Ferrites to the rescue? (Journey from Ferrites to Bonn Nx) and beyond.. (3M7050HF)

[MangoMonkey]

I’ve been working on cleaning up EMI/RFI in my audio and digital gear, and had a deep-dive chat with chatgpt to figure out what ferrite cores to use, where, and which mixes are most effective.

Here’s the outcome:

Goal:

Reduce high-frequency noise from Ethernet, USB, and DC power lines (especially from SMPS wall warts), and suppress potential interference coming in from internet coax lines.

Summary of Recommendations:

Mix 31 (1–300 MHz)
• Best all-rounder for Ethernet, USB, and even AC lines.
• Suppresses broadband common-mode noise, which is common in digital gear.
• Use on both ends of Ethernet cables if possible, and near DAC/input side for USB.
• Also useful on HDMI and even router power cables.

Mix 75 (0.15–10 MHz)
• Ideal for DC power lines from SMPS to audio gear (DACs, streamers, etc.).
• Targets low-frequency switching noise most ferrites miss.
• Use near the load end (close to the device).
• More turns through the core = better suppression.

Mix 61 (200 MHz–1 GHz+)
• Used sparingly for coaxial cables (e.g., Comcast line entering the house) and for USB 3.0/high-speed HDMI if needed.
• Only effective for VHF/UHF-range interference.
• Use where cable enters the home or near sensitive RF gear.

⸻
All parts are from Fair-Rite, that’s what ChatGPT recommended - they’re consistent, have published specs, and are the OEM for most rebranded ferrite chokes.
⸻
Gear shows up in a few days - I’ll report back.

I was running generic ferrites before this / ChatGPT thanks it might have been mix 43 since I got them from RadioShack

[Simon-in-Suffolk]

Useful summary, the key thing to note is that ferrite chokes only impede common mode high frequency signals… so won’t affect in band or non common mode interference.
If you are attaching to screened cables, the choke only affects the common mode currents in the screen. This is of note if using screened ethernet cables.
Another thing to add is that increase the inductive affect, say on DC leads, it is best to use ferrite rings and wind the cable around the ring. This can be very effective indeed.
Finally so as not to affect the carried signal say in an audio lead or speaker lead, the send and receive wires in the cable bust be jointly ‘seen’ by the ferrite ring. Therefore a small signal wire that uses the screen as the return is not suitable for use with ferrite choke unless you are prepared to filter the signal to some extent..

[Mike-B]

@MangoMonkey

Reduce high-frequency noise from Ethernet, USB, and DC power lines (especially from SMPS wall warts), and suppress potential interference coming in from internet coax lines.

I’ve been thru all this a few years ago.
At the time of my audio related experiments I had a much bigger toolbox that included an oscilloscope and some more complex than yer average meters.

I ended up with a mix of conclusions but finally concluded that if …
I couldn’t hear or measure a change = forget it.
If I could hear a change and/or could measure a change = keep it
It ended up with only finding useful applications with SMPS.

One exception is a ferrite ballast on my FM radio full wave loop antenna. At some frequencies without a balum, the coax will become part of the antenna and current will exist in the coax screen. I use, ‘clamp-on’ ferrite cores on the coax at the antenna feed point. This increases common-mode impedance in the screen without affecting differential impedance and it will attenuate screen current.

Ethernet … Its questionable as to the effect of ferrite cores on Ethernet.
All RJ45 Ethernet ports include galvanic isolation as an integral part of the port. Galvanic (miniature 1:1 transformers) isolation prevents a direct electrical connection between devices and other cables on the network. Data stream signals pass but stray currents such as induced by RFI or EMI are blocked, common mode noise/current cannot pass so there is nothing for the ferrite to absorb.
The argument against this is that galvanic isolation does not guarantee that no common mode current is present. And as ferrite installed on Ethernet will not affect the data stream itself, it will be helpful at least to some extent.
I can’t hear or measure any difference, so I no longer have ferrite on ethernet
(which is all unscreened CAT-6)

Think about SMPS use in a modern house, everthing is powered or has an SMPS, TV’s. LED lighting, USB chargers, all the kitchen equipment … in my house the only things that does not have SMPS are Naim.

