Shielded Loops: What the York Study Means for Our Designs

A paper about measuring the shielding of a carbon-fiber sheet deserves a close reading if you design shielded receiving loops. Yi Wang, Andrew C. Marvin, Simon J. Bale, Ali Ghaffarlouy Raef, John Dawson and Martin Robinson have examined something that matters directly to my work on OctaLoop, OctaLoop Mini and TerraBooster: the feed and shield interruption are part of the electromagnetic structure, even when the feed is balanced.

The source is their EMC Europe 2026 paper, Investigation of the feasibility of near-field magnetic shielding measurements in the 10 MHz to 200 MHz frequency range. I am discussing the complete accepted manuscript available from White Rose Research Online. The repository records acceptance on 27 April and publication on 4 September 2026. Its experiment concerns material shielding; it does not measure any RF.Guru antenna.

My conclusion is useful and specific. Shielding, a balanced input and feedline-current control remain sensible design choices, but they solve different coupling problems. A shielded loop can still have a direction-dependent electric response. The right design question is how the actual gap, feed, load and surroundings affect the complete receiving system. This extends the approach I described in how our receive antennas evolved: follow the unwanted path all the way to the receiver.

What the York experiment actually compares

The authors first calculate the field of an ideal electrically small circular loop, then simulate a physical unshielded loop with a finite-impedance feed, and finally simulate and build a shielded loop. The last step matters: the laboratory loop is not a perfectly symmetric magnetic dipole merely because its outline is circular.

For the field plots, the two observation directions are in the loop plane. The 0° ray crosses the gap; the 90° ray is orthogonal to it. In the sheet experiment, 0° puts the gaps nearest the sheet. The 90° case rotates the gap/feed axes within the loop planes. The loop normals do not tilt. This is different from the ordinary polarization experiment in which we turn a loop normal toward or away from the incident magnetic field. Figures 6, 8, 9 and 14 of the accepted manuscript establish that distinction.

“Coplanar” means the two loop planes coincide, with the loops side by side across the barrier. “Coaxial” means their normals lie on the same axis, like two facing rings. Here, coaxial describes geometry, not a coaxial feed cable. The paper uses the coplanar arrangement; changing to coaxial geometry changes which field components meet the sheet.

The quoted 150 mm field-observation distance is explicitly measured from the source-loop center. The sheet-test text says loop-to-veil distance but supplies no dimensioned center-versus-edge reference drawing. A replication should specify both distances instead of silently equating them.

Transverse impedance is a local field ratio

Let the circular loop lie in the equatorial plane of spherical coordinates. The observation distance is r, the angle from the loop normal is θ, and azimuth around that normal is φ. In the ideal model, θ = 90° gives an electric field E_φ and a magnetic field H_θ; the radial magnetic component H_r vanishes there. Equation (3) defines the transverse ratio as E_φ/H_θ, in ohms.

Transverse means perpendicular to the local radial direction. Longitudinal means along it. Away from the equatorial plane, a small loop also has a radial magnetic near-field component. It is an ordinary part of Maxwell's solution, not evidence for a separate long-range longitudinal radio wave. On the ideal loop axis the transverse E/H ratio can become undefined because both transverse components vanish while radial H remains. A field ratio always needs its position, components and reference directions.

It is also different from the loop's terminal impedance, the coax's characteristic impedance and the amplifier input impedance. All have units of ohms, but they relate different quantities. Likewise, “longitudinal” relative to the source-to-observer radius does not mean common-mode current flowing along a feed cable.

To interpret the ideal-loop calculation, set x = kr = 2πfr/c and η₀ = √(μ₀/ε₀) ≈ 377 Ω. With the transverse field directions chosen for outward power flow, the magnetic-dipole expression reduces to:

These are independent algebraic reductions of the paper's ideal-dipole equations, not fits to its measurements. The common current, area and propagation factors cancel. At a fixed location deep in the reactive near field, the E/H magnitude is small relative to 377 Ω: that is the useful meaning of H-dominant here. The ratio also has phase. A mostly reactive E/H ratio is not the real power-flow relation of a plane wave.

