The Destructive Null: When Directionality Bites Back

Directionality is built from addition and cancellation. Calling every deep cancellation “good” is as misleading as calling every null a fault. A useful receive null must suppress the unwanted field while preserving the wanted signal over the frequency, arrival angles, polarization and time that matter.

A Null Is a Complex-Voltage Cancellation

An array combines complex channel voltages: each contribution has both amplitude and phase. For weight vector w and array response vector a for one direction, frequency and polarization, a compact expression for the combined response is:

The superscript H denotes the complex-conjugate transpose. A null occurs where that weighted sum approaches zero. The cancellation includes spatial phase from element positions, electrical phase and loss through the channels, the installed element responses and mutual coupling. It is not explained by a delay cable or hybrid angle alone.

With two nominally equal channels, the destructive result can be written as 1 − ge^jδ, where g is the amplitude ratio and δ is the remaining phase error. Perfect equality with 180° cancellation gives zero in the ideal model. Small amplitude or phase errors leave a residual. That is why the bottom of a very deep null is usually more sensitive than the main lobe to cable loss, component tolerance, temperature, coupling and calibration.

Null depth also needs a reference. “30 dB null” is incomplete unless it says whether the residual is compared with the main-lobe response, the opposite switch state, one element or another declared reference, and whether the value came from voltage, power, a model or an installed measurement.

The Null Must Follow the Interference

A fixed array pattern suppresses a source only while the source energy arrives through the nulled part of the pattern. That can work extremely well for a stable local carrier or a geographically fixed source dominated by one arrival direction.

The same sharp null becomes a liability when:

• the wanted signal arrives near the nulled azimuth and elevation;
• the interference reaches the array over several reflected or skywave paths;
• arrival angle or polarization changes with propagation;
• the operating frequency moves far enough to change electrical spacing and channel phase;
• nearby conductors reradiate the field from another direction; or
• channel amplitude, phase or group delay drifts after calibration.

A compass heading is therefore not enough. A receive pattern is three-dimensional and polarization-sensitive. An azimuth null measured at one elevation can leave energy arriving at another elevation almost untouched. A single narrow null also cannot erase a spatially distributed noise field or multipath energy that occupies several angles.

ITU guidance on antenna measurements warns that reflections and reradiation can create or fill apparent pattern nulls. LZ1AQ's practical receive-array work makes the same HF point from field measurements: a controlled local source can produce a deep repeatable cancellation, while skywave amplitude and phase variation can make an instantaneous front-to-back result move substantially.

Null Depth, Front-to-Back and RDF Are Different

A pattern with a spectacular single-point null can still have a modest RDF if broad sidelobes admit noise from much of the sky. Conversely, an array with useful RDF may not put its deepest null on the local source that is ruining reception. The operator needs the metric that matches the problem.

Noise Covariance Changes the Best Combination

In a receiving array, the channels do not necessarily contain independent noise. Atmospheric noise, local RFI and reradiated fields can appear in several elements with correlated amplitudes and phases. Receiver noise may be partly independent; external noise often is not.

If R_i+n is the interference-plus-noise covariance matrix, the combined unwanted output power is:

This is why a free-space polar plot is not a complete SNR prediction. It shows the response to coherent plane waves under the model assumptions; it does not by itself describe the site's spatial noise distribution, channel correlation, receiver noise or overload.

An adaptive array can estimate channel correlations and adjust complex weights to improve signal-to-interference-plus-noise ratio or to place a null on an interferer. The desired response still needs a constraint or trusted reference. Otherwise, an optimiser that only minimizes output power can make the receiver wonderfully quiet by cancelling part of the wanted signal too.

The NTIA's steerable-nulling work describes this as spatial filtering: the processor senses the signal, noise and interference environment, then adjusts amplitude and phase weights. ITU-R's interference-mitigation report adds an important boundary: adaptive nulling works best when the relevant interferers can be detected and tracked and remain sufficiently stable during the observation.

A Hybrid Angle Is Not an Insurance Policy

A 45° or 90° network can be the right building block for a declared geometry. The angle by itself does not guarantee safe multiband behaviour, a particular RDF or freedom from unwanted notches.

As frequency changes, physical spacing becomes a different fraction of a wavelength. Element response, mutual coupling, feedline transformation and network phase and amplitude tracking also change. The resulting lobes and nulls come from the complete complex response. There is no universal spacing such as 0.139λ or 0.16λ that guarantees optimum RDF or prevents destructive nulls for