and capacitor, every time the rectifiers conduct. This causes a pulse of high frequency in common on both sides of the power line twice per cycle. The common mode choke gets in the way of that current completing the loop. Without it, the radio noise radiated by the pair of power lines because of the current they each carry half of, acting as a single conductor at RF, would be much worse.

Harmonic leakage from the cables & any gaps on the metalwork, mostly.

--
   W  "Some people are alive only because it is illegal to kill them."
 . | ,. w ,      
  \\|/  \\|/              Perna condita delenda est
---^----^---------------------------------------------------------------

I would be thinking in terms of using a choke input DC filter. That way, you not only clean up the differential noise on the line, you improve the power factor by eliminating the high current peaks each time the rectifiers turn on. Think in terms of the choke used on the output of a simple DC welder. Small shoe box size.

So why don't I just put a big fat choke on the ground line? It's not like it has to carry any current!

How about the other points: two-line supplies and my situation?

Tim

-- Deep Fryer: A very philos> (snip)

Ouch, there goes rail regulation. Seems kind of a waste to run 600V IGBTs from +/-100V. I don't even have much capacitance, at least yet.

Besides, that's gotta be what, 0.1H at 50A? Yech! Could buy the IGBTs to build a PFC for the price of the *wire* for that sucker!

Tim

-- Deep Fryer: A very philosophical monk. Website @

PI-filter on your DC bus then. But without some line inductance (before or after rectifier, no major diff) then as JJ points out, current peaks will be HIGH. A Pi-filter also means the cap lifetime calcs are easier - rectifier-side caps see BRUTAL 100/120Hz ripple, but no HF. vice-versa for SMPS-side caps.

Anecdote:

testing a range of 230V and 400V AC motor controllers. Tech measures RMS input current. 4.7kW drive has 30% *more* Irms than 12kW drive. WTF says I, then goes and does calcs. large drive has 5% line chokes, small drives none. Voila, math = measurement. ouch.

Cheers Terry

no worries. thanks for the pics. hairy-faced gits flock together?

rather than a WAG, use a scientific WAG (SWAG). this is for the DM filter.

you ought to know what the HF ripple in the DC bus capacitance is. its usually easier to set up a crude simulation than an analytic expression. Then do a reasonabole (include ESL,ESR) model of the DC bus caps, so you have a reasonable idea of the ripple voltage.

now you can look at a roughly accurate spectrum of said ripple voltage.

I typically do the design with a copy of the relevant EMI spec - that way you know what your allowable DM current into 50R is. Because EMC is all about parasitics, I pick the lowest allowable level, and use that to design the filter. but I include reasonable values of parasitics for the DM filter components.

I usually have 2 models - a switching-level model, so I can get fairly accurate RMS currents, and a cruder model for the filter, using some sort of a vandalised current source to represent the current into the filtering network.

the rectifier complicates things somewhat. but it tends to look pretty capacitive, so is a short at EMC frequencies. which is what I use.

then fiddle with your components (AC sweep) until you have something that looks like it will meet the requisite EMC standards. inductor end-to-end capacitance is hugely important - a massive inductor does naff all at EMC frequencies because of this capacitance, which is dominated by the start wire and the final layer of turns + the finish wire. Adding more L does naff all, the C doesnt really change but the corner frequency (at which the L stops helping) gets lower...... if you make the inductors, you can control this - eg bank winding, with N+0.5 turns so the start and finish stay the hell away from each other.....

with reasonably large caps that work at HF (ESL, ESL, ESL) and a moderate switching frequency, you can probably use a fairly small L,

for a reasonable AC (or DC) inductance, use 5% chokes. 3% is the critical value below which not much happens; 5% gives a huge reduction in peak current, for 5% drop in AC volts across said chokes at 100% rated current (do the calc ignoring the nasty(ish) current waveshape) so the line regulation is ~ 5%.

5% for a 10kW 240V single-phase supply is:

Vbase = 240Vrms

Pbase = 10kW

Ibase = 41.7Arms

Zbase = 5.8R

Wbase = 120*pi

Lbase = 15.3mH

5% L = 763uH

Epeak = 1.32J - and that DOESNT INCLUDE the evil current peaks. its a ~2J inductor.

again its easy to simulate the overall rectifier behaviour for AC line current, but hard to calculate. try stepping L, have a look. then use the actual peak current to calculate the required energy, so your L doesnt saturate at the current peaks (thereby defeating its purpose)

HTH

Cheers Terry

Tim Williams a écrit :

Sure there is. A PFC will handle all that nicely :-)

--
Thanks,
Fred.

Thanks for the reply.

So I should say, split my electrolytic cap bank with a few microhenries (or maybe as much as mH, I forget) to decouple the HF from one side? Or just toss on more film caps (maybe even one of those swanky "film-lytics") for the HF side of the filter?

So, at the line, I take it there's no way to get away from high ripple or bad regulation or bad power factor (PFC aside).

Tim

-- Deep Fryer: A very philos> >

IGBTs

to

expression ...

...

Bugger... I don't like simulation. I don't have a simulator on hand, either.

Don't have any idea what kind of EMI spec I want. I mean, I'm building this for myself, is the FCC going to give a damn about a couple kilowatts of radiated hash running for at most a couple hours?

Tim

-- "Librarians are hiding something." - Steven Colbert Website @

LTspice. I dont use it, but its free and good (a rare combination)

In that case, size your rectifier caps for the 120Hz ripple current, and the caps on the HF side of the Pi filter to prevent bus overshoot and device destruction. then stick as much inductance as you can be bothered making between the two; 20dB is a good start wrt attenuation. remember to put something HF in parallel with the LF electros.

then smack in as much CM inductance in the AC line as you can be bothered winding. an X2 cap across the AC of the rectifier, and a pair of Y caps from P,N to E at the same point - so any capacitive coupling to chassis flows thru these caps and back into your smps.

or do nothing. after all, your work coil is air-cored so spews flux everywhere (although its pretty sinusoidal). this is the quickest option.

Cheers Terry

If you don't recall, I have electrolytics from computer supplies, 16 x 470uF

200V. Any bearing on the capacity of them? I recall typical specs for similar product are around 2-4A ripple, and I have no recollection of ESR, ESL...

Any recommendations on film caps? It's looking like I'll have to buy some more... I've got Cornell Dubilier DMEs (2 x 10 x 2.2uF 400V) on there right now. Seems to be the cheapest, densest mylar line available, at least at the time I checked. No good for current (I have a clip of some 0.1uF's on a tank circuit and one starts oozing smoke) and notable ESL (series resonant circa 300kHz for...0.47uF I think?), so basically bulk nonpolar capacitance.

With some nice polypropylenes on the inverter itself, HF crud should be low enough by the time the ripple gets to the film cap bank, huh?

Oh...do I have links to any recent pictures? I forget... This is about the most recent I have online...

Except the big coupling capacitor is now split in two, so it makes a fake CT off the supply voltage.

Got it. Common mode should be pretty easy, although I don't really have a core big enough to loop all that 6AWG through. Or the #6, for that matter...Maybe another pi-wound dealie from 1/2 x 0.04" copper (like the Lmatch)...

Yeah, but that's cheating. The power line conducts crud, while the coil drops off as what, inverse cubed distance?

Tim

-- "Librarians are hiding something." - Steven Colbert Website @