Shielded Receive Loops: Floating Shields, Grounding and Common Mode
Shielded Receive Loops: Floating Shields, Grounding and Common Mode
My short split-screen loops favored floating shields. My large loop on ground benefited from an RC reference. Electrical size, reference impedance and balance explain why these choices can coexist.
RF.Guru working definition: Common-mode current is the non-cancelling phasor-sum current in a specified set of conductors, evaluated at a defined cross-section and using a declared current-direction convention. In the intended differential transmission-line mode, the outgoing and return currents are equal and opposite, so their phasor sum is zero. When they do not cancel, the remaining current must close through another reference or return path—such as the outside of a coax shield, a mast, equipment chassis, station wiring, nearby structures, earth, the operator, or distributed coupling through the environment.
This broader working definition is especially useful in practical antenna systems. On transmit, non-cancelling current on the outside of the coax can make the feedline and connected structures part of the radiating antenna system unless that path is intentional, clearly defined and properly controlled—for example by providing the required return path and placing a suitable common-mode choke at the correct boundary.
Why I tested the ground connection
I regularly receive questions about shielded receiving loops that start with the same assumption: the shield must be grounded, and connecting it to the center tap of a push-pull input must improve rejection. My own comparisons give me a reason to examine that connection much more carefully.
For short receiving elements with a circumference of approximately 1.2–3 m, I use a shield interrupted at both the top and bottom. In my far-field checks and near-field probe tests, leaving those shield sections floating made the loop less prone to unwanted pickup than connecting the shield to the center tap of the push-pull input.
With a much larger shielded loop on ground, I use a different arrangement. The shield has a connection to the common reference, but inserting an RC network in that connection reduced unwanted common-mode coupling compared with the direct connection in my tests. Once the circumference reaches 40 m or more, the shield is electrically long over much of HF.
These are observations from my arrangements, without a universal rejection figure attached. They support a useful engineering conclusion: the reference connection is part of the receiving antenna and must be designed and tested as such. A ground symbol by itself does not explain what the shield will do.
The shield, the receiving conductor and the feed cable are different paths
Keep three physical structures separate: the intended receiving conductor and differential input, the conducting screen around that element, and the output cable with its power and station connections. A change to the screen connection can change currents on all three through conductive, capacitive and inductive coupling.
A conductive screen redistributes charge and changes the electric field around and inside its structure. It does not require an earth wire for charge redistribution to occur. However, an open, split antenna screen is not a closed enclosure with a guaranteed shielding effectiveness. Its performance depends on geometry, frequency, connections and the impedances of the circuit it surrounds.
The screen also supports induced surface currents. A continuous low-impedance conducting turn can oppose the magnetic flux that the loop is intended to receive. An interruption prevents that particular galvanic shorted turn. The magnetic field does not simply enter through the gap as though the gap were a window; the complete electromagnetic boundary and the receiving circuit determine the response.
For an electrically small loop in an approximately uniform field, the familiar open-circuit magnetic response is:
Here ω = 2πf, Hn is the magnetic-field component normal to the loop, and Aeff includes the effective area and turns under the assumed excitation. The sign depends on the chosen loop orientation; the phasor convention is ejωt. The voltage delivered to the amplifier also depends on the loop impedance and input loading. This small-loop expression is not a complete model for an electrically long loop near soil.
Two gaps create two shield sections
With a gap at the top and another at the bottom, there are two separate conducting screen sections, provided that neither the feed box nor the mounting hardware bridges them. Each section can have its own RF potential and current distribution. They should not be drawn as one continuous grounded ring.
“Floating” means that there is no deliberate galvanic reference connection. It does not mean that there is no RF return path. Capacitance remains between each section and the receiving conductor, the other section, the amplifier, the feed cable, the ground and nearby objects. Gap capacitance also remains. Moisture, fasteners and connector shells can change these paths or accidentally bridge an intended interruption.
As a scale illustration, the magnitude of the reactance of an ideal 10 pF capacitance is about 15.9 kΩ at 1 MHz, 1.59 kΩ at 10 MHz and 531 Ω at 30 MHz. These are calculated examples, not measurements of my loop. They show why a section that is isolated on a multimeter can still have a significant RF connection.
The gaps described here are in the screen. They must not be confused with arbitrarily cutting the receiving conductor or changing the intended differential input circuit. Those changes produce a different antenna.
