CM inductor in DC link

Probably not -- on the AC side, it can take care of diode recovery noise. Not that it's much. The diodes will also act as PIN diodes, gating the RF only when conducting; some still gets through (they have fairly high capacitance, say >100s pF), and most still gets through during the conduction peaks anyway ("most", given the peak-detecting nature of the analyzer), but it might still save a few dB, like spread spectrum does. (And finally, the parametric mixing with line frequency may split peaks itself, saving another dB or two even without SS.)

Plus, the filter should be as close to the outside world as possible, so the length of wire exposed to internal fields can be as short as possible. Having the rectifier and AC line components (fuse, switch, precharge if applicable) exposed makes the design more difficult.

Now, the DC output may also benefit from common mode stuff. I have a project in mind that's probably going to be easier to build with a DC link board followed by an inverter board, and keeping stray fields down inside the enclosure is most likely going to require (or at least benefit from) a common mode choke on that cable between. On the plus side, the boards can be RF-grounded to the enclosure, so the requirements are more like, a ferrite bead to add some impedance (reducing ground currents), rather than sheer attenuation to keep noise off an indefinite AC line run. But, it's still important, as it helps avoid noise induced in signal level cables and circuits, which means fewer things to isolate (ugh!). Plus, I kind of just want to see how quiet I can make something (advantages of personal projects!).

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com 

"Klaus Kragelund"  wrote in message  
news:06ff8735-6527-407e-b4d0-8b20551ce963@googlegroups.com... 
> Hi 
> 
> Lets say we have a frequency inverter (VFD). Classic with RFI, bridge  
> rectifier, DC link and 6 phase inverter 
> 
> Anyone tried to move the CM inductor from the RFI filter to the DC link  
> side (thus on the other side of the bridge rectifier) 
> 
> Results, experiences? 
> 
> Cheers 
> 
> Klaus

F

The design we have right now has two common mode inductors (2 stage filter) , so I will move one of them to the DC link side, which means the other one still can help with the filtering of reverse recovery effects (which are s ignificant in our design due to the diodes used by the power module supplie r)

The CM inductor in the DC link side has lower peak currents - in my design about 30% (due to a intermediate DC link capacitor just after the bridge), so it can be designed with higher inductance for the same form factor. It m ay also be better due to symmetrical filtering of the DC link rails.

Also it will be a 2 winding version instead of a 3 winding version so that may be advantageous in combination with the placement after the bridge rect ifier, which means the 3mm creepage requirements are gone.

I will try it, and report back here :-)

(often in EMC, what seems a good idea on paper, it not in real life)

Cheers

Klaus

Maybe, maybe not.

Think FILTER! where the idea is to 'mismatch' impedances for the unwanted. you want SHORT to GND, OPEN through, SHORT to GND, OPEN through ...that type of thing.

But the important thing to notice is that sometimes the first stage of a filter gets thrown away and sometimes the last stage too [albeit inadvertantly] because the NOISE IMPEDANCE negates that stage's effect. For example, if the the first stage is short to GND and the noise Z is very low, well that stage has little effect. Or the correlary, the noise impedance is high Z and the fist stage is now OPEN through, again not the affect you wanted, practically lose that first stage's effect.

Now, most noise from an SMPS is BOTH types, depending on what frequency you'r at, or where in the box the circuitry is picking up the noise [like magnetic fields from the transformer] At low frequency the noise is usually extremely low Z, at high frequency the noise usually becomes a high Z, but depends.

BEFORE you go moving stages aroudn will-nilly, you should characterize the noise source(s) you're trying to remove to make certain that you're not shooting yourself in the foot. AND to maintain control of the design, AND [you have no idea how often I've found this] you've been wasting money on filtering that is unneeded!

Well, I know what the source is. It's the bridge rectifier shorting (500mA reverse recovery current when it turns off), and that's low impedance

However, when you have a 2 stage filter, the effects of the impedance mismatch is low, but I am trying to remove one of the commom mode inductors, so it becomes very important again.

Right. That's why I am attacking the EMC filter. It's not my design, but now when we have problems, I have inherited it ;-)

It's not always easy to find the cause. In this case the reverse recovery of the diodes cause a current glitch and that current glitch triggers resonances in the EMC filter, so the EMI spectrum is deceiving at first glance.

I tried feeding the entire converter with high voltage DC (so that the bridge is not commutating) and then the noise is very low, which means I think I can exchange one of the common mode inductors with minor DM filtering instead (in front of the bridge)

Cheers

Klaus

That's why LTspice is your friend! Don't forget to include the parasitics of the required setup, AND the cabling back to your LISN.

That is already done, except in this case Orcad Pspice is my friend ;-)

Cheers

Klaus

in that case here's a Belden 'medical grade' AC mains cord:

• The parameters in this model library were derived

• from data using a test setup

* ** Example use: XLNCRD1 1 2 3 4 5 6 LINCRDH * *

• - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

- -

• LINE CORD 120VAC 10A from Belden for Hospitals

* *

• Terminations are AC power plug and AC power receptacle

• Ratings: 10 Amp, 120/250V 50-60Hz

*

• AC LINE Supply

• | GND Chassis Neutral

• | | AC LINE Return

• | | |

• | | |

AC LOAD Supply

• | | |

| GND Chassis Neutral

• | | |

| | AC LOAD Return

• | | |

| | | .SUBCKT LINCRDH 1 2 3 4 5 6

• LSUP 1 7 2.5UH LRET 2 8 2.5UH LGND 3 9 2.5UH KSUPRET LSUP LRET .46 KSUPGND LSUP LGND .46 KRETGND LRET LGND .46 RSUP 7 4 .06 RRET 8 5 .06 RGND 9 6 .06 CSUPRET 4 5 160PF CSUPGND 4 6 160PF CRETGND 5 6 160PF .ENDS

The simplest application of the above is to note that capacitor-input EMI filters often do not deliver their specsheet performance, because the input capacitor is shorted out by the source impedance of the power source, and so has zero effect.

The datasheets often specify the filter transfer function in a 50-ohm system, following a benighted MIL SPEC, which impedance is what the usual test equipment provides at both input and output. And which is not real-world at power frequencies.

Joe Gwinn

They do what the spec sheet says, into 50 ohms. Whether you're using it at 50 ohms or not, that's your problem. :-)

Ah, but:

1a. At the magic frequencies where the line does radiate (around various combinations of straight lengths, bends and branches), it looks kind-of-sorta-more-or-less 50 ohms. Maybe in the 25-100 ohm range, but not ridiculously out of whack. These are the frequencies where real power is absorbed by the line, and radiated into space. This is B-A-D, hence the filter.

1b. At anti-resonant frequencies, the line does not radiate appreciably, and the impedance is either very high or very low (limited by losses in the cabling, presumably; which should be fairly high given the common use of PVC, which is a fairly lossy dielectric). In these cases, the voltage or current transmitted to the wiring may be large (and so you'll see ugly signals on the oscilloscope), but because it's not radiating much, the impaired filter performance isn't as big of a deal.

1. So as to avoid any possible "special case" scenarios, let alone gaming the system for a given setup, they simply assume the line radiates at all frequencies -- a perfect match. Real line is probably not exactly 50 ohms (a long stretch in free space could be hundreds of ohms), but it's at least a start, and sure beats trying to get electricians to run impedance controlled wiring. ( :) ) Perhaps the published threshold levels have been adjusted to account for a safety margin beyond the intended threshold, as well.

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

All true, but still I'd bet an inductor-input filter will fare better in a real power system.The datasheets of capacitor-input EMI filters are strangely silent here.

Joe Gwinn