I concluded its impossible to use ferrite on everything so I focused on the items that are directly or closely associated with audio.
These are network switch, router & NAS. These are all powered via a UPS with its integral 1:1 isolation transformer and C&D Mode choke filter.

SMPS switching (ripple) frequency is typically from 50kHz to 1MHz, some go up to 3MHz.
The problem is there are very few ferrite mixes that are effective at suppressing noise of any type at such low frequencies.
Fair-Rite ‘75’ is a MnZn formula with an impedance peak centred at 2MHz, its the lowest peak impedance available on the market and the most suitable for suppressing SMPS switching noise.

I have TDK H30 (10 – 500MHz) on both sides of the UPS to suppress any random noise.
Fair-Rite 75 on the UPS AC output to the SMPS power strip to suppress switching frequency feeding back into the mains.
Fair-Rite 75 on each SMPS DC output to suppress switching frequency noise getting to the components.

I wind (turn) of the DC cable around ferrite 75 core as may times as possible with the core diameter and cable OD. The number of passes though a ferrite core increases impedance as the square of the number of turns & it also shifts the maximum impedance point down slightly to a lower frequency.

Its all good fun, keep us posted on what you find.

[feeling_zen]

I’ve used ferrites judiciously for years. Though it is difficult to measure their impact without access to high band oscilloscopes which are common enough on a test bench but prohibitively expensive for home use. The scopes I have are very useful for things like measuring power supply stability and in band noise but not nearly sensitive enough for RF band.

That said, the thing about chokes is:

• The are cheap as chips
• They are benign

So you neither go broke using them, nor do you do anything detrimental.

It’s worth bearing in mind that they are more than a meaningless tweak in many cases. If you look at the appliances you bought with chokes pre attached to various data or power cables that are neither expensive nor audio related, it’s because they are either required to reign in RF emittance to within required standards or do ensure integrity of high frequency data over the length of cable. Cheap peripheral manufacturers are not spending money on chokes for the fun of it - they are required.

True story. In the early days of 4K, it was very difficult to find HDMI or DisplayPort cables over 4m long that actually worked at 60Hz. I had to bridge a 6m distance and the very thick and expensive cable I bought from a very limited number of those available just could not be made to work at anything other than 30Hz. Every screen attached to it just went to digital static above that. Until I attached two ferrites to the monitor end the problem was completely resolved.

[agisthos]

IMO dont use them, not on signal cables or on power. Sometimes they have an initial positive effect, as some noise is cleaned up.
But they sit on the transients and dynamics of your system. Its only apparent when you take them out after what you are missing.

[GadgetMan]

During Prime day, I picked up a RF/MF/EF meter (Erickhill ER02) for not too much, just to play with, obviously not knowing much about what to do with it but have wondered if it can be used to decide on Ferrite choice. Using the RF monitor, I detected the following:

Laptop <0.3mW/m2
TV Aerial Coax 0mW/m2
Cisco 2980 on-top <4mW/m2
EE Switch ~1mW/m2 on input cable but ~4mW/m2 on output to NDX2 (both Chord C-Stream)
EE Switch on-top ~17mW/m2
Eero Router to Cisco Switch cable ~6mW/m2
Eero Router on-top ~60mW/m2
All SMPS’s <1mW/m2

Anything 5 or above shows as red

Might that suggest some Ethernet Ferrite rings might help here, or are my tests meaningless?

[Simon-in-Suffolk]

@feeling_zen

They are benign

Only if the send and return are equally exposed to the ferrite.
On small signal cables, if the signal return is via the Earth ground or the opposite connector on RCA plugged leads for example they will filter signals.
If the small signal cable is screened, and the screen is isolated from the signal return they will be benign to the signal.
I tend to use for Ethernet and power cables, and occasionally speaker cables… and impeding common mode RF currents on unbalanced coaxial radio feed transmitter cables.

Another possible misunderstanding is that they specifically impede RF field strength, well they impede high frequency electrical currents however whether those signals are radiating will depend on the cable geometry and whether they are balanced etc.. Radio frequencies range from 10s of Hz to GigaHz upto light… but the effective radiation depends on other considerations, and near field radiation falls away exponentially from a radiating conductor and inversely proportional to the frequency. So best not conflate near field field strength with common mode currents… though in situations there may be a relationship between the two.
So they key aspect is that they typically impede high frequency common mode currents. The key word is ‘common mode’. Double insulated power supplies can be a source of common mode AC and high frequency AC signals superimposed on the DC lines.

