1:1 RF Transformer Design SWR problems

[ChristianFletcher]

Hi Mike

Many Thanks

At the moment, I’m using a single RG174 cable to form both halves of the transformer. The inner for the primary and the outer braid as the secondary etc. This appeared to work really well so far but I made a change to the core material when I changed the winding style so I need to go back and try a traditional winding style and repeat the testing again to validate what caused the improvements.

The RG175 method puts the connectors side by side and I would rather have greater separation but also maintaining the shortest possible winding length.

DC BLOCKING

I don’t really want to consider alternative solution to the transformer method at moment as I would like keep this thread running regarding optimising the current transformer design. I also want to try and explore some of the potential issues that have be raised by the modelling.

Many thanks for the suggestion I have become sufficiently interested in the issue of providing isolation from external ground that I would like to explore all the alternatives in more detailer and investigate their various merits over my transformer method. I think this would have to be done in a new future thread as this one has already become very long.

I may open a new thread in the future and offer up these findings as I guessing lots of people have potential issues with PME.
Would you confirm if you are suggesting that I put one capacitor in series between input and output braid and another in series between the input and output centre conductor. Just straight DC block. I think this is often used when power is sent over coax along with the RF so your suggestion must work.

Best Regards Chris

[G0HZU_JMR]

I stuck some connectors on my old grey ferrite transformer with 50R twinline on it and tried measuring the cable leakage when this thing was inline.

I used various probes and various methods and the picture below is just one of them. In this image I'm sending a signal through from port 1 to port 2 on the VNA and using port 4 as a sniffer (blue trace)

The whole of the cable is lively with RF when the xfmr is in line and you can see the difference in the two blue traces as the stored trace shows what I see with the transformer inline and the current blue trace shows how much lower the pickup is from the cable without the transformer. Easily 30-40dB better. Note that this is with good quality RG58 cable. If I'd done this with cheapo RG58 sold at rallies or in CB shops then the leakage up at the top end would have been much worse in the cable only plot. You can also see a peak at about 12MHz in the leakage. I think this is the frequency that the cables and transformer are resonant with a direct ground loop and leaking the most.

Note that it isn't just a case of the transformer not being screened. The whole cable is leaking RF along the braid, even if I probe near the VNA. It doesn't seem much different in leakage if I probe very near the transformer itself. Adding ferrites helps a bit but not across the whole band. But I only tried a couple of basic ferrite clamps.

I also tested the setup using my spectrum analyser (with a screened preamp) as a receiver and ran a cable out of the window into the garden. This breaks the ground loop inside the room and grounds the cable outside. I put a known noisy laptop charger in the next room and compared the leakage with/without the transformer inline and I saw the same 30-40dB difference in RFI from the laptop across 2-30MHz. It's not just the transformer picking stuff up because if I use very short cables with the transformer the RFI problem almost vanishes. It is the whole cable and transformer that is leaky because the screening integrity of the cable has been breached across a significant bandwidth.

[G0HZU_JMR]

I also tried using a portable SW receiver and I sent a signal up the cable at about 18MHz with 80% AM and then compared the difference with the transformer in and out. The difference in leakage was probably about 30-35dB but it's hard to tell. I measured how far down I could turn the sig gen and still hear the interference when the transformer was inline. The I took it out and measured how much louder the sig gen had to be to hear the signal with a straight cable. Note that this was with some cheaper RG 58 cable.

Maybe none of this matters. In my youth I've seen CBers use a matcher/wire to various items like a domestic appliance as an antenna. You would be surprised what they did in those (CB craze) days if they lived in a flat or bedsit with no space for an antenna. But they seemed to get away with it with no serious RFI issues.

[ChristianFletcher]

Thanks, Jeremy

That’s very interesting and looks pretty bad more than a 1000 times the leakage with the transformer in the circuit. I hope to repeat similar test with my setup. I wish I had a spare core to send you but I don’t have a spare for type 43.

I’m just wondering how much RF leakage was from the coaxial cables itself and how much was from leakage was from the transformer and the fact that parts of the transformer assembly and hook wiring was unshielded. I think you said you used a RF sniffer so I assume you did confirm that main leakage was from the coaxial cable itself. I’m not disagreeing just asking your opinion.