A center tap is a symmetry point with a real impedance
Under ideal differential excitation, equal and opposite signals can make the midpoint of a symmetrical winding a voltage null. That result belongs to that mode and its symmetry assumptions. It does not establish zero common-mode impedance, nor does it establish that the node is free of noise.
At RF, the center tap sees winding inductance and capacitance, lead and PCB impedance, bias components, supply decoupling and whatever cable or chassis connections complete the return path. A connection that looks short at DC may develop a substantial voltage when RF current flows through it.
A direct shield connection can therefore do two things at once: conduct screen-induced current into the input reference and transfer reference-node voltage onto the screen. A passive bond does not distinguish between wanted grounding and unwanted noise injection. The resulting screen voltage and current can couple back into the receiving conductor.
Leaving the two screen sections floating removes that deliberate conductive route. In a short, sufficiently symmetrical structure, capacitive coupling may then produce less unwanted differential voltage than the hard connection did. This is a plausible mechanism for my observed improvement. The comparison alone does not identify the contribution of every parasitic path.
The reverse result is also physically possible. A floating section can develop a large RF voltage, resonate with its surroundings or couple unequally to the two input sides. A well-designed quiet reference connection can then help. My preference for the floating arrangement follows the behavior of this short split-screen design; it is not a rule that every receiving screen should float.
Common-mode rejection can be lost before the amplifier
For two input voltages measured relative to the same declared reference, define:
Vcm = (Va + Vb)/2
Vout = AdVd + AcmVcm
The amplifier's voltage common-mode rejection ratio is 20 log10|Ad/Acm| under the specified frequency, loading and source conditions. It does not describe all field pickup by the antenna, and it does not remove interference that has already become a differential input voltage.
Consider a simplified weak-coupling model. A screen node at voltage Vs couples through capacitances Csa and Csb into input sides with effective impedances Za and Zb to the reference. To first order:
This model assumes small injected currents, linear operation and negligible change in Vs due to loading. It illustrates a specific conversion mechanism; a real transformer-coupled input generally needs a coupled impedance model. Equal capacitances are insufficient when the impedances differ, and equal impedances are insufficient when the capacitances differ. Two floating sections require separate voltages and coupling terms.
Input protection capacitance, winding asymmetry, unequal leads, an offset feed cable or a nearby metal support can all affect this balance. Once those paths generate Vd,err, the push-pull stage amplifies it along with the wanted differential signal.
The same general conversion principle is described in Analog Devices' MT-070 input-filter tutorial: mismatched common-mode input networks create differential interference. Its instrumentation-amplifier circuits and component values are not prescriptions for an HF receiving loop.
“Short” must be stated in wavelengths
For a circumference P, the free-space electrical-size indicator is P/λ0 = Pf/c. The table uses c = 299,792,458 m/s. It is a scale check, not a prediction of shield resonances.
| Frequency | 1.2 m circumference | 3 m circumference | 40 m circumference |
|---|---|---|---|
| 3.5 MHz | 0.014 λ0 | 0.035 λ0 | 0.467 λ0 |
| 7 MHz | 0.028 λ0 | 0.070 λ0 | 0.934 λ0 |
| 14 MHz | 0.056 λ0 | 0.140 λ0 | 1.868 λ0 |
| 21 MHz | 0.084 λ0 | 0.210 λ0 | 2.802 λ0 |
| 30 MHz | 0.120 λ0 | 0.300 λ0 | 4.003 λ0 |
A 3 m circumference is comfortably short relative to a wavelength on some lower bands, but reaches 0.3 λ0 at 30 MHz. Uniform-current and single-node approximations become progressively less reliable as size grows; there is no abrupt physical switch at a particular decimal fraction of a wavelength.
A 40 m circumference equals one free-space wavelength near 7.49 MHz and spans several wavelengths on higher HF bands. A two-gap arrangement divides the screen into shorter arcs, but each arc can still be electrically long. A single gap opposite the feed interrupts the continuous turn while leaving substantial conducting paths on either side.
Do not use the coax manufacturer's internal velocity factor to predict a resonance of the outside of the shield. That figure describes the guided differential mode between the coax conductors. The exterior mode interacts with air, insulation, soil and nearby conductors and can have a different propagation behavior.
A large ground-loop shield can carry standing currents
On an electrically large loop, screen voltage and current vary with position. Incident fields provide distributed excitation; the gap, feed junction, reference network and ground loading impose boundary conditions. The screen can support nonuniform currents and resonances even though it does not form a DC-closed ring.