There is a sign convention to catch when reproducing the paper. Taking its printed equations (1)–(3) literally gives E_φ/H_θ = −Z_out, because φ-hat crossed with θ-hat points inward. Its Figure 3 instead shows the outward-reference phase, tending from +90° toward 0°. The magnitudes agree; a complex implementation needs the explicit minus sign or an equivalent reversal of one field reference.

The independence from loop area is a property of the ideal dipole approximation. It does not license rescaling the physical 100 mm loop to a meter-scale receive antenna at the same frequency. Electrical size, finite-distance field variation, feed geometry, gap capacitance and loading return as soon as the real conductors are modeled.

A balanced feed does not erase the gap

The unshielded simulations already differ between observation azimuths once a 50-ohm drive point is introduced. For the shielded loop, the contrast becomes especially strong toward the upper part of the sweep: the gap-crossing direction has a higher E/H magnitude than the orthogonal direction. This occurs with the authors' balanced 100-ohm excitation already in place. It cannot be dismissed as merely the mistake of feeding the loop single-ended.

The physical explanation is that the shield controls charge and current paths rather than making electric fields disappear. A deliberate interruption prevents a continuous low-impedance shield turn, but leaves a region with fringe electric fields and displacement current. The feed and its load provide another distinguished part of the structure. A globally balanced port does not restore rotational symmetry to those local boundary conditions.

The paper does not separately vary gap width, shield termination and feed position, so it does not establish an optimum gap or apportion every effect between them. It demonstrates the combined structure's directional ratio. One sentence says “opposite to the shield gap,” but the surrounding definition, Figure 6 and the explicit φ = 0° description identify the ray crossing the gap. That is the geometry used here.

For a receiver, distinguish field coupling from circuit rejection. In a simple small-element model, a differential input voltage can contain both a magnetic-flux term and an unwanted electric term:

A_eff is an effective magnetic area, h_E an effective electric length, and both may be complex and frequency dependent. H_n is magnetic field normal to the loop; E_t is the relevant electric component. A_d and A_cm are differential and common-mode gains at a declared input interface. This is an explanatory model, not a fitted RF.Guru equivalent circuit. Even an ideal common-mode rejection ratio cannot subtract an unwanted term that has already become differential voltage.

For the passive reciprocal loop and feed structure, transmit-field asymmetry gives a reason to examine the corresponding receive coupling. It does not make a powered receive amplifier reciprocal, nor does one point's E/H ratio equal an antenna's E-field rejection rating. That requires calibrated receiving tests. The same distinction underlies my discussion of installed loop nulls.

How the paper measures E/H and material shielding

The authors calibrate a balanced electric probe and a balanced magnetic probe in an air-filled parallel-plate transmission line. Its port impedance is 50 Ω, while its approximately TEM field ratio is 377 Ω. Those numbers need not be equal: voltage and current depend on the plate dimensions; local E/H in the air-filled TEM region does not.

With the complex VNA transfer measurements identified by probe and source, their equation (4) is:

If probe outputs are K_EE and K_HH, the calibration ratio cancels K_E/K_H. That is why an uncalibrated voltage ratio is insufficient. Probe orientation, phase reference, loading and cross-sensitivity must remain controlled; the published comparison plots show magnitudes, although the acquisition and equation are complex.

The material experiment then measures the coupling between two loops with and without the sheet. In the usual positive-attenuation convention:

This is a transfer ratio for the specified source, receiving probe, sheet, distances and orientation. The sheet is a carbon-fiber veil, 300 μm thick, with reported measured conductivity 670 S/m. It is supported between polystyrene sheets and assembled into a 1.2 m square from two strips. In this setting, “shielding effectiveness” describes the barrier's effect on coupling; it is not the electrostatic screening of the loop's own inner conductor.