I don’t think I will have time to do any testing again until the weekend, but I will try and film it so you can see how I get on testing my transformer.

Its amazing what you can use as an antenna and will match I have worked much portable dx on my 5 watt 817 using an old barbed wire fence as an antenna.

Chris

[G0HZU_JMR]

I dug out some very short BNC cables and with my ferrite based transformer with long windings , the suckout/radiation is still in band when the cables are this short. I did a quick youtube video, see below. You can see that the resonance has shifted up quite a bit because the cables are shorter. You can see that the RF radiates away about the same all along the whole cable assembly and you can see that the rate of decay is slow as I move the probe away. This shows it is radiating and not just the reactive near field.
You can also see me put a ferrite clamp on the cable and this kills the Q in the ground loop. But you can see that the cable is still leaky with a flat highpass response.

If I replace the transformer with a thru adaptor the leakage is much better although there is still some near field leakage if I hold the probe very close to the BNC connections. But I think this is just near field as it falls away at a very fast rate as I move the probe away. The leakage from the xfmr+cable looks like radiation because the rate of decay is very gradual as I move the probe away.
The frequency range is 0-50MHz and the marker is at about 22MHz. The yellow plot is S21 (thru) on 2dB/div, the purple plot is S11 (return loss) on 5dB/div and the other plot is the probe and it measures S41 on 10dB/div.


https://youtu.be/uTNEdhiv9oQ

Obviously, this is an artificial scenario with a perfect ground loop but you can see from the yellow S21 trace that a significant amount of the energy is leaving the system via radiation at 22MHz if there is a decent ground loop in the system with this transformer in line. Over half of the power?

[ChristianFletcher]

That looks quite conclusive. I still would to try the same test on my transformer. I assume I can use the spectrum analyser tracking generator to feed a dummy load via the transformer and then make a loop connected to the spectrum analyser input.

Just out of interest what was level output was coming from your generator. I wish I had a nice probe set, I'm going to have to spend some money with Rigol.

you have a really nice setup. do you have normalise the loop input to calibrate the frequency response ?

Thanks Jeremy great work and video

Thanks

[G0HZU_JMR]

I can't be certain of the level. The VNA was probably set to +6dBm to get a bit better S/N but it defaults to 0dBm. It's all relative anyway. The display is in dB.

If you try and replicate this make sure you touch the ground of the flying dummy load to the analyser to complete the ground loop or you won't see any peaking effects. You could probably do this test using a simple loop wire as a probe. The H field probe I used in the video is homemade using semi rigid cable. I've got a few of them all different sizes. They aren't cheap to buy so I made my own. I think the frequency you may or may not see some peaking depends on the cable lengths and also the physics of the transformer. But all this is an artificial worst case test of an ideal ground loop.

Once you test your real setup with your real antenna grounding etc the peaking may not be an issue but I don't think the screening will be as good as unbroken coax.

You could just fit ferrite to the cable to get rid of the worst of this issue I guess but I still don't like the idea of a broken screen and one of these transformers part way up an unbalanced cable. But it did seem to work well in your video on the ham bands.

[G0HZU_JMR]

Originally Posted by ChristianFletcher

At the moment, I’m using a single RG174 cable to form both halves of the transformer. The inner for the primary and the outer braid as the secondary etc. This appeared to work really well so far but I made a change to the core material when I changed the winding style so I need to go back and try a traditional winding style and repeat the testing again to validate what caused the improvements.

One obvious thing to be wary of is that the coax (acts as a transmission line) will have a velocity factor of 0.66 typically and the enamelled wire (also acts as a transmission line) will have a velocity factor affected by the winding style (is there an air gap or not?) and the thickness of the enamel(?) wire coating and its dielectric constant. If you end up twisting the wires closely together I think these coating related issues become more significant. So be wary how you choose and compare your wire lengths because the actual electrical length may be quite different to the physical length.

