Sunday, January 14, 2024

Reducing RFI (Radio Frequency Interference) for a POE (Power Over Ethernet) camera or wireless access point

One of the (many) banes of the amateur radio operator's existence is often found at the end of an Ethernet cable - specifically a device that is being powered via "Ethernet":  It is often the case that interference - from HF through UHF - emanates from such devices.

Figure 1:
POE camera with both snap-on ferrites installed -
including one as close to the camera as possible -
and other snap-on/toroids to suppress HF through VHF.
Click on the image for a larger version.

Why this happens

Ethernet by itself is usually relatively quiet from an (HF) RF standpoint:  The base frequency of modern 100 Megabit and gigabit Ethernet is typically above much of HF and owing to the fact that the data lines are coupled via transformers making them inherently balanced and less prone to radiate.  Were this not the case, the integrity of the data itself would be strongly affected by the adjacent wires within the cable or even if the cable was routed near metallic objects as it would radiate a strong electromagnetic field - and any such coupling would surely affect the signal by causing reflections, attenuation, etc.

This is NOT the case with power that is run via the same (Ethernet) cable.  Typically, this power is sourced by a switching power supply - too often one that is not filtered well - and worse, the device at the far end of the cable (e.g. a camera or WiFi access point - to name two examples) is often built "down to a cost" and itself contains a switching voltage converter with rather poor filtering that is prone to radiation of RF energy over a wide spectrum.  Typically lacking effective common-mode filtering - particularly at HF frequencies (it would add expense and increase bulk) - the effect of RF radiating from the power-conducting wires in an Ethernet cable can be severe.

Even worse than this, Ethernet cables are typically long - often running in walls or ceilings - effectively making them long, wire antennas, capable of radiating (and intercepting) signals even at HF.  The "noisy" power supply at one or both ends of this cable can act as transmitters.

What to do

While some POE configurations convey the DC power on the "spare" conductors in an eight conductor cable (e.g. the blue and brown pairs), some versions use the data pairs themselves (often using center-tapped transformers in the Ethernet PHY) meaning that it may not be easy to filter just the DC power.

While it is theoretically possible to extract the power from the Ethernet cable, filter it and and reinsert it on the cable, the various (different) methods of doing this complicate the matters and doing so - particularly if the DC is carried on the data pairs - can degrade the data integrity by requiring the data to transit two transforms incurring potential signal attenuation, additional reflection and affecting frequency response - to name just a few.  Doing this is complicated by the fact that the method of power conveyance varies as you may not know which method is used by your device(s).

It is possible to subject the entire cable and its conductors to a common-mode inductance to help quash RFI - but this must be done carefully to maintain signal integrity.


Some POE cameras also have a coaxial power jack that permits it to be powered locally rather than needing to use POE.  I've observed that it is often the case that using this local power - which is often 12-24 volts DC (depending on the device) - will greatly reduce the noise/interference generated by the camera and conducted on the cable - provided, of course, that the power supply itself is not a noise source.  Even if a power supply is used near the camera, I would still suggest putting its DC power cable through ferrite devices as described below to further-reduce possible emissions.
There are some devices (such as those sold by DX Engineering) that are essentially back-to-back signal transformers that can reduce radiation of signals from Ethernet cable, but these typically do not permit the passage of power and are not candidates for use with POE devices.

Ferrite can be your friend

For VHF and UHF, simple snap-on ferrites can significantly attenuate the conduction of RF along, but these devices are unlikely to be effective at HF - particularly on the lower bands - as they simply cannot add enough impedance at lower frequencies.

To effectively reduce the conduction of RF energy at HF, one could wrap the Ethernet cable around a ferrite toroidal core, but this is often fraught with peril, particularly with cable carrying Gigabit Ethernet - as tight radius turns can distort the geometry of typical CAT-5/6 cable, affect the impedance and cause cross-coupling into other wire pairs.  If this happens, one often finds that the Ethernet cable doesn't work reliably at Gigabit speeds anymore (being stuck at 100 or even 10 Megabits/second) or starts to "flap" - switching between different speeds and/or slowing down due to retransmissions on the LAN.

One type of Ethernet cable that is quite resistant to geometric distortion caused by wrapping around a toroidal core is the flat Ethernet cable (sometimes erroneously referred to as "CAT6" or "CAT7").  This cable is available as short jumpers around 6 feet (2 meters) long and, with the aid of a female-female 8P8C (often called "RJ-45") coupler can be inserted into an existing Ethernet cable run - just be sure that it is from a reputable source and rated for "Gig-E" service.  As it is quite forgiving to being wrapped around ferrites, this flat cable can be pre-wound with such devices and inserted at the Ethernet switch end and/or the device end at a later time.  I have found that with reasonable quality cable and couplers that this does not seem to degrade the integrity of the data on the LAN cable - at least for moderate lengths (e.g. 50 feet/15 meters or less) - your mileage may vary with very long cable runs.

