Every dB of received noise enters through one of four doors. They have different causes, different signatures and different cures, and a common-mode choke only reaches one of them. This page is about finding out which door your noise uses before you spend money, because on a quiet station the honest answer is that you should buy nothing at all. Every claim carries a C for a checkable source or an 算式 for arithmetic you can redo on paper.
Received noise enters an EFHW station through four paths. The ARRL’s own guidance for interference hunting is to ask “where is this coming from” before asking “what is producing it”, and the same discipline pays here: you cannot pick the right cure until you know the path.
| Door | Path | Coupling | How it shows up |
|---|---|---|---|
| ① Element | The antenna wire picks the noise up directly. | Near-field electric, and far field | Cut 2 barely changes it |
| ② Shield | Common-mode current on the outside of the coax braid, running through the noise field all the way from the feedpoint to the shack. | Conducted along the braid, plus field coupling | Cut 2 collapses it; a clip-on bead changes it |
| ③ Mains | Noise conducted in on the supply wiring, through the PSU, into the receiver’s ground reference. | Conducted | Cut 4 changes it |
| ④ Near field | The rig, the SDR, a USB lead or an Ethernet run is itself the receiving antenna. | Near-field | Cut 5 makes the noise follow the device |
All five fit in one afternoon and need nothing you do not already own. Cut 1 carries the most information per minute, so start there and stop as soon as a door is eliminated.
| Cut | What you do | What the reading means |
|---|---|---|
| 1 | Screw a 50 Ω dummy load onto the antenna port and listen on the same frequency. | Noise disappears → it arrives via the antenna system (door ① or ②). Noise stays → it comes in on the mains or from the near field (door ③ or ④). This one cut halves the search space. |
| 2 | Disconnect the antenna wire at the transformer but leave the coax connected and running to the shack. | Noise stays → the braid is receiving it, and a choke has a target (door ②). Noise collapses → the antenna wire itself is doing the listening, and no choke will help (door ①). |
| 3 | Clip a ferrite bead onto the coax and A/B it in and out, several times. | Level steps with the bead → you have confirmed a common-mode path, and a permanent choke is worth building. Level moves but not back, or moves only at some frequencies → you are shifting a resonance, not removing a current (see the divider section). |
| 4 | Run the radio from a battery and unplug every charger, every wall-wart and every Ethernet run in the room. | Noise changes → door ③, and the cure belongs on the mains lead, not on the coax. |
| 5 | Walk the house with a portable shortwave receiver, holding it close to suspect devices. | Noise rises sharply near one device → door ④, and you have found the culprit. Noise that is uniform everywhere points back at doors ①, ② or ③. |
The advice to add a receiving loop is repeated everywhere as if every EFHW station needs one. On this station the opposite is true, and it is worth showing how a measurement settles the question rather than a reputation. Values below are from a full year of PSK Reporter records analysed for BG1SB (grid ON80da), and they describe this site only.
| Quantity | Measured value | Reading |
|---|---|---|
| Weakest signal decoded, quietest direction | −23 dB SNR | The FT8 decoder’s limit is about −24 dB, so the station is within a dB of theoretical sensitivity |
| Noise floor by direction | Uniform, no directional source | No direction is being swallowed by a local noise source |
| DX share, receive vs transmit | 29.9% RX against 20.6% TX | Receive reaches further than transmit, so receive is not the bottleneck |
“Environmental interference” gets used as one problem. It is six. Read the cards as a comparison rather than as six essays: every card uses the same five headings, so the verdict on chokes can be read straight down the column and the differences are the point.
Mechanism. Almost all power-line noise comes from a spark or an arc across utility hardware: a gap between two conductors, caused by fittings that are loose, broken or spaced too closely. Current crosses the gap, ionises the air, and radiates broadband energy.
Signature. On a 50 Hz grid the arcing fires 100 times a second, because every cycle has two voltage peaks and two zero crossings. On a 60 Hz grid it is 120 Hz. On an oscilloscope the noise appears as two horizontal bands drifting slowly upward, one from each half of the cycle. Two free tests confirm it: rain quietens it, because water bridges the gap, and wind makes it vary, because loose hardware moves.