[ChristianFletcher]

Thanks Jeremy

In the very first transformer I tried I had the wires twisted together and this worked really badly even worse than the details that I first published. Winding side by side work better but we got the voltage imbalance doing the 25 ohm resistor test.

The coax I think worked best and I did not get the voltage imbalance until 30 Mhz. Unfortunately I did change the core material at the same time.

Just asking about the output power. I think I can go upto 0dbm. I wanted to be sure the problem would be bad enough for me to detect at my power level.

I had twigged based on your earlier comments about grounding the load. I suspect my auto ATU is very poorly grounded and perhaps that why I am not getting the same issues when driving with the transceiver. However I do actually want to ground the ATU. So this need to be replicated as you advise regarding my test setup. I was thinking of assembling everything on a ground plane copper sheet.

Thanks Again Chris

[G0HZU_JMR]

There are a few factors to consider when comparing twisted or 'not' twisted. I think that the dielectric constant of the enamel will begin to dominate if you have the wires twisted snugly together. So the transmission line will slow down and it will be electrically longer. The characteristic impedance Zo will be lower with twisted wires which should improve the performance but the electrical length will be longer (not so good, and will tend to spoil the benefit of the lower Zo).

I think the difference in electrical length could easily be 20% if you compared 'flat side by side with a small air gap' vs 'twisted' because the twisted pair line will be slower in terms of velocity factor. However, if you shorten the wires and then twist them to get the same electrical length for the tline, the shorter 'physical' wire length means that the choking impedance with the ferrite will be less so the performance at LF will begin to degrade.

Also, if you start using different wires with different coating materials and different coating thickness then you could see a significant variation in electrical length and Zo when you twist the wires. It could get quite confusing when you start testing them against each other...

[G0HZU_JMR]

I had a rummage earlier in the loft and found an old ferrite ring that was used as a TVI choke. I used to give these to neighbours when I was active on the bands in my student days. I recall one neighbour saying "They look just like an exhaust gasket" when I told him to wrap it into his TV coax This one must be 35years old. It's a bit small and a bit skinny but I followed my own advice given in my second post on this thread and wound two windings in parallel using twisted pair wire. Having two tline windings in parallel halves the Zo.

The VSWR into a floating 50R load is given below. The equivalent tline model predicted 1.38:1 at 28MHz and 1.6:1 at 50MHz if I allow a few nH for the connection wires. As you can see, the measured response agrees with the model and also you can see that the LF performance has degraded as predicted. It's just under 1.2:1 down at 3.5MHz. I think your RG174 transformer was up around 1.8:1 at 28MHz and 2.8:1 by 50MHz by comparison. But I think your match down at 3.5MHz was a bit better. I'm tempted to try a triple winding to get the effective Zo even lower but I think the LF performance would degrade too much.

Note that I built this to do well in your VNA VSWR test. But this doesn't meant this transformer will work better in your actual application. I made this point earlier in the thread. The VNA test into a floating 50R load is a bit artificial and a transformer optimised to work well here might not be optimal for your actual installation.

You could try experimenting with 25 ohm (or lower) coaxial cable if you can find some? You'd have to check the ratings obviously.

[ChristianFletcher]

looks goods to me. Can you post some pictures of your transformer.

Regards Chris

[G0HZU_JMR]

You just caught me in time. I was in the process of taking it apart but I've put it back together with some connectors next to a ruler. See below.

Before you get too excited, the ferrite material is very lossy. This was a ferrite ring used for TVI suppression so it is not designed for transformer use like this. It's all I could find in the loft. The loss is quite poor (0.7dB) and I think it would get quite hot at high power. I'm back at work on Monday and I might be able to find something less lossy.

To prove that it is the ferrite, if I tie and float the gnd ends of the transformer in free air such that the transformer is no longer shunting to gnd but is a floating (shorted) tline the loss drops to a negligible level. So this shows that the ferrite is the problem. The VSWR gets slightly better like this across the whole range too. Also, it works down to LF but it isn't of any practical use like this. This test is just to show the contribution to loss from the ferrite. It also shows that the transformer windings are behaving like a transmission line.

