One Month of Phasing: What We Learned About Receive Arrays

This was not a single formal campaign with one frozen array and a complete uncertainty budget. It was a month of iterating: change the geometry, rerun the model, measure the channels, listen, restore the baseline and find the next weak assumption. The value is not a universal recipe. It is the order in which the array started telling the truth.

RDF Became Important—Not Omnipotent

Receiving directivity factor, or RDF, compares an antenna's response in its best direction with its average response over all directions. It is a useful way to describe how strongly a receive pattern favours one region of space. It does not by itself mean “more wanted signal and less noise.”

If the wanted signal arrives near the main response and unwanted noise is broadly distributed, higher RDF can improve SNR. If one coherent local source dominates, a correctly placed null may matter more. If the wanted path arrives at another azimuth, elevation or polarization, the same high-RDF pattern can reject the signal we wanted.

The month changed the question from “How high is the RDF?” to “Does this pattern improve the wanted-signal SNR for this path, frequency and noise field?” No universal RDF target answers that.

A Rear Null Is Useful Only Where the Interference Is

Deep simulated rear nulls are seductive. In the field they can be narrow in angle and frequency, sensitive to channel mismatch and easily filled by common-mode pickup, local scattering or multipath.

That does not make rear nulls pointless. A stable interferer from a known bearing can make a narrow null extremely valuable. The lesson was to stop treating “front-to-back” as a substitute for the full pattern. Record null direction, width, depth above the measurement floor, frequency span and stability after reconnecting the array.

Spacing Is Mathematics With Installation Attached

Element spacing changes spatial phase, lobe width, ambiguity and mutual coupling. It does not have one sweet spot for every element, band, soil and objective. A value that gives a useful ideal array factor can still fail when embedded element patterns, feedline currents or ground interaction are included.

For a plane wave, the phase contribution from an element-position difference Δr is proportional to 2πf(s · Δr)/c. The frequency f, arrival-direction vector s and physical geometry therefore belong in the same calculation. Quoting spacing only as a fraction of wavelength hides which frequency and which desired direction were used.

During the month, the useful spacing was the one that survived three checks: converged modelling, measured complex element channels and an installed over-the-air pattern or SNR comparison. A recurring model value was a candidate—not a law.

The Geometric Mean Is a Starting Frequency, Not an Optimizer

For two widely separated bands, the geometric-mean frequency or wavelength can provide a logarithmically centred design point. That can be convenient when the error metric is symmetric in fractional frequency. It does not automatically balance RDF, null depth or pattern quality.

The two bands can have different element transfer, external noise, ground interaction, mutual coupling, feed loss and operating priorities. A multi-objective sweep is more honest: calculate the complete pattern on every required band, apply a stated weighting to the metrics, and show what each compromise gives up.

Fixed Phase Did Not Defeat Delay

A fixed-phase hybrid can make a compact, repeatable passive network around a design region. A true time delay produces phase proportional to frequency and can preserve an arrival-time compensation over a wider fractional span. Neither architecture universally wins.

The hybrid's phase difference, amplitude balance, insertion loss, isolation and return loss vary with frequency and port termination. The delay line has attenuation, velocity-factor tolerance, dispersion, connectors and temperature effects. The right comparison uses measured complex transfer at the same reference planes—not the labels “broadband” and “band-specific.”

Low Mounting Did Not Remove the Ground

Small receive elements can often be installed low and still form useful directional arrays. That operational convenience is real. It does not prove that elevation has little effect or that a low array automatically has a low-angle response.

Element height changes ground coupling, loss, embedded pattern, mutual coupling and the phase and amplitude reaching the combiner. At short range, a test source may also be in the near field rather than representing a skywave arrival. Height must be included in the complete geometry and varied without silently changing cable routes or common-mode boundaries.

Common-Mode Control Became Part of Calibration

An array assumes that each channel represents its intended receiving element. Exterior current on coax, power or control leads adds uncontrolled antennas. Those conductors can fill a null, bias the apparent main direction and make two nominally identical branches respond differently.

Electrical isolation can help, but “isolated” is not a measured common-mode result. Measure exterior current on every branch, vary cable routes and common-mode impedances one change at a time, and recheck the complex channel transfer afterward. A choke can alter both unwanted pickup and the intended element reference, so its effect belongs in the array calibration.

Models Were Most Valuable When They Could Be Falsified

A simple array factor is excellent for exposing phase conventions and broad spacing behaviour. NEC can add conductor geometry, ground and coupling when the model is converged and its limitations are respected. Neither automatically contains active-device noise, compression, real cable common mode or the site's moving noise sources.

The month worked best when every model output led to a field test capable of disproving it. If moving one cable changed the null, the cable belonged in the model. If swapping two channels moved the asymmetry, the electronics or feed path needed calibration. If neither model nor test repeated after the baseline was restored, there was no result yet.

Ears Found Problems; Records Established Results

Listening remains useful. The ear quickly notices an impulsive source disappearing or weak speech becoming intelligible. It is also influenced by level, AGC, bandwidth, fading and expectation.

So the final habit was simple: keep the receiver state fixed, record signal and noise in the same bandwidth, alternate configurations quickly or simultaneously, and return to the first condition. Trust the ears to find a promising change; trust repeatable records to decide whether it survives.

The Measurement Order We Kept

• Define the task: wanted headings and elevation region, bands, polarization, interfering directions, bandwidth and receiver arrangement.
• Measure each branch: complex element-channel transfer, noise, overload and common-mode current at declared planes.
• Characterize the combiner: amplitude/phase balance, insertion loss, isolation, return loss, group delay and termination state across frequency.
• Model the whole geometry: embedded elements, ground, coupling, supports and cable routes—not only isotropic points.
• Inspect the full pattern: RDF, wanted-direction response, nulls, lobe count and ambiguity over every required band.
• Test installed SNR: use a stable signal or source, fixed receiver settings and simultaneous or A/B/B/A comparisons.
• Repeat after disturbance: reconnect, swap channels and remeasure so a fragile calibration does not masquerade as a robust array.

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