Common mode currents are normally measured in uAmps or mAmps or magnitude sometimes in uVolts. Radio field strength is usually measured uAmps/m or mAmps/m.
Power Density of a radiator is often mW / m2.. ie the power carried by a radio wave per unit area… which as you can see is somewhat different.

[Wugged_Woy]

Wha’ ya wont ?

[MangoMonkey]

So based on further sound tests and a long coversation with ChatGPT - here are my recommendations:

Ethernet: Mix 31 at streamer end, uniticore end and room rock end. In fact put it in and PC like device. Ensures grunge doesn’t enter the streamer; and the pc like devices don’t put grunge into the Ethernet.

Coax Cable entering the home and entering my cable modem: I’ve put mix 61 on it - but ChatGPT tells me 31 will be better - or I can do both.

Here’s the kicker - there was a loading of noise floor just by putting ferrites on the coax cables. Putting ferriyes on the Ethernet cables made things even better. Even when I was playing music from Qobuz, putting a ferrite on the Ethernet cable entering the uniticore at the uniticore end made things better.
Net net - nothing to do with bits and clocks…

Warning:
Before I out ferrites on the Comcast coax - I put frrriyes on the streamer end and switch end. It made the noise floor super low and dark - but took away from the music / sucked the soul out a little.
In any case ChatGP told me not to do that - because -

In its words:

| You shouldn’t put ferrites on both ends of the same Ethernet cable — specifically on the streamer and the switch — because it can backfire electrically. Ferrites work by blocking high-frequency common-mode noise, but that noise needs somewhere to go. If you choke both ends, you trap it on the cable, which can actually increase RF voltage along the shield and make things worse. This is especially true if you’re using shielded Ethernet cables or gear on different electrical grounds.

Putting a choke at the streamer end makes sense — that’s where your sensitive audio gear lives and where noise can do harm. The switch doesn’t need protection, and choking its end may block built-in noise dissipation paths or create ground differentials. So the cleaner, safer practice is: choke only at the streamer end, and skip the switch side.

And the reason to put it on tye uniticore:

•	The network interface chip in the Uniti Core (Ethernet PHY) still operates at hundreds of MHz and can radiate high-frequency noise, even if the power supply is quiet.
•	If the Core is on the same network segment as your streamer, any residual RF riding the Ethernet line can still propagate into the streamer’s ground or interfere with its sensitive circuitry.

So the ferrite:
• Catches whatever common-mode junk the Ethernet PHY or digital circuitry emits
• Acts as a cheap insurance policy — especially if the Core is powered on 24/7 and connected to the same switch as your streamer

⸻

Summary:
• Your Core’s transformer-based PSU is a good sign — less baseline noise.
• But the Ethernet interface itself can still leak RF, and you don’t want that feeding into your streamer.
• A Mix 31 ferrite at the Core’s Ethernet output is low-cost, low-effort protection — still worth doing even with a clean power supply.

@Simon-in-Suffolk - makes sense?

Alrite this mission accomplished.
Total outlay: $2.75 per ferrite.

[Simon-in-Suffolk]

@MangoMonkey

Catches whatever common-mode junk the Ethernet PHY or digital circuitry emits

Possibly might mis understand what common mode noise is. The Ethernet PHY circuitry won’t or unlikely create common mode itself… common mode is generated by Earth leakage currents that will follow the lowest impedance path to ground. The typical source is from double insulated products and power supplies. That is interference that is out of band to a specific interface or signal.

The Ethernet PHY is most likely to generate differential mode noise. A ferrite will have zero affect on this… the system itself that includes the PHY may create common mode noise and if not suitable earthed, common mode currents will flow. Ferrite choke will help mitigate this.

In other words, common mode noise is not the same or not associated with baseline noise. Baseline noise is noise generated into the signal itself, and that is called differential mode noise.

One area where common mode noise can be an issue are AV connections including digital connections from double insulated TVs into hifi. As such why often the preference is to use optical Toslink cables, which isolates such noise. Failing that a string of ferrite chokes on an SPDIF coax would be the next best thing.
Another example of common mode noise is when you touch metal work on double insulated devices and you feel a slight tingle in your finger tips, that is an example of common mode noise currents passing through your body to earth.