Consequently, a direct bond plus a gap on the opposite side can still produce significant unwanted coupling. The bond changes the boundary at one location. It does not force tens of meters of screen to remain at the same RF potential, and it does not eliminate capacitive return current through the surroundings.
For a loop on ground, soil conductivity and permittivity, height above the surface, moisture and nearby wires all enter the problem. The reference lead, feed cable and any earth electrode add further impedances. “Common ground” must identify an actual node and connection; amplifier ground, coax exterior, chassis and a stake in the soil cannot simply be assumed equipotential.
That is why I treat the large-loop reference connection as a circuit element within a distributed antenna. My earlier TerraBooster placement article discusses the installation side of this problem. A favorable bond arrangement does not remove the need to consider nearby conductors and the output cable.
What an RC reference network can change
The improvement I found with the RC connection tells me that changing the reference impedance changed the unwanted coupling. Possible contributions include a change in screen voltage, redistribution of current between the screen and cable, and resistive damping of a mode that has appreciable voltage across the network. The measurement does not, by itself, identify one of these as the sole cause.
Write the installed network as Zref(f), including lead and layout parasitics. For a single simplified mode driven by an equivalent voltage Vdrive through an external impedance Zext:
This is an explanatory equivalent circuit, not a complete model of a 40 m antenna. Increasing |Zref| does not guarantee less current at every frequency: its reactance can cancel part of Zext, and the driving voltage and coupling can change with the boundary condition. A resistive component can dissipate energy in an excited mode, but only where current or voltage actually couples that mode into the resistor.
The letters “RC” do not specify a filter response. Two common two-terminal possibilities illustrate the difference:
| Ideal network in the reference connection | Impedance | What it actually implies |
|---|---|---|
| R and C in parallel | Zref = R / (1 + jωRC) | Finite DC resistance; the capacitor increasingly bypasses the resistor as frequency rises. |
| R and C in series | Zref = R + 1/(jωC) | No ideal DC path; at high frequency the impedance approaches R, before parasitics dominate. |
For the parallel arrangement, f = 1/(2πRC) is the frequency where the ideal resistor and capacitor admittances have equal magnitudes. It is not automatically the cutoff frequency of noise reaching the receiver. At higher frequencies the capacitive branch makes the connection stronger, and a resistor bypassed by that capacitor may have little influence on a particular resonance. A series-resistor/shunt-capacitor low-pass filter is a different circuit again: its two ports, reference node and source/load impedances must be specified.
Component parasitics, resistor thermal noise coupled into the input, wanted-signal loading and balance belong in the assessment. The aim is a useful received signal-to-noise ratio with controlled cable participation. A lower noise indication caused by reduced overall sensitivity would not establish that improvement. This discussion concerns a functional RF reference; it does not prescribe inserting components into a protective-earth conductor.
Separate field pickup, cable current and amplifier rejection
Electric-field pickup, common-mode current and common-mode voltage at an input are related through the installation, but they are not interchangeable measurements. A probe result should identify the excited field or port, the receiving quantity and the reference used.
The EMC Europe study by Wang and colleagues is useful context. Their balanced shielded-loop arrangement showed that feed and gap geometry affected the local electric-to-magnetic field ratio and the shielding measurement. A balanced feed therefore did not make the physical loop an ideal magnetic-field sampler. Their experiment did not compare my two-gap floating screen against my center-tap connection, or test a 40 m loop on soil. See the accepted paper and my earlier analysis for OctaLoop and TerraBooster.
A near-field source can create a strongly electric or strongly magnetic local field, but real probes have cross-sensitivity and their leads can radiate or pick up signal. Moving the probe changes the excitation geometry; moving its cable may change it too. A source described as a magnetic probe is not, by that description alone, a calibrated pure-H-field source.
In the far field of a source in free space, E and H have a fixed wave relationship. A single far-field comparison cannot independently separate electric and magnetic coupling. Its value is testing the complete receiving response under that illumination. Source dimensions, wavelength, distance and reflections determine whether a proposed test is actually in the far-field regime.
A comparison that can distinguish the mechanisms
To extend the observations into a repeatable characterization, I would use the following sequence. It builds on the controlled-comparison approach in my long-term receive-antenna A/B article.
- Document the conductors and nodes. Record circumference, receiving-conductor wiring, screen gaps, which sections are joined, center-tap wiring, RC topology, cable route and the exact reference node. Check for unintended conductive bridges.