As the flat cable and female-female Ethernet coupler are to be inserted into the cable run, they must be of known, good quality so it is best to test the couplers and cable that you obtain prior to installation to be sure that their use doesn't cause a reduction in signal quality/speed.

Practical examples

Best attenuation across HF

Figure 2:
Three toroids wound on "flat" Ethernet cable.  An FT114-43
is used on each end with an FT114-31 in the middle.
Click on the image for a larger version.
Using a test fixture with a VNA, I determined that for best overall attenuation across the entire HF spectrum I needed three ferrite toroids on the 2 meter long flat Ethernet jumper.  All three of these were FT-114 size (1.14", 29mm O.D.) with the first and last being of material type 43 and the center being type 31:  Both types 31 and 43 offer good impedance to low HF but 43 is more effective on the higher bands - namely 10 and 6 meters - and types 52 and 61 may be useful at even higher bands:  The three toroids, separated by a few inches/cm, offer better all-around rejection from 160 meters through 10/6 meters than just one.  Having said this, it is unrealistic to expect more than 20dB or so of attenuation to be afforded by ferrite devices at high HF/low VHF - "because physics".

One might be tempted to use the more-available FT-240 size of toroids (2.4", 60mm O.D.) but this is unnecessarily large, comparatively fragile and expensive:  While you can fit more turns on the larger toroid, one hits the "point of diminishing returns" (e.g. little improvement with additional turns) very quickly owing to the nature of the ferrite and coupling between turns.  Using the FT114 or FT140 sizes is the best balance as it may be much less expensive than a larger device, it can accept 6-8 turns with the cable's connector installed, and more than 6-8 turns is rapidly approaching the point of diminishing returns for a single ferrite device, anyway.

In bench testing with a fixture, it was found that three toroids on a piece of flat Ethernet cable provided the best, overall attenuation across HF and to 6 meters - significantly better than any combination of FT114, FT140 or FT240 toroids of either 43 or 31 mix alone:  Figure 2, above, shows what this looks like.  Two FT114-43 and one FT114-31 toroid were used - the #31 toroid being placed in the center, providing the majority of series impedance at low HF and a #43 at each end being more effective at higher HF through 6 meters.

To construct this, the flat Ethernet cable was then marked with a silver marker in the center and four turns were wound from each end, in turn, for a total of eight turns on the FT114-31.  Placing an FT114-43 at 12 inches (25cm) and winding seven turns puts the FT114-43 fairly close to each connector, allowing the installation of one or two snap-on ferrites very close  to the connector if it is determined that more suppression is required to suppress radiation at VHF frequencies.  Small zip-ties (not shown in Figure 2) are used to help keep the turns from bunching up too much and also to prevent the start and stop turns from getting too close to each other:  Do not cinch these ties up enough to distort the geometry of the Ethernet cable as that could impact speed - particularly when using Gig Ethernet.

It is important that, as much as possible, one NOT place a "noisy" cable in a bundle with other cables or to loop it back onto itself - both of which could cause inadvertent coupling of the RFI that you are trying to suppress into the other conductors - or to the far side of the cable you are installing.

Best attenuation at VHF and HF

If you are experiencing interference from HF through VHF, you will need to take a hybrid approach:  The use of appropriate snap-on and toroidal ferrite devices.  While snap-on ferrite devices are not particularly useful for HF - especially below about 20 MHz - they can be quite effective at VHF, which is to be expected as that is the purpose for which they are typically designed.  Similarly, a ferrite toroid such as that described above - particularly with type 43 or 31 material - will likely have little effect on VHF radiation - particularly in the near field.

Figure 3:
A combination of a snap-on device with an extra turn looped
through it and two ferrites to offer wide-band suppression
from HF through VHF.
Click on the image for a larger version.

Figure 3 shows such a hybrid approach with a snap-on device on the left and two toroids on the right to better-suppress a wider range of frequencies.  In this case it is important that the snap-on device be placed as close to the interference source as possible (typically the camera) as even short lead lengths can function as effective antennas at VHF/UHF.  You may also notice that the snap-on has two turns through its center as this greatly improves efficacy at medium/low VHF frequencies but may be counter-productive at high VHF/UHF frequencies owing to coupling between turns.