[G0HZU_JMR]

I just took one winding out so the Zo is now doubled. The model predicts the VSWR at 28MHz to degrade to be 1.58:1 and the VSWR at 50MHz to be 2.2:1. The actual measurements on the VNA were 1.55:1 at 28MHz and 2.05:1. at 50MHz. Pretty close!

This was the real reason I made this transformer so I could show you the changes wrt a change from a lower Zo by removing an identical winding in parallel. I tried to do the above demo with a degree of test control between the two transformers. When you asked for the picture I was part way through unwinding one of the windings but it was easy to put back

It might not work so well in every case but you have to be able to get that lower Zo without compromising something else. So I think you may end up having to try quite a few things and you may end up going back to RG174 or RG58 at Zo =50R just for the simplicity. Note that if you do try out the parallel winding method it is very easy to get in a muddle and wire up the 8 free ends incorrectly. Obviously, if you make the slightest error here the transformer will behave oddly.

[ChristianFletcher]

Thanks Jeremy

I was a bit confused but the picture is great. Hopefully I can do some of my own testing this weekend and we can compare notes.

Chris

[G0HZU_JMR]

I had another rummage and found some big ferrites in my EMC survival kit. They are cylinder shaped and not really ideal in this respect. But I transferred the windings to one of these ferrites and I also added another winding in parallel. So now there are three parallel twisted pair windings each of about 50R Zo. It looks really ugly and I deliberately wound the three windings sloppily to help show that the three lines behave as transmission lines.

The insertion loss is about 0.13dB at 30MHz and is now what I would expect from this tline topology with a decent ferrite. Ideally, I would have used thicker wires but I don't want to waste my limited supply of thicker En Cu wire. The VSWR is as attached. You can see that the VSWR is improved to 1.2:1 at 28MHz. This agrees exactly with the tline model prediction for a Zo of 16.7 ohm and a tline of this electrical length. I still don't like what you are doing with this transformer configuration but hopefully I've showed you that you can improve the VSWR with a shorter line and also a lower Zo and a switch to a suitable ferrite. All of this info was in my second post on the thread. But here we are at post #116

[G0HZU_JMR]

The other thing to watch out for is that (in theory at least) there is a risk that you will end up with a ground loop 'suckout' effect that lies on top of a ham band and this will depend on your transformer and also your coax length and maybe even your ATU setting. If this happens and you transmit 100W there is the risk that the ferrite transformer could overheat very suddenly to the point of some serious smoke. I doubt that this will be the fault of the transformer on its own, it will be due to a system resonance due to grounding, cable length, transformer physics and operational frequency. A bit like playing Russian Roulette I guess. You could probably reduce the risk a fair bit by adding some ferrite clamps but I still don't like the idea of this thing part way along some coax.

[ChristianFletcher]

thanks Jeremy

Yes we kind of got there in the end. I'm going to try and continue with the design but also try the other methods that have been suggested. I suspect that I'm probably not seeing the suck out because the ATU is so badly earthed. But I will watch out. I kind of want it to go wrong because the would be interesting

Regards Chris

[G0HZU_JMR]

One thing I noticed with this latest xfmr of mine is that the resonant suckout is very small even on short test leads. On your test with the Rigol you saw about a 15dB dip and that seemed quite alarming if it was affected by cable length. With my latest transformer, the dip is only 2dB with very short leads and it falls to about 0.2dB with 1m long leads. With even longer leads it is hard to see it. But the cable is still leaky to RFI when the transformer is inline. But you might find that the leaky braid might not matter once you attach a real antenna because the antenna signals might be stronger than the RFI anyway

I did wonder if you could do it with an isolation balun to twin feeder and then have another balun at your ATU/antenna. But I'm not sure if that would bring its own set of issues. I really don't have the experience with respect to HF antennas and the options for feeding them.

[ChristianFletcher]

https://youtu.be/bLwpLC2Xib4

Made a quick video showing some testing on the transformer. fitting the ferrite clamps appears to get everything back under control. Also worked for near field measurements but forgot to film this. Don't think I'm in for an Oscar this time.

Regards Chris