A correction worth having. Corona discharge is rarely, if ever, the cause. The ARRL states it plainly: corona is confined to lower frequencies, it is a low-current phenomenon, and it does not propagate far from the source. Blaming corona sends you looking for the wrong thing on the wrong pole.
CMC verdict: no. The energy travels along the power wiring and reaches you by radiation into the antenna wire — door ①. A choke acts on the coax braid’s common-mode path, and this noise is not on it.
Remedy. Complain, with evidence. This is the one category where the fix belongs to somebody else, and in many jurisdictions that somebody is legally obliged to act.
Source: ARRL, Power Line Noise · IEEE Std 1897-2024
Mechanism. A switching converter chops DC at tens to hundreds of kilohertz. The harmonics of that chop, and the ringing on every edge, run straight up through the HF spectrum. The noise leaves the box in two forms: differential mode between live and neutral, which an X capacitor suppresses, and common mode between the conductors and earth, which needs a Y capacitor and a common-mode choke. Cheap supplies cut cost by omitting precisely those last two parts — the only two that work at HF.
Signature. A comb of harmonics spaced at the switching frequency, drifting as the supply’s load changes. Often accompanied by a slow periodic wobble as the supply enters and leaves burst mode on light load.
What compliance actually buys you. Household equipment must meet the Class B conducted-emission limits. Here they are, with the voltage limit converted to the power it represents across 50 Ω:
| Frequency | Limit | Voltage | Into 50 Ω |
|---|---|---|---|
| 0.15–0.5 MHz | 66 → 56 dBμV | 2.00 → 0.63 mV | −41.0 → −51.0 dBm |
| 0.5–5 MHz | 56 dBμV | 0.63 mV | −51.0 dBm |
| 5–30 MHz | 60 dBμV | 1.00 mV | −47.0 dBm |
| for reference | quiet rural HF noise floor | typical order of magnitude | ≈ −120 dBm |
| for reference | suburban HF noise floor | typical order of magnitude | ≈ −100 dBm |
Read the gap, not the rows. A device that passes the standard with no margin is putting a signal at the mains port comparable to the first three rows. A quiet band sits at the last two. That gap is the whole problem, and it is why compliance is not the same thing as not interfering — the limit is a compromise between commerce and spectrum, not a promise that your HF reception will be usable. The bottom two rows are typical orders of magnitude rather than a standard, which is exactly why they are labelled that way.
CMC verdict: no, for the coax. Noise entering on the mains never touches the coax, so a choke there is decoration.
Remedy. The same device, on the right conductor: a common-mode choke on the mains lead or the DC lead. Also check USB and Ethernet runs, which are the most commonly missed members of this family.
Source: GB/T 9254.1-2021 (≈ CISPR 32) · EMC practice for X and Y capacitors and common-mode chokes
Mechanism. AC charging runs the mains through an on-board charger in the car — a switching converter of 3.3 to 11 kW. DC fast charging does the conversion in the post itself, at even higher power, increasingly with silicon-carbide devices and their steeper switching edges. Either way the charging cable is several metres of multi-conductor cable running from a building to a vehicle, and its protective earth conductor is exactly as long as the common-mode antenna it forms. A kilowatt-class switching converter driving a long unshielded conductor, with no RF reference ground at either end, is the mechanism written out in full.
Signature. Broadband, strongest while a car is actually charging, and gone when it unplugs. That on/off correlation is the most useful thing about it, because it dates and fingerprints the event even when you cannot see the charger.
CMC verdict: partly. The dominant path is door ① — the charging cable radiates into your antenna wire. A coax choke reaches only the fraction that arrives on the shield.
Remedy. Negotiation and geometry, in that order. Establish the on/off correlation first; that is the evidence that makes a conversation with the owner or the property manager possible. Then move your own antenna further away, or accept the hours.
Source: GB/T 18487.1-2015 · published EV-charger EMC remediation literature · 125 kHz keyless-entry case reports
Mechanism. Modern string inverters are usually transformerless, so the panels and the grid share a common-mode path with no isolation between them. Add the long DC run down from the roof, which forms a large common-mode loop, and switching devices with steep edges, and you have a strong common-mode source bolted to the same building as your antenna.