An independent scale check gives sheet resistance R_s = 1/(σt) ≈ 4.98 Ω/square. An ideal thin resistive sheet in free space at normal plane-wave incidence gives SE ≈ 20 log₁₀[1 + η₀/(2R_s)] ≈ 31.8 dB. This explains the approximate scale of the paper's roughly 32 dB plane-wave reference. It does not reproduce the finite-sheet near-field curves.

What the results establish—and where I would be stricter

The strongest result is the experimentally supported dependence on gap/feed azimuth. For much of the measured range, the orientation with higher transverse impedance gives lower material SE. Around the middle of the sweep the separation is substantial. But Figure 15 also shows the curves crossing toward the high-frequency end. I would preserve the authors' word “generally”; I would not turn it into a universal monotonic rule relating |E/H| to SE. Near-field excitation has spatial structure and phase, and the receiving probe samples the changed field.

Finite-sheet edge leakage is part of the answer. Figure 16 compares the 1.2 m square with a sheet extending to the computational boundary. The finite sheet gives lower SE, with a frequency-dependent difference attributed to fields reaching around its edges. A result from that fixture is therefore not an intrinsic material constant and cannot be reassigned to an antenna shield.

The measurements are open-laboratory measurements. No suitable anechoic chamber was available. The authors report sensitivity to nearby objects, particularly at larger separation, and omit field-ratio measurements beyond 350 mm. Differences from simulation become more pronounced above 100 MHz. Figures 12 and 13 support the orientation trend; they do not justify treating each simulated and measured ordinate as interchangeable. A formal uncertainty budget, raw complex traces and repeated rearrangements would be needed for a tighter numerical claim.

Figure 17 extends simulations to 500 MHz and shows the sheet SE curves approaching the plane-wave reference. That is a simulation comparison beyond the measured 10–200 MHz range. At 500 MHz, 150 mm is about a quarter wavelength; it is not a general far-field boundary. Even the ideal-dipole expression above gives about 512 Ω and 14.4° at that distance and frequency. That is not a prediction for the physical loop at 500 MHz, but it demonstrates why “approaching” should not become “already a 377-ohm plane wave.”

The standards introduction also needs a qualification. IEEE 299-2006 covers enclosure measurements over 9 kHz–18 GHz, with extensions; the paper associates its low-frequency loop method with the range through 20 MHz. Twenty megahertz is not the upper scope of the entire standard. The IEEE record now marks that edition inactive-reserved.

For NSA 94-106, 24 October 1994, §4.3.1, the publicly available transcription of the NSA-supplied specification lists magnetic tests at 1, 10 and 100 kHz and 1 MHz; electric tests extend to 10 MHz, with separate plane-wave points from 100 MHz to 10 GHz. The paper's grouped description of both standards as loop methods “up to 20 MHz” is therefore too broad. Its useful research question survives: can a carefully characterized loop fixture extend near-field material measurements to higher frequencies?

Mapping the mechanisms onto OctaLoop and OctaLoop Mini

For both products, my design rationale is to reduce unwanted capacitive coupling while preserving a useful differential loop signal, then control additional paths through the electronics and feedline. The paper strengthens the case for testing those pieces together. It supplies no numerical OctaLoop rejection, null-depth or SNR result.

The OctaLoop Mini overview identifies the documented OctaLoop3 Mini configuration: approximately 60 cm diameter and 1.9 m circumference, miniature 75-ohm coax, an approximately 4 cm shield gap at the top, a balanced input, symmetrical filtered gain paths and output combining. The assembly guide and photographs show perimeter cable routing and two terminal connections. They establish mechanical routing; the exterior photographs do not reveal the complete shield-to-reference termination network.

