Why are there still common mode currents in this system?

[sjgallagher2]

(disclosure: made a similar post from Electronics SE, but I tend to trust the EEVBlog RF guys more )

I have the following system set up in the lab:


imgur album in case inline image doesn't work: https://imgur.com/a/cy19lDs

An RF signal generator feeds 2 meters of RG-178 coax, this coax connects through a 1:1 balun (a "voltage balun" configuration) and connects to around 1.5 meters of 300 ohm twin lead, terminated in 300R. The balun is a Mini-Circuits T1-1T. The cables are draped in free space, as far as I could manage from nearby metal, around waist height.

I am able to measure antenna mode (common mode) currents on both the coax and the twin lead by measuring the induced current on a loop encircling the transmission line. I have a couple spectrum analyzers, a Rogowski coil (good to 30 MHz), and a clip-on ferrite bead with a dozen turns of wire. The ferrite bead is most sensitive and works to a couple hundred MHz for qualitative measurements.

Here is what I've observed:

• When the coax is disconnected from the balun and twin lead, its antenna mode currents disappear, regardless of the load (open, short, 50 ohm).
• When the coax connects to the balun and twin lead, both transmission lines show strong antenna mode currents, which fall off as frequency drops below 10 MHz.
• The antenna mode currents form standing waves with a propagation velocity equal to free space c, based on the frequency and separation between nodes.
• Rotating the balun ("current balun" configuration) does not affect the results at all. See image below.
• Using a 1:4 (pri:sec) balun, Mini-Circuits ADT4-6T, did not significantly change the measured currents. Return loss improved though.

imgur album in case inline image doesn't work: https://imgur.com/a/cy19lDs

Above: System with current mode balun.

I experimented with frequencies from 10 MHz through 300 MHz (note that the T1:1T balun is only spec'd to 200 MHz, additional IL and poor RL occur outside that range).

My question: Why do I still get common mode/antenna mode currents on the coax in these configurations? And what can I do to fix the imbalance difference?

[mtwieg]

It's likely just due to primary-secondary capacitance across the transformer itself. Even if the construction of the transformer is perfectly symmetric, this creates a common mode current path from input to output(s).

I believe these sorts of parts are meant to be used with the center tap node grounded (or AC grounded, at least). Note this won't actually decrease the amount of CM current in the circuit, but it will redirect the CM current to that GND so it doesn't contaminate the balanced signals.

It's also possible that your balanced load is actually not balanced. Hard to say without more info on your setup or some actual measurements.

[sjgallagher2]

Thanks mtwieg, the capacitive coupling is an idea I hadn't considered. That seems like something I should be able to measure with a good LCR from one tap of one winding to the other? Or a series (S21) or shunt (S11) measurement with a VNA? I agree about the center tap being grounded, I've done this in the past to keep things from getting ugly with balance. But I'm also interested in better understanding why the system is behaving this way, so I can avoid these sorts of things better in the future.

I've added some photos of the setup (in all its rapid prototyped glory ) to this imgur album: https://imgur.com/a/YYJ7Klk
This includes pictures of my ferrite bead "probe", the Rogowski coil, a couple versions of the transformer board, and a hole I drilled in the twin lead to try and measure current through a single conductor using the Rogowski. The load is a 300R resistor tacked on to the end of the twin lead. It's probably got its own issues above ~10MHz, but S11 is fine so I'm overly concerned. If I keep having issues I'll refine the test a bit more.

[sjgallagher2]

Update: I measured the impedance from the primary phase dot to the secondary phase dot, transformer on its own, see attached image. Measured capacitance is around 4pF (3.9-4.3pF). Impedance magnitude from winding to winding at 3MHz gets down to 10k , and is trending down, which aligns with what I saw with decreasing imbalance currents as frequency gets lower than 10 MHz, although there might be other reasons.

[mtwieg]

Ok so your balanced load is just floating out there with no conductive path back to GND. So transmission line effects must be dominating the common mode currents. If you sweep frequency, you should observe the CM current has peaks and valleys, the frequencies of these will depend on your cable lengths, and on the transformer parasitics. The amplitude of the observed CM current will also depend greatly on where your probe is along the length of the cable.

I'm surprised you observed that using it as a current mode balun didn't change the result. In my experience, current mode baluns work better for this, especially as frequency increases.

[sjgallagher2]

 I have been experimenting further, and I verified that connecting the center tap to the coax shield reduces the CM currents by up to 25dB at 40 MHz, a more modest 14dB at 100 MHz. See attached photo of the connection.