Doing this by itself is not likely to be as effective in reducing radiation at VHF/UHF from the cable itself, often requiring the placement of additional ferrite devices.  Figure 1 shows the installation of several snap-on devices placed as close to the POE camera as physically possible - mainly to reduce radiation at VHF and UHF as at those frequencies where even a few inches or centimeters of cable emerging from the noise-generating device can act as an effective antenna.

Determining efficacy

During the installation of these devices on my POE cameras I was interested in how much attenuation would be afforded at VHF:  Since I'd already used the "chokes on a flat cable" approach like that in Figures 2 and 3 I knew that this would likely be as effective as was practical at HF - but because the VHF/UHF noise could be radiated by comparatively short lengths of "noisy" cable - and that the 43 and 31 mix ferrites were probably not as effective at those frequencies - I needed to be able to quantify that what I did made a difference - or not.

Figure 4:
The cable in Figure 3 installed, but not yet
tucked into place as depicted in Figure 1.
(This does not show the snap-on ferrites installed
where the wire exits the camera housing.
The female-female RJ45/8P8C "splice" can be
seen in the upper-left corner of the picture.
Click on the image for a larger version.

For HF this was quite simple:  I simply tuned my HF receiver - connected to my main antenna - to a frequency where I knew that I could hear the noise from the cameras and compared S-meter readings with the system powered up and powered down.  This approach is best done at a time during which the frequency in question is "dead" or at least weak (e.g. poor propagation) - 80/40 meters during the midday and 15/10 meters at night is typical.

For VHF this required a bit more specialized equipment.  My "Go-To" device for finding VHF signals - including noise - is my VK3YNG DF sniffer which has extremely good sensitivity - but it also has an audible "S-meter" in terms of a tone that rises with increasing signal level:  This allowed an "eyes and hands off" approach in determining efficacy of the installation of a ferrite device simply by hearing the pitch of the tone..  Switching it to this mode and placing it and its antenna at a constant distance fairly close to the device being investigated allowed me to "hear" - in the form of a lower-pitched tone - whether or not the application of a ferrite device made a difference.

Slightly less exotic would be an all-mode receiver capable of tuning 2 meters such as the Yaesu FT-817, Icom IC-706, 703 or 705.  In this case the AM mode would be selected and the RF gain control advanced such that the noise amplitude audibly decreased:  This step is important as not doing this could mean that if the noise decreased, the AGC in the receiver would simply compensate and hiding the fact that the signal level changed.  By listening for a decrease in the noise level one can "hear" when installing a snap-on ferrite made a difference - or not.

One cannot use a receiver in FM mode for this as an FM detector is designed to produce the same amount of audio (including noise) at any signal level:  A strong noise source and a weak one will sound exactly the same.  It's also worth noting that the S-meter on a receiver in FM mode - or an FM-only receiver - are typically terrible in the sense that their indications typically start with a very low signal and "peg" the meter at a signal that isn't very strong at all which means that if you try to use one, you'll have to situate the receiver/antenna such that you get a reading that is neither full-scale or at the bottom of the scale to leave room for the indication of change.

Of course, a device like a "Tiny SA" (Spectrum Analyzer) could be used to provide a visual indication, using the "Display Line", markers and stored traces to allow a quick "before and after" determination.  As mentioned above, one would want to place the antenna and the receiving device (an actual receiver or spectrum analyzer) at a fairly close distance to the device being investigated - but keep it and its antenna in precisely the same location (or connected to a fixed-location antenna) during the entire time so that one can get meaningful "before and after" readings.


With the use of ferrites alone, one should not expect to be able to completely suppress radiation of RF noise from an Ethernet cable - the typical maximum to be reasonably expected is on the order of about 20dB (a bit over 3 "S" units) and this can vary wildly with frequency.  In a situation where the POE device is very close to the antenna, it may not be possible to knock the interference down to the point of inaudibility in which case relocation to place the two farther apart - or trying similar devices of different models/brands to try to find a combination to reduce it..

The most effective use will be for noise sources will be at some distance from the receive antenna - particularly if a long cable is used that may act as an antenna.  Additionally, these measures can be effective in situations where your transmitter causes problems with the device itself due to ingress of RF energy along the Ethernet cable.

Be prepared to install appropriate ferrite devices at both ends of the cable as it's often the case that not only does the POE device itself (camera, wireless device) radiates noise but also the POE switch itself:  No-name brand POE power supplies and Ethernet switches are, themselves often very noisy and the proper course of action would be to first swap out the supply or POE switch with a known quiet device before attaching ferrite.

As every interference situation is unique, your mileage may vary, and the best road to success is being able to quantify that changes you have made made things better or worse.

This page stolen from