Signature. Correlated with daylight and with inverter output. Often strongest at midday, absent at night, and it changes when a cloud crosses.
CMC verdict: partly. Mixed paths. The fraction radiated into the antenna is out of reach; the fraction conducted into the building wiring is reachable, but with a choke on the mains, not on the coax.
Remedy. Almost entirely source-side, which means not yours: LCL filters, common-mode reactors, active filtering, and gate-drive optimisation all live in the inverter. What you own is the same on/off evidence, and the same conversation.
Source: ARRL RFI programme, which lists solar alongside power lines and lighting as its three tracked categories
Mechanism. Broadband over Power Line is defined as carrying data over power wiring in the 2–30 MHz band. That is not adjacent to the amateur HF bands. It is the amateur HF bands, plus everything between them.
Signature. This is the one entry in this list that is not a by-product. It is a modulated, deliberate signal inside your band, and it sounds like one.
CMC verdict: no, and a choke will never help. The signal arrives in band, on the antenna, as the wanted signal would. There is no common-mode path to interrupt.
Remedy. Frequency coordination and complaint. This is a regulatory problem, and treating it as an antenna problem wastes money.
Source: BPL band definition (2–30 MHz) · national RFI complaint procedure
Mechanism. Every LED lamp has a driver, and the driver is a small switching supply built to the tightest possible budget. The same cost pressure described in card 2 applies, at a smaller scale and with less room for filtering. High-power ballasts for grow lights are the industrial version of the same story, and the ARRL names them as a category it fields a steady stream of questions about.
Signature. Tied to a switch. The noise appears when a lamp is on and vanishes when it is off, which makes cut 5 the decisive test and makes this the easiest source on the list to identify.
CMC verdict: partly. Whether a choke helps depends entirely on which door the noise used to get in, which is why the diagnostic comes first.
Remedy. Replace the lamp with a better-behaved one. Cheaper and more reliable than filtering the consequences.
Source: ARRL RFI programme (lighting devices category)
Eight rows, and only the last one is where a coax choke does its full job. This is the whole page in one object.
| Source | Dominant door | Does a coax choke help? | Where the fix lives |
|---|---|---|---|
| Power-line arcing | ① via re-radiation | No | Complaint, and antenna geometry |
| Your own switching supply | ③ | No | A choke on the mains or DC lead |
| Your own PC, USB and Ethernet | ④ | No | Ferrites on those cables |
| A neighbour’s LED or supply | ① mostly | Partly | Source side, plus geometry |
| EV charger | ① mostly | Partly | Negotiation, plus geometry |
| Solar inverter | ① and ③ | Partly | Source side, plus the mains |
| PLC / BPL | ① in band | No | Frequency coordination, complaint |
| Near-field noise the braid picks up | ② | Yes, fully | The coaxial choke — this page |
This is the section that most choke articles skip, and it is the one that decides whether the device can work at all. Get the mechanism right and the placements, the failures and the recipes all follow from it.
A receiver responds to one thing only: the differential voltage between the inner conductor and the inside surface of the braid. Common-mode current flowing on the outside of the braid cannot reach the receiver by itself. It first has to be converted into a differential voltage, and that conversion happens at an asymmetry — the imperfect balance of the 49:1 transformer, the feedpoint, the rig’s chassis ground. The braid is a receiving antenna that is wired straight into a mode converter you cannot remove.
An ideal choke presents almost zero impedance to differential mode and a high impedance to common mode. Your wanted signal travels in differential mode and passes through untouched. The noise riding on the braid is blocked. That is the whole mechanism. It also explains what a choke does not do: it does not improve your match, raise efficiency or change SWR, because none of those have anything to do with the common-mode path. A choke that changed your SWR was changing something else, and you should find out what.
Inserting a series impedance into the common-mode loop divides the current down. To first order:
A ≈ 20 · log₁₀( 1 + Z_cm / Z_loop )
| Z_cm of the choke | Noise reduction, Z_loop = 50 Ω | Noise reduction, Z_loop = 200 Ω |
|---|---|---|
| 50 Ω | 6.0 dB | 1.9 dB |
| 100 Ω | 9.5 dB | 3.5 dB |
| 250 Ω | 15.6 dB | 7.0 dB |
| 500 Ω | 20.8 dB | 10.9 dB |
| 1000 Ω | 26.4 dB | 15.6 dB |
| 2000 Ω | 32.3 dB | 20.8 dB |
| 5000 Ω | 40.1 dB | 28.3 dB |
The divider is a first-order picture and it is wrong in three directions that matter in the field.
