For as useful as computers are in the modern ham shack, they also tend to be a strong source of unwanted radio frequency interference. Common wisdom says applying a few ferrite beads to things like Ethernet cables will help, but does that really work?
It surely appears to, for the most part at least, according to experiments done by [Ham Radio DX]. With a particular interest in lowering the noise floor for operations in the 2-meter band, his test setup consisted of a NanoVNA and a simple chunk of wire standing in for the twisted-pair conductors inside an Ethernet cable. The NanoVNA was set to sweep across the entire HF band and up into the VHF; various styles of ferrite were then added to the conductor and the frequency response observed. Simply clamping a single ferrite on the wire helped a little, with marginal improvement seen by adding one or two more ferrites. A much more dramatic improvement was seen by looping the conductor back through the ferrite for an additional turn, with diminishing returns at higher frequencies as more turns were added. The best performance seemed to come from two ferrites with two turns each, which gave 17 dB of suppression across the tested bandwidth.
The question then becomes: How do the ferrites affect Ethernet performance? [Ham Radio DX] tested that too, and it looks like good news there. Using a 30-meter-long Cat 5 cable and testing file transfer speed with iPerf, he found no measurable effect on throughput no matter what ferrites he added to the cable. In fact, some ferrites actually seemed to boost the file transfer speed slightly.
Ferrite beads for RFI suppression are nothing new, of course, but it’s nice to see a real-world test that tells you both how and where to apply them. The fact that you won’t be borking your connection is nice to know, too. Then again, maybe it’s not your Ethernet that’s causing the problem, in which case maybe you’ll need a little help from a thunderstorm to track down the issue.
So what’s it mean that it’s only 25dB down at 12:38 when he has disconnected the wire to put some more turns on his ferrite core?
Measuring RF stuff is much more involved then just juggling around with some wires and ferrite cores and he also fails to explain what is really happening and why. It’s just some simplistic measurements and “look, the curve changed”. Maybe there is a better explanation in the second half, but I could not get that far.
Someone else who didn’t watch the entire video…
Unfortunately, HaD commenters rarely do.
Probably because some of us prefer to read things written here, rather than watch videos elsewhere.
There’s nothing wrong with that, just with deliberately looking for things to complain about and ignoring the rest of the content as soon as they think they’ve found it.
Ethernet (the twisted pair kind, not the old co-ax kind) is terminated by baluns, to make it immune to common-mode interference. So a ferrite around the pairs should make no difference either way – the signal is totally differential, balanced.
Ethernet isn’t fully immune from common mode interference. Since the interference is common to both wires, it can sometimes be rejected by the receiver, but only if the interference is perfectly balanced and the cable and system are well-designed. In practice, perfect balance is hard to achieve, especially over long cable runs or in environments with significant electromagnetic interference (EMI).
Shielded cables provide better protection, but only with proper grounding. If the shield isn’t grounded correctly, it can act as an antenna and potentially increase interference. As mentioned in the video, quality of cable can also be a factor.
There’s lot of documented experiences where ham radio operators using high power disrupt their LAN or neighbours network (high RFI environment).
I found a metal strip of 8 sockets which I loaded up with decent LED’s, gobs of light and cheap too. Then I noticed buzz on FM broadcast with the lights on. I did a few turns inside the metal box’s wiring through a ferrite donut. No buzz with lights on or off. Good in my book.
There are as many different kind of ferrites as there are muddy river banks in China (where some of them come from, I swear).
Different ferrites (like the mix 31 and mix 43 he shows but does not describe) have radically different properties for frequency response, permeability (‘inductance’), saturation, loss, coercivity, retentivity & hysteresis. A switching power supply transformer ferrite is not the same as a filter ferrite is not the same as a high frequency RF coil ferrite is not the same as a UHF ferrite, and is especially not the same as a RF circulator ferrite.
Just saying “use a ferrite” without qualifying it with its characteristics is near useless. But go ahead, grab that VNA and random ‘ferrites’ and start throwing proverbial darts blindfolded. You might hit upon a solution.
You obviously watched the 2nd video, but not the first…
As someone who has recently fought with EMI in a 10m test chamber (spoiler – it was a part of the original design I was told was OK…it wasn’t) let me chime in on the benefits of ferrite beads.
They can do wonders, but you need to choose the right bead and put it on the right wire. Then you need to have a way to know if it had any effect. The bead manufacturers do all they can to prevent you from comparing beads from different companies. There’s no standard for mixture naming/numbering, and the beads come in what appear to be random sizes with no common part numbering scheme. Not easy if the original design used beads from one manufacturer but the new design used a different manufacturer. And don’t get me started on the clip-on beads that need a damn tool to remove.