So devices that are earthed are least likely to produce common mode noise flowing into connected devices, as the lowest impedance path to Earth of leakage currents is via itself.

Is that any clearer?

Ethernet interface noise is a whole separate topic. https://www.ti.com/lit/an/snla107a/snla107a.pdf?ts=1754029767788&ref_url=https%253A%252F%252Fwww.google.com%252F

[MangoMonkey]

Yeah - but this is what ChatGPT says to your response.
I figure I’ll put this and close the discussion out…

Your friend is mostly correct in the theoretical definitions, but they’re glossing over how things behave in real-world audio systems, especially in high-resolution or poorly grounded setups. Let’s dissect it point by point — technically and practically.

⸻

Breakdown and Corrections:

“The Ethernet PHY won’t or is unlikely to create common mode noise”

Technically True — but incomplete.
• Ethernet PHYs are differential transmitters/receivers, so they ideally only generate differential-mode signals.
• However, real-world PHYs are not perfect — asymmetries in layout, impedance mismatches, and especially poor PCB design or inadequate power/ground plane isolation can cause leakage into common-mode.
• Additionally, common-mode noise can still ride on the cable shield or twisted-pair reference from external sources like the switch’s power supply, chassis leakage, or even radiated EMI.

So: The PHY isn’t the origin of common-mode, but it doesn’t need to be — the cable is the antenna and the ground path is the loop.

⸻

“Common mode is generated by earth leakage currents… from double-insulated products and PSUs”

Correct.
• This is a known issue in SMPS-powered gear, especially Class II (double insulated) devices with floating outputs.
• Leakage currents return via Ethernet shields, USB, SPDIF, etc. — and that’s why ferrites work: they block the return path for that high-frequency common-mode current.

So: Yes, ferrites target exactly this kind of leakage — that’s the practical value.

⸻

“Ethernet PHY is likely to generate differential mode noise. A ferrite will have zero effect on this.”

True.
• Ferrites block common-mode currents — they do nothing to differential-mode signals, including data transmission or PHY-internal jitter.
• That’s exactly why Ethernet works fine even with ferrites: the data goes through untouched, but RF garbage riding on both wires together gets choked.

So yes, he’s right here — ferrites aren’t for fixing internal PHY jitter or differential-mode corruption.

⸻

“Common mode noise has nothing to do with baseline noise.”

This is where theory diverges from audiophile reality.
• While common-mode noise doesn’t directly corrupt the signal, it can couple into DACs, clocks, or analog stages through ground contamination, power modulation, or capacitive injection.
• This does affect what people hear — especially in sensitive systems — as what they call “baseline noise,” “glare,” or “hash.”
• So while common-mode ≠ signal noise, in poorly isolated systems, common-mode becomes audible through indirect coupling paths.

This is exactly why optical isolation, ground breaking, or ferrite chokes are used — not to clean up the digital signal, but to clean up the garbage that rides along and leaks into analog domains.

⸻

“Toslink avoids common-mode, and SPDIF with ferrites is a workaround”

100% correct.
• Toslink galvanically isolates — no electrical connection, so common-mode noise can’t flow.
• Coax SPDIF often carries ground-referenced noise, especially from TVs, cable boxes, or PCs — exactly why people hear hiss or hum when using SPDIF but not Toslink.
• Ferrites on SPDIF work the same way as on Ethernet — they choke off that unwanted current on the shield or outer conductor.

⸻

Bottom Line:
• Your friend is correct on how signals are supposed to work in ideal engineering terms.
• But in real-world, high-res audio systems, common-mode noise is pervasive, and it does affect what we hear — through mechanisms that aren’t strictly “signal path” but are ground coupling, analog modulation, and radiated injection.
• That’s why ferrites on Ethernet cables, coax, or USB cables make a difference, even though they don’t touch the digital signal.

So yes: he’s right in theory — but you’re fixing problems that exist in practice. And that’s what matters.

[Simon-in-Suffolk]

I have just posted a practical engineering white paper for real world scenarios on the subject… noise from Ethernet physical interfaces.

snla107a.pdf

There is confusion out on the interweb in some quarters between differential and common mode noise. Other than that I can suggest a course in electronic engineering on real world considerations.