The coax and twin lead are long particularly so that I get different behavior along the transmission line length at high frequencies, that way I can move the ferrite bead along the line and look for peaks and valleys in the standing wave pattern of the CM current surface wave. I try to make all my measurements at peaks. I'll also take a look at the frequency behavior of these peaks and look for changes there.

So now I understand how common mode currents can propagate through the balun, and I can mitigate them, but why are they present in the first place? I set up this experiment to prove out some concepts in imbalance difference modeling (see e.g. https://learnemc.com/introduction-to-imbalance-difference-modeling), hoping that the balun would eliminate the mode conversion excitation, but clearly there's still a strong excitation for common mode in the system. And I thought this would be the simplest system to use! Maybe I'm still misunderstanding the source of the common mode?

[A.Z.]

A question if I'm allowed to ask...

Your drawing shows a generator connected to a coax and from there a 1:1 BalUn (let aside the current/voltage for the moment) which in turn is connected to a balanced line with 300 Ohm characteristic impedance and the latter is terminated with a 300 Ohm resistor (hopefully a non inductive one, since we're at RF); now, having a 50 Ohm coax directly connected to a 300 Ohm line/load will cause quite a mismatch, why aren't you using a 6:1 transformer to match the coax to the balanced line ?

Also, changing the transformer connection from "voltage" to "current" won't necessarily work, since the transformer may be somewhat unbalanced, you'd better using the transformer to balance voltages and add between it and the coax a decent current balun, since the latter will balance currents, it will also act as a choke, so reducing the common mode currents flowing over the external surface of the coax braid

[edit]

Forgot, in your schematic there's a 50 Ohm resistor, in series, between the generator and the center conductor of the coax, what's that supposed to do there ? I think you'd better place it in parallel, not in series if I understood what you wanted to use the resistor for.

[sjgallagher2]

Ideally I would actually have a 1:6 balun! but I didn't have one on hand I did have a 1:4 balun which I've been using as a stand in, a bit better. The resistor shown on the schematic represents the output impedance of the generator, practically speaking the coax connects directly from the RF signal generator to the balun board through to the twin lead and finally to the 300 ohm termination resistor.

[A.Z.]

which frequency are you dealing with ?
is the 300 Ohm termination resistor a non inductive one ?
if you have a 4:1 use it, you'll have less mismatch

[sjgallagher2]

Actually there's an error here, something was nagging me... The input impedance looking into a transformer is (1/T)^2 * ZL, so rather than a 1:6 transformer I would want a 1:sqrt(6), or roughly 1:2.45. A 2:5 transformer would be closer to ideal. You can verify e.g. here.

Anyway, this mismatch is a secondary concern, since mismatching doesn't impact excitation of common mode currents (except insofar as it changes the excitation amplitude, but this is dependent on position along the line, and isn't related to the excitation mechanism itself). It's still not clear how the common mode is being excited, and how to avoid its generation in the first place?

Re: frequency, this is a lab experiment to test a theory (imbalance difference modeling), so I'm trying to work across 10-100 MHz to have at least a decade of frequencies to compare. Can go up to 200 MHz.

[T3sl4co1l]

Might as well put the link for posterity:
https://electronics.stackexchange.com/questions/722802/why-are-there-still-coax-shield-currents-in-this-system
Not that there's been much activity yet, but it sounds like a self-answer is in the works, which is fun.

It's still not clear how the common mode is being excited, and how to avoid its generation in the first place?

Well, how's a transformer constructed?

Or, if you have a few of these parts to spare... why not wrench one open and see?

What does the construction tell you about things?
- The origin of the inter-winding capacitance
- What is the centroid of that capacitance? Is it closer to one end or the other, of either winding?
- How well balanced is the common-mode at frequencies where the inter-winding capacitance dominates?
- How else could we express the transformer's construction?

(I'll spoiler the last one: it's transmission lines. How could you identify the transmission lines in the construction, then?)

Tim

[sjgallagher2]

Good idea, I'll break one open! And in case anyone is playing along, I'll post some microscope photos of the T1-1T

The interwinding capacitance originates with the primary and secondary windings overlapping inside the small core. This one can be expected to have higher primary-secondary capacitance than e.g. a transformer with primary on one side of a toroid, secondary on the opposite side (diametrically). That sort of construction, if it works at RF and the ferrite gets enough flux linkage through the magnetic circuit, might be better as a common mode choke. Just a thought.