- Change only the reference arrangement. Compare floating sections, the direct bond and the specified impedance network on the same element and amplifier. Keep geometry, supply, input loading and cable placement fixed. A switch or jumper fixture adds capacitance of its own and must be accounted for.
- Measure the wanted response. Use a repeatable intended-signal excitation across frequency and maintain receiver gain and bandwidth. Disable changing AGC behavior for amplitude comparisons. Normalize the interference response to wanted response so that a sensitivity reduction cannot masquerade as better rejection.
- Excite interference paths separately. Compare localized electric-dominant and magnetic-dominant probe arrangements, a declared common-mode injection port, and suitable distant illumination. Record probe orientation, distance, drive and feed routing. These tests answer different questions.
- Observe the cable. A calibrated current probe around the complete coax measures the non-cancelling current through its aperture; ideally the internal differential currents cancel. It does not directly map current over both loop-screen sections. Check position dependence and probe loading. See my explanation of whole-coax current measurement.
- Sweep and repeat. Look for narrow peaks, frequency-dependent reversals, overload and changes when realistic cable routes or soil conditions change. Repeat A/B ordering; for two simultaneous receivers, swap channels to expose channel bias. Retain complex amplitude and phase where the instruments allow it.
For a controlled wanted excitation of fixed level, let S be its output amplitude and N the output amplitude caused by a fixed interference excitation, measured separately. A useful comparative rejection change is:
A positive value indicates improved discrimination in that test. Use 10 log10 for power ratios; random-noise comparisons also require equal bandwidth and suitable averaging. This is a proposed reporting method, not a numerical result from my existing tests. It is not the amplifier's intrinsic CMRR.
How I choose the shield connection
| Arrangement | Design reason | What must remain under control |
|---|---|---|
| Short screen, gaps at top and bottom, floating sections | Removes the deliberate screen-to-reference current path; favored by my reported comparison. | Section symmetry, capacitive returns, wanted sensitivity and increasing electrical size at higher frequencies. |
| Screen tied to the push-pull center tap | Provides a defined conductive reference where that reference and geometry are suitable. | Center-tap RF impedance and noise, winding balance, screen-current return path and conversion into the differential input. |
| Large ground-loop screen with an RC reference | Changes the termination seen by distributed screen modes; reduced unwanted coupling in my tests. | Actual network topology, soil and cable paths, modal damping, parasitics and retained signal-to-noise performance. |
For my short split-screen receiving elements, the floating arrangement is a deliberate design choice supported by the comparisons I made. For the large loop on ground, the reference impedance is part of the antenna design. The different choices follow from different structures and current paths.
A shield connection is useful when it improves the intended receiving behavior. A center tap, a ground symbol and an opposite gap each describe a feature of the circuit. None of them, alone, establishes rejection of unwanted pickup.
Mini-FAQ
- Does a receiving-loop shield always need to be grounded? No. Its RF behavior depends on electrical size, segmentation, parasitic coupling and the actual reference connection. Joeri found less unwanted pickup with floating sections on his short, two-gap arrangement; that does not establish a universal rule.
- Why can connecting the shield to a center tap make pickup worse? A center tap can be a differential symmetry point while still having finite common-mode impedance and noise. Bonding the shield can carry current into that reference or drive the shield from it. Unequal coupling can then create differential interference before the amplifier.
- Do gaps at the top and bottom remove every RF current path? No. They divide the screen into separate conducting sections when no other conductor bridges them. Capacitance between sections, the receiving conductor, cables and the environment still permits displacement current.
- Is a 3 m circumference electrically small throughout HF? No. It is about 0.070 free-space wavelength at 7 MHz but 0.300 wavelength at 30 MHz. Small-loop and single-node approximations become less reliable as electrical size increases.
- Why does a large ground-loop shield need different treatment? At 40 m circumference it is already close to a free-space wavelength at 7 MHz and spans several wavelengths on higher HF bands. Distributed excitation, soil loading, gaps and reference connections can create nonuniform currents and resonances.
- Is an RC shield connection automatically a low-pass noise filter? No. Its topology and surrounding impedances determine the response. A parallel RC connection approaches capacitive behavior at higher frequencies; a series RC connection approaches its resistance before parasitics dominate. Neither label alone specifies noise attenuation at the receiver.
- How can I tell whether a quieter arrangement actually improved reception? Compare the interference response after normalizing to a repeatable wanted-signal response, with fixed bandwidth, gain and geometry. Check cable common-mode current and frequency-dependent peaks separately. Lower indicated noise alone can result from reduced sensitivity.