This design had motor noise which resisted all attempts to “bead” it away. What I finally did, in desperation, was to remove two innocent-appearing capacitors between the motor leads and chassis ground…which dropped the noise by 20dB and got us a pass.
“When you have eliminated all which is impossible, then whatever remains, however improbable, must be the truth.”
― Arthur Conan Doyle, The Case-Book of Sherlock Holmes
That video by HRDX is embarrassing … it clearly shows that he knows very little about the subject. There is a real science to making a suitable fixture that can produce valid results, and it doesn’t look like anything he did there.
You obviously didn’t watch the video(s) in their entirety…
I watched the first one entirely. He had no idea how to accurately set up a test like that.
You’re welcome to make your own video explaining it….
I would if I could. The difference is that I know I don’t know how to do it right whereas HRDX incorrectly assumes he does. He is literally just putzing around. There are resources on the web that describe the complexities of building a fixture for two port RF measurements … go find them yourself if you really want to know.
“Everything the poster has said is wrong, only I know how to do [arbitrary technology] right. But I’m not going to give you any actual evidence for my claims, I just want an easy ego boost and don’t care about actually helping anyone other than myself”
You sound like a very smart person that I’d like to listen to
Nah, my only skill is laughing at people making low-effort critical HaD comments. It doesn’t have many wider applications, alas.
Actually, I know enough about how hard it is to accurately test this sort of thing that I know I don’t know how to do it properly either. You should maybe educate yourself similarly.
Ah, so you’re pulling the “do your own research” card.
Check out Whisky Poppa’s post below. The information on K9YC’s website (which I have followed for years) is how I know that neither you nor HRDX know what they are talking about.
With the cheaper, more prevalent SFP connections on the market you can always go with fiber and have a near zero RF network (transceivers have some RF noise).
And this is why i used shielded cable cat 6
Whit a ground at one end
…and a nightmare to terminate. I have never needed STP for Ethernet, but I will acknowledge that there are environments (datacenters?) so noisy that it might be necessary.
Maybe go look at Steve G3TXQ, who got tons of his information from Jim, K9YC, on ferrite.
They actually explain how to make a real vna test fixture. Jim has spent a dozen years plus going through most ferrite mixes available through multiple vendors. . None are the same, even with the same vendor.
He literally wrote the book. Ham_RFI, he wrote the interference section of the ARRL manual, sits on the IEEE, etc.
Way more of a reputable source than a clown with a 20 dollar vna (which has its purposes. I have 3 different nano models as well as a non toy vna).
–Shane
WP2ASS
Let’s make the move and use optical fibre cabling for ethernet.
RJ45 is as dated as BNC/10Base2 used to be.
It’s time to look forwards. 😃
Using iPerf to see if the beads affected the Ethernet? And finding that they might even “boost the file transfer speed slightly”? Ummm ….
Unless the negotiated speed changes or it starts to get errors reported by the interface, it won’t have any measurable effect on how fast the network is.
Does where these are placed matter? (at router? Mid line? Near terminal?)
I have a micro-controlled CNC where the endtops can trigger from EFI that was probably coming from my spindle.
I solved the problem, I think, with shielding but I wonder if any of the EFI was also coming from the ethernet that runs through my shop.
In traditional amateur radio cabling you would add the ferrites at both ends, not far away from the connectors.
In principle, ethernet cables are no different here.
You can also see this with higher quality USB A to USB B cables, as used by USB printers. They have ferrites built-in.
If you want to maximize effectivity of the noise blocking, you can form a little “coil” (cable ring).
You can run the cable twice or trice through the same ferrite clip, if the hole is big enough.
Just don’t squeeze or bend the cable too much, be generous, be gentle.
As a rule of thumb, cables have a bend radius and shouldn’t be stressed too much..
Bending them too much can cause reflections, a bottle neck etc.
Especially coaxial cables in RF applications are affected.
But you’re likely not using 10Base2 cabling anymore, so it’s not that critical. ;)
Still, it’s usually wise to try to not put mechanical stress on any type of cable.
Avoid hard edges. 90° edges may look cool or tidy, but they cause reflections, even as traces on PCBs.
https://www.picwire.com/Files/Technical-Articles/Bend-Radius_PIC_Technical-Article.pdf
If the issue persists, you can use galvanic insulation.
There are ethernet gadgets who have a pair opto-couplers inside.
So both sides, port A and B have mo electrical connection.
There are also optical network cables, of course.
Converters from 10BaseT to optical fibres exist.
You can run your CNC through such a connection and have it separated from your control computer.
In industrial applications, such connections aren’t uncommon.