R + jX, and the resistive
part is the useful part. A choke that is purely reactive absorbs
nothing; it stores energy and returns it somewhere else, which is how
you get resonances rather than attenuation.
Turns give you impedance as Z_cm ∝ N², but the material’s permeability is itself a function of frequency. Measured values from the Fair-Rite material data, for the two mixes you will actually use:
| Frequency | Mix 43 μ′ | Mix 43 μ″ | Mix 51 μ′ | Mix 51 μ″ |
|---|---|---|---|---|
| 7 MHz | — | — | 364 | 156 |
| 10 MHz | 293 | 340 | — | — |
| 14 MHz | — | — | 235 | 186 |
| 28.6 MHz | — | — | 129 | 163 |
| 30 MHz | 117 | 202 | — | — |
Working recipes, all of them quoted rather than invented:
| Covers | Mix | Form |
|---|---|---|
| 80–20 m | Mix 31 | FT240-31, 5–9 turns of RG-8X |
| 40–10 m | Mix 43 | FT240-43, 6–10 turns of RG-8X |
| Portable | Snap-on Mix 31 | 3–5 spaced along the coax — convenient, and markedly worse |
Three of the four doors are out of a coax choke’s reach. Each has its own signature and its own cure, and none of the cures is a ferrite on the feedline.
The noise source sits on the same supply as your rig. Its rubbish travels down the wiring, through the power supply, into the chassis — and the chassis is the reference against which the receiver measures everything. Nothing about this route involves the coax, which is why cut 1 does not clear it.
An end-fed half wave is fed at a voltage antinode — the point of maximum electric field on the wire. A high-impedance node is inherently sensitive to near-field electric coupling, which is exactly how lamp drivers and small switching supplies mostly couple. The topology that makes the EFHW convenient to hang is also the topology that makes it a good listener for everything in the building.
Your rig, your SDR, a USB lead, an Ethernet run: any of these is a small antenna sitting on your desk, listening to whatever is switching next to it. It is the least exotic door and the easiest to overlook, because the culprit is often something you installed yourself and therefore trust.
The two rules from section 01 apply here in full, and they are the whole method. Everything below is about making them easy to obey.
You do not need a laboratory. If your station runs through the MRRC stack, you already have an S-meter feed and a DSP spectrum display — both are sufficient to A/B a choke, provided you keep the receiver settings frozen. A receiver’s spectrum display is the better instrument of the two, because it shows you the shape of the noise as well as its level, and a choke that moves a resonance looks obviously different from one that removes a current.
For an objective, repeatable number that survives across days, the same PSK Reporter analysis this station uses is available to any station with a callsign and some patience: it produces a noise-floor figure from which stations you can hear, and the method is documented on the measured antenna profile page. Run it before the change, run it after, and compare the two.
Section 02 built a verdict table. This is the same table read as a list of things not to waste a ferrite on.
The standard answer for a station buried in local noise is to split transmit from receive: keep the EFHW for transmitting, add an active magnetic loop for receiving, aim its null at the noise source, and switch between them with a T/R relay or the rig’s own receive antenna port. It works, it is well established, and the products exist.
Power-line arcing and a neighbour’s charger are the two categories where you cannot treat the source yourself. Both need evidence and a channel. In China the channel is the 12381 hotline, opened on 20 March 2023 as the radio-interference complaint centre under the MIIT public-service platform.
Build the log before you call: when it started, which bands, which modes, what it sounds like, and a 30 to 60 second recording in both AM and SSB. This is the same list the ARRL recommends to its own members, and it is what turns “my radio is noisy” into a case.
Every quantitative statement above carries one of two labels, or it carries a dash.
| Label | Means |
|---|---|
| C | An engineering claim with a checkable source. The sources are listed below. |
| 算式 | Arithmetic you can redo. The formula is always printed next to the numbers, so you can reproduce every cell yourself. |
| — | Not measured, and not guessed at either. |