That top gap and the feed at the enclosure define a privileged axis. The direct hypothesis is that rotating this axis relative to a nearby source, while keeping the loop normal fixed, can alter unwanted electric coupling. The paper makes this a well-motivated test, not an already measured Mini characteristic. Its 10 mm gap in a 100 mm loop is also not geometrically equivalent to the Mini's roughly 40 mm gap in a roughly 600 mm loop. Even their approximate gap-to-circumference fractions differ: about 3.2% and 2.1%. Neither fraction predicts rejection on its own.

The full-size OctaLoop overview describes a larger shielded coaxial loop with differential push-pull electronics. The assembly instructions confirm that the larger model uses the same assembly method with larger tubing. That supports the shared mechanism, but does not establish identical electrical loading or an identical gap. I would not assign the Mini's gap dimension or termination map to the full-size model without checking that model's actual revision.

This is also why I keep outline geometry separate from shielding. An octagon can give repeatable construction and near-circular area, as explained in the loop-shape comparison. Eight sides do not cancel the feed discontinuity or establish electric-field immunity.

The balanced electronics address input symmetry. Feedline suppression addresses current on the cable exterior. Shielding addresses particular capacitive paths. Equal treatment of both input legs, including protection and parasitic capacitance, helps avoid converting unwanted common-mode excitation into differential signal. Nevertheless, the York loop already had a balanced source and still showed gap-related field asymmetry. That is the important constraint on my own design claims.

TerraBooster shares the screening idea, but the ground changes the problem

The TerraBooster configurations in my placement article are shielded 16, 32 and 56 m loops and an unshielded 84 m loop. The first two have 4 × 4 m and 8 × 8 m layouts. The 84 m version has no loop shield gap to which the York gap result could directly apply; feed asymmetry and environmental coupling still matter.

For the shielded configurations, the supplied TerraBooster3 design drawing specifies a shield interruption at the corner opposite the feed. Its circuitry shows transformer input coupling, paired filtering and attenuation paths, push-pull amplification, output combining, and an input shunt controlled by power state. The additional sheets show the supply/feedline network and separate RC reference arrangements for different loop groups. Those reference paths belong in a field-coupling model; they cannot be replaced by an assumed perfect RF ground.

The drawing also includes a transfer-sweep image. An electronics S₂₁ trace, without a calibrated E/H stimulus and complete test definition, cannot establish the assembled loop's electric rejection or installed SNR. I use the schematic to identify topology, not to turn its annotations into new antenna ratings.

The useful physical mapping is the deliberate shield interruption, feed symmetry and controlled referencing. The numerical comparison stops there. A ground-loaded loop interacts with lossy soil over its entire footprint. Soil permittivity and conductivity, moisture, insulation, nearby radials and the route to the receiver affect charge, current and field distribution. This is not a 100 mm loop in free space.

These ratios are not predicted resonances, operating-band approvals or measured patterns. At 10 MHz even the 16 m loop is about 0.53 wavelength around. The paper's frequency range therefore cannot be transferred merely because it overlaps HF. Maxwell scaling would require scaling the entire geometry, distances, loads and material behavior; the soil and powered front end make that a different experiment.

For TerraBooster I would examine the gap/feed direction relative to nearby wiring and radials, but also exchange physical positions and control soil conditions. Rotating a large ground loop can change which patch of soil and which nearby conductor each segment encounters. A measured change must be separated from those environmental changes before attributing it to the gap.

Field day and confined sites: separate the coupling mechanisms

A practical question behind this discussion is how closely antennas can coexist at field day or on a confined site. In the reader experience discussed in the TerraBooster placement article, a loop that usually improved SNR at home did not show the same advantage at a temporary site with a nearby quarter-wave vertical and radials. That observation motivates a coupling investigation; it does not identify the cause.

Mutual coupling means the antennas interact electromagnetically: a field from one induces current in the other, and that current changes the field and loading. The neighboring antenna need not be transmitting. Its connected receiver, open connector, short or tuner state changes its termination and can change the interaction. Reradiation is the induced current's secondary radiation, one way that interaction modifies the field at the receive loop. A fence, radial system or idle transmitting antenna can therefore change a receive pattern without producing noise of its own.