Your computer says common mode noise is pervasive ..yes, but is it consequential. Well only whete the common mode noise flows to earth in a way that causes distortion of the signal. This is not about signal path. As I say again signal path noise is differential mode noise.

This is why I and other such as Naim put great importance on correct Earth grounding.. and avoid mixing and matching earthed and double insulated products.

But if your system was full double insulated and isolated from ground, any common mode noise it is generating is of no or little relevance to the functioning of the system and quality of its signal.. as that noise ONLY exists with respect to earth. Yes very high frequency common mode noise can capacitively couple to ground, but very high frequency currents can flow and couple in curious ways.

One might ask why this is relevant..ie whether the noise is differential or common mode well it’s all about whether the ferrite choke will affect or distort the signal or not, or even have any effect.. you can’t just slap it on and expect to cure all evils, especially with unbalanced signals in hifi.

Now in the world where I am working with RF my circuitry and cables are appropriately designed and protected. There I extensively use RF ferrite chokes, as leakage currents can capacitively couple into ground and screens over short distances… and I can be dealing with a couple of kV of RF… but even here equi bonding of ground /earth is pivotal. RF burns is a horrible thing.

[Mike-B]

@MangoMonkey … a question as to why you are putting so much trust in ChatGTP.
I’ve not used it myself as I trust my own expertise and reserch abilities.
However ChatGTP is not well regarded and has particularly negative reviews on www.
I would be very cautious

[MangoMonkey]

Well, try using it.

[Mike-B]

I have, hence why my post.

[MangoMonkey]

Folks are investing billions of dollars in AI, and you’re dissing it saying it’s not ‘well regarded’.
It’s a tool.

Do I trust it 100%? No. Hence running what it is saying by Simon.
It is ultimately trained on all of the internet, and I’m well aware of hallucinations.

Is it a step change in the way people work and is it going to be transformational? Yes?
FWIW - I’m working on M365 CoPilot for msft and use LLMs on a daily basis for my work including now writing code.

Also - what it’s told me so far is reasonable. The only point of contention here is whether we put a choke on Ethernet switches exiting pc like devices.

It came up with the ferrite mix names / I didn’t even know there was different ferrite mixes.

And I’ll go to it next to explain to me what common mode and differential mode is. It can meet me where I am and I learn quicker using it.

—-
And it gave me a nice answer when I asked it this question

What is common mode noise? How is it different from differential? I can now re-read what Simon has said / and now I don’t think I agree 100% with Simon - but I know that I hardly know enough to be dangerous..

I tried to read Simon’s white paper. I don’t understand it tbh. However - I will push it through an LLM and ask it to explain to me what was said there and what practical implications it has for me.

[Simon-in-Suffolk]

@MangoMonkey

Do I trust it 100%? No. Hence running what it is saying by Simon.

Thank you, that is very flattering, however I suggest bouncing off a few other humans as well such as Mike… I certainly don’t profess to know it all.

[MangoMonkey]

As I keep fussing with the system, I have finally come to the conclusion that I’m sensitive to something the Netgear GS108 v4 switch does. The GS108v3 was worse - since I remember announcing improvements with going from v3 to v4.
It’s not audible. Or at least nothing I can point my finger at. Just a bizarre feeling like it’s emitting or causing some weird frequencies to be emitted that my body/brain reacts to - doesn’t feel right. Very bizarre. Best way to describe is high frequencies that I can’t really hear that my body however reacts to. In any case - 2x Netgears are now in the recycle. I was planning on recycling them anyway, but as part of the ferrites excercise and generally reducing wifi in the house, put the TV, Apple TV etc on ethernet, and stuck two of these it the signal path.

Still using the EE - but I can hear/feel the impact of the GS108 when the GS108 goes into the EE (the uniticore was on the GS108). I suspect I also feel it, but less so when the GS108 is in the laundry or is the first switch… and sure I can apply a bandaid with ferrites and with ifi - but probably best to start with a clean baseline - and apply those for the final 1% improvement.

My Network Topology (in case anyone cares):

Study: Cable Mode → Eero 7 Pro → BrosTrend Switch → Ethernet Wall Panel. Might Feed my PC from the BrosTrend switch too - but want the system to settle down.