I'll revisit this when I have some time later today (hopefully).

[A.Z.]

Ok, after looking at the pics of that MiniCircuits transformer internals, I think I'll stay with CoilCraft stuff

https://www.coilcraft.com/en-us/products/transformers/wideband-rf-transformers/#/SMT

probably not the best around, but they worked/work decently well for me

[mtwieg]

Ok, after looking at the pics of that MiniCircuits transformer internals, I think I'll stay with CoilCraft stuff

The fact that this one has a center tap means the windings are going to be more complex and therefore increase parasitics. Mini-Circuits offers transformers in tons of different configurations, and I'm betting ones without the center tap would perform much better vs an unused center tap.

But I'm perplexed by the photos. There must be some splices underneath.

probably not the best around, but they worked/work decently well for me

What series are you referring to?

One thing I like about Mini-Circuits is they always provide S parameter data for their parts. For the ADT4-6T, the data is a 3-port network, and it clearly states that pin 5 is grounded (the center tap). Their eval board also grounds the center tap. So that's clearly how it's intended to be used.

I also looked at coilcaft's WBC series a bit (has some center-tapped variants). They provide four port S parameters, neat! But their README file says something odd:

Port #1 of the model corresponds to pin #1 of the part. Port #2 corresponds to pin #3, port #3 corresponds to pin #4, and port #4 corresponds to pin #6. Each port is referenced to a 50 Ohm measurement but the measurement is not limited to this impedance. Performance will change by connecting ports to different impedances and attaching ports to ground. Center-tap pins (#2 and #5) are not connected.

Strange that they don't connect the center tap(s), or include them as ports in the model. So if you don't intend to connect the center tap(s), these models are great and can predict how well it will work as a balun. But if you do use the tap(s), then their model isn't as useful...

[eb4fbz]

Actually there's an error here, something was nagging me... The input impedance looking into a transformer is (1/T)^2 * ZL, so rather than a 1:6 transformer I would want a 1:sqrt(6), or roughly 1:2.45. A 2:5 transformer would be closer to ideal. You can verify e.g. here.

Baluns are usually specified in terms of impedance ratio, instead of turns ratio. So you need a 1:6 to convert 50 into 300ohm.

[glenenglish]

is the twinlead hung in free space, or sitting on a desk ? if on a desk, is it twisted ever couple of inches ? The desk will be enough to unbalance one leg

also, you say "you still have common mode currents". But what is the magnitude  WRT odd mode currents ?

Transformers are not perfect.

Try doing your experiment at 10 MHz with a 1:4 guanella balun from a 75 ohm source (use a  50:75 resistive PI pad) , and then gradually work up in frequency, optimizing each component as you go up in frequency,
-glen

[sjgallagher2]

Baluns are usually specified in terms of impedance ratio, instead of turns ratio. So you need a 1:6 to convert 50 into 300ohm.

Welp, that makes sense! Thanks for clarifying, I had my wires crossed.

is the twinlead hung in free space, or sitting on a desk ? if on a desk, is it twisted ever couple of inches ? The desk will be enough to unbalance one leg

See pictures of the setup. The twinlead is hung between two traffic cones along a bar.

also, you say "you still have common mode currents". But what is the magnitude  WRT odd mode currents ?

Transformers are not perfect.

Based on my measurements so far, which are not very rigorous, the even mode currents appear to be only a few dB lower than odd mode. This is estimated by my measurements with the Rogowski, but it's not really designed for such small currents at relatively high frequencies. I've had better luck with a qualitative measurement using a ferrite bead, but this can't fit around a single wire in a twin lead line, so I don't have much to compare there, although it has better response. Generator is 0dBm, load is 300R, see configuration details in my previous posts.

Try doing your experiment at 10 MHz with a 1:4 guanella balun from a 75 ohm source (use a  50:75 resistive PI pad) , and then gradually work up in frequency, optimizing each component as you go up in frequency,
-glen

I think this is straying from what I'm actually after in this thread, which is an explanation of how to eliminate the mechanism causing the common mode currents. I can considerably reduce common mode currents by tying the center tap of the secondary to the shield of the coax, but I was hoping for something that addresses the actual transition without exciting any common mode. I might have been confused about this, however, and confused about the difference between eliminating the excitation and attenuating any CM currents (they might be two perspectives on the same phenomenon). Without a good analytical model of the situation, I don't have a way of clarifying this for myself.