Overload is different. A sufficiently strong coupled signal can compress the active antenna or receiver, cause desensitization, or generate intermodulation. A shield and a balanced input do not reject energy already coupled into the wanted differential mode. Strong-signal behavior is a separate question from small-signal SE and from permanent-damage limits. My receive/transmit proximity discussion treats those distinctions.

Start by measuring from the nearest loop conductor to the nearest radial, antenna wire or metal object, not only from enclosure to mast. With nearby transmitters inactive, compare the wanted signal and noise while changing one neighbor's termination or the receive-loop position, then restore the baseline. A repeatable change under these small-signal conditions supports a coupling or field-distribution explanation, not transmitter-induced overload. Independently examine feedline pickup. Strong-signal tests need a controlled stimulus within the documented limits of the complete receive chain; attenuation after an already overloaded antenna amplifier cannot undo its distortion.

The York fixture deliberately uses close spacing to investigate a conductive barrier. Its 150 mm dimension is neither recommended antenna separation nor a transmitter safety distance. For a field-day layout, wavelength, element dimensions, relative orientation, neighboring resonances, termination, power, radials and soil all matter. Give the whole loop clear space, then verify the actual layout. There is no minimum separation or immunity radius established by this paper.

The tests that would settle the product questions

I would carry the paper's discipline into the receiving tests: characterize the exciting field, record the geometry and distinguish the desired response from unwanted conversion. I would not use its material SE curve as the acceptance curve for an antenna.

A local source called an “E-field source” still has H, and a magnetic source still has E. With two independently characterized excitations, solve V₁ = K_EE₁ + K_HH₁ and V₂ = K_EE₂ + K_HH₂ for the complex response coefficients. The excitation matrix must be well conditioned; two sources with almost the same E/H ratio cannot reliably separate the coefficients. Use additional positions and orientations to check the model. For a large ground loop, replace a single-point approximation with measured field distributions and a distributed model.

To compare electric contamination fairly, K_E/K_H alone is dimensional. Under a specified E/H ratio, report |K_EE|/|K_HH| and retain relative phase when predicting total output. This prevents an impressive-looking rejection number from depending on an unstated normalization.

For a paper-fixture replication, retain no-sheet/sheet/no-sheet baselines, measure both S₁₁ and S₂₁, and repeat the gap-azimuth comparison with the same support and cables. Sweep sample size to expose edge bypass. Check background pickup, probe cross-sensitivity, VNA noise floor and room sensitivity. Repeat at reduced drive to establish linearity. The geometric drawing should specify center and edge distances, not just “150 mm.”

For the receive antennas, use their normal powered state; power-off shunting changes the input boundary condition. Never infer the powered receive response by measuring the unpowered input. Normalize wanted-field sensitivity before claiming an unwanted-coupling improvement, so attenuation of everything does not masquerade as better rejection.

My design conclusion

I take this paper as a strong reason to keep designing and measuring the shield, gap, input and feedline as one system. The screening method used in shielded OctaLoop-family and TerraBooster elements is physically relevant to the study, but it does not make their behavior identical to the York fixture.

For the Mini, the documented gap gives a concrete azimuthal-coupling test. For the full-size OctaLoop, that same test starts by confirming the exact gap and termination of the revision under test. For shielded TerraBooster, it must include the reference network, electrical length and ground. For the unshielded 84 m element, it is a feed-and-environment study without a shield-gap claim.

The useful objective is lower unwanted coupling while retaining the wanted signal, with enough headroom for the actual site. A gap is necessary to the intended screening topology, balance is valuable, and neither grants universal E-field immunity. The York results sharpen that engineering argument; the product result still belongs to a calibrated measurement of the product.

Questions, antenna-factor records or height trials to share? Contact RF.Guru.