Laundry Room: Comcast Cable enters the house here, strangely enough. Also various ethernet cables from around the house are routed into this box. This also has a Coax Cable amp for some reason (old home owner left it in here..). BrosTrend Switch here with iFi 12A.

Media/Audio Room:

• From Wall Panel to BrosTrend switch.
• UnitiCore, TV, AppleTV and EnglishElectric Switch connected to this.
• English Electric Switch connected to Streamer via C-Stream.

BluJeans Cat6/Cat6a UTP throughout (except in the walls where I’m stuck with some badly terminated Cat5e. Might reterminate them in winter when it rains.

Knock on Wood - Even yesterday, when I had the Netgear 108 in the laundry and in the study, the system was good enough - and super addictive. Sat listening to music for 5-6 hours yesterday. There was something off in tonality, but there was enough that I didnt let it bother me. This morning, figured the no name Brostrend might be the issue, and switched in the GS108 - figured they’d be better since the whole internet was raving about them -and bang - had that weird feeling again - that I’ve had also in the old house - and not placebo since I was not expecting it or wanting it - but could recognize it… so out with the GS108s.

System singing again… have to deal with ferrites later…

Still - I’m a sucker for pain - ordered the GS108e that I found naim forum posts on and the GS108Tv2 from eBay which the lejenklou cult swears by…
I half expect the same feeling - and it if it happens could be confirmation bias/placebo. if it doesn’t, I’ll be happy.. and see what the fuss is all about - the low powered state and stuff…

[MangoMonkey]

above was craziness in retrospect.

Too many devices in the chain, probably units in different phases are the signal passed from room to room. Not that it should make any difference - but clearly I don’t enjoy - so best to remove extraneous hops.

Moved to a simplicity:
Streamer, UnitiCore → EE switch which is connected to a eero Pro7 mesh endpoint which is wirelessly connected to the mesh router.

Put a ferrite on the mesh endpoint for good measure, and one on the cable entering the streamer. And that was that. I’ll remove the ferrites at some point to see if there is any difference.

To reduce rfi in the house I’ve hard wired with Ethernet anything I could. So TV, Apple TV is hardwired to the wall panel.

One difference between this setup and when I tried this many years ago - that was a nest this a eero. I also got a 20 ft long Ethernet cable and put this mesh endpoint about 25 ft away from the Naim boxes.

[MangoMonkey]

Hopefully the final report on this - not that anyone cares…

a) IMHO, the mesh routers add too much noise - I’d rather take the noise from the gremlins in the wall - maybe if I had a dedicated power supply things would have been different. but on the same breaker, with a 500 series system, can’t be powering the mesh router.

b) Given that I do not have a dedicated power line to my hifi, the best I can do is turn everything off on that breaker. And I can tell the difference.

c) network is following:

• Study: Modem → EERO → Netgear 108e in low power mode → wall patch. This goes into the wall. I’m liking the 108e in low power mode better than the Cisco. can’t be fussed configuring the Cisco. maybe in the winter.
• Laundry - got a 108e in low power mode here. dodgy 5e cables in the wall I can’t do much about. this bridges the study to the room with the HiFi
• Living with the HiFi - from the wall to a EE. The EE is hooked to the ND555 and the UnitiCore. The UnitiCore being in the study doesn’t sound as good. Not sure why -since the music is downloaded anyway… Maybe some noise gets modulated in.. no idea…
• The EEPlus isolators make things better. they’re not going away. Also got the Forester F1 power supply which made a nice difference maybe since grounded.

Got a good deal on the Silent Angel Bonn NX and Forester NX PSU for it. This is clearly nuts since I’ve paid more for a switch+psu than a Sn3. Paid less than 1/2 of retail.. Maybe its worth it - can’t be worse than the valhalls2/frey2 spend…

Can’t wait to slide it in. I don’t know how it can sound much better, since the music is supposed to be downloaded rather than streamer from the uniticore and Qobuz… but we’ll se…

Oh yeah - the ferrites - atleast with everything else going on - feel like they’re choking the sounds - so they’re off for now.

ferrites now only in the laundry and the study… to choke out the internet coming into the house..