Need help with RF CM-Filter Design Problem

[Jieve]

I've only recently started studying EMI and common-mode filtering and need some design advice, so please bear with me.

I have an audio circuit project that I am working on, to which I want to add common-mode filtering at the output terminals to prevent any RF (especially in FM range) picked up by the cables from being demodulated into the audio. A schematic of the output circuit including a common-mode filter consisting of a CM-choke and 2 caps to chassis is attached. I'm assuming the source to be a common-mode voltage source located along the cable somewhere, with the source impedance being the cable itself. I am attempting to minimize the RF voltage at the load by shunting the CM current to the chassis (chassis is earth grounded, and signal gnd is connected to chassis in one place).

In order to simulate the frequency response accurately, I need to include source and load cm impedances in the simulation. I have calculated the cm load impedance as approx 24 Ohms, looking into the circuit from the filter caps, and have measured it on the actual device up through 5MHz, and results were pretty much dead on. However, I don't exactly know how to measure the source impedance (I assume this is simply the cable impedance applying a common-mode signal to both conductors, likely twisted-pair). I've simulated the results with a wide range of different source impedances to get an idea of where the response is going to lie (graph attached in next post), but now I'd like to confirm it via measurement.

I have a 30MHz signal generator and 100MHz scope, but nothing fancy (expensive) like a spectrum analyzer. So now my questions: 1) Is the way I'm visualizing the RF noise circuit correct (voltage mid-cable, common-mode on both conductors)? 2) Is the way I'm thinking about this design process correct (minimizing RF voltage at the "load impedance", or the other side of the filter)? and finally 3) If this is correct, what is the best way to measure twisted pair cable impedance with the equipment I have? I was thinking of just using a shunt resistor and something like 100-1000ft of cable to get enough resistance to make an accurate measurement. I'd be happy with a value up through 10MHz for now, just to give me an idea of how close my simulation matches reality.

Thanks for any input.

 

[Jieve]

Here are the simulation results.

 

[sunnysky]

There are a bunch of errors here.

First start off with measurable specs.
What is your signal?
What is your noise attenuation.
What is differential impedance and signal?
What is Common Mode impedance and signal?

Normally in CM chokes the idea is balance a differential signal so that CM noise is cancelled. Here it is not balanced since there is no signal driven on R12..

 

[VEBill]

Simulation has its limits, but if you invest more and more time and effort, you can push it as far as you like.

Let me give you an example. Your circuit diagram shows the bypass capacitors going to ground. And your simulation software can do nothing but assume that the ground is perfect. But when you build it, the ground point is itself going to have some complicated parameters and unexpected coupling. In general, these may come into view at higher frequencies. You can only imagine what's required to include such physical effects in a simulation.

Designing for E3 compliance requires a much wider view than what you'll get from a simple simulation based on the schematic. Root cause of E3 issues is often a design issue that doesn't even show up on the schematic. Things like shielding, correct bonding of connectors, routing of high current paths. Physical inspection of the build is often more revealing than the schematic.

Keeping a broadcast station out of an audio amplifier should not be difficult, unless you live in the shadow of the tower.

 

[sunnysky]

Until you define the specs for Signal impedance, bandwidth and CM noise spectrum and impedance ( usually high like xx pF or x uH ) you can never design a perfect filter.

Never start a design before good specs !! N.B.

e.g.
pass band , band stop, differential attenuation, CM attenuation, CM impedance

A design is perfect only if it meets the defined specs with margin.

guess if you dont know then verify. Noise can be conducted , radiated, induced by capacitance from E field, dV/dt or inductance from H field currents dI/dt

For a CM audio amp, on a long cable like an intercom, then to avoid AC hum you may have both E and H field interfence with noise much greater than signal so your CM filter approach might make sense . Telephone landline handsets use a combination of CM and DM hybrid transformer for use in long connections to Central Office.

Having done ingress tests on CATV coax lines near railroads, we discovered how much shielding fails to suppress traction motor noise in the midband where payTV once was above 100MHz. After looking at CNR ratios on a Spectrum Analyzer, we concluded the ground connections were degrading raising the Common Mode impedance and resulting in TV interference on CATV whenever the trains went by a district with at least 150 m distance to tracks.

Just today I was checking my strings of LEDs along the back fence due to an open connection. I used to DMM just to measure stray AC line noise on DC using the rectifcation of the LEDs and it generated 6V on a 12V string into 10Mohm DMM. Since the positive side would clip I knew that the Positive side would read a negative DC voltage being parallel strings of 3x3V.

If you want a great noise generator. Just get an automotive SPDT 12V relay and use the NC contacts in series with the coil current and you have a buzzer that will arc the contacts with maybe 50~100mA of current depnding on V/DCR but spread over the entire spectrum of your scope. Make a big area loop antenna and you can probably jam some cell phones.

Then try to measure 1uV off your scalp with an instrument amplifier (INA IC) and learn that shielding and high impedance is not enough and active Vcm guarding may be the solution on the shield of twisted pairs to eliminate noise in EEG signals.

 

[Jieve]

Normally in CM chokes the idea is balance a differential signal so that CM noise is cancelled. Here it is not balanced since there is no signal driven on R12..

To be balanced, an audio signal does not need to be driven equal and opposite on both lines (although this is commonly done). It only needs to have equal impedances on both balanced lines to signal ground.

Let me give you an example. Your circuit diagram shows the bypass capacitors going to ground. And your simulation software can do nothing but assume that the ground is perfect. But when you build it, the ground point is itself going to have some complicated parameters and unexpected coupling. In general, these may come into view at higher frequencies. You can only imagine what's required to include such physical effects in a simulation.

Designing for E3 compliance requires a much wider view than what you'll get from a simple simulation based on the schematic.

Right, definitely agree with you. The schematic didn't actually show the ESR, ESL I included, for example, although they were used in the simulation. I also tried to lay things out on the board according to best practices from Ott's book and other sources I've found around the web to try to cover as many bases as possible, or at least the ones I could model practically with my level of knowledge/understanding.

Until you define the specs for Signal impedance, bandwidth and CM noise spectrum and impedance ( usually high like xx pF or x uH ) you can never design a perfect filter.

Never start a design before good specs !! N.B.

I can't argue with you here, but this was exactly my reasoning for posting this question. I knew I wanted to filter CM in the FM range and beyond (although not exactly how much). My problem was that I needed impedances to be able to model any filter effectiveness, but didn't know how to get them (specifically the source impedance from the cable). In Ott's book "Electromagnetic compatibility engineering" he claims cables to typically be high CM impedance and ground planes to be low, but I wanted to see this for myself.

I was eventually able to do some current/voltage measurements from 5MHz to 30MHz on 10-100ft lengths of twisted pair cable, and calculate the impedance. I found it to behave like an unterminated transmission line depending on length, with the CM impedance of the cable generally being much higher than that of the balanced output to chassis ground (~24ohms). So I ended up switching the positions of the caps and choke, so the caps are nearest the output.

I recently got the boards and assembled the prototype circuit, and it seems to work quite well (in the audio range and beyond). I have not yet been able to do any testing with regard to RF interference though, so haven't been able to test the RF filtering yet. Still working out what equipment I need to get realistic results without breaking the bank. Something like that noise generator idea might work well (I think this was mentioned in the book as well, or maybe he was recommending using a dremel tool in the vicinity, would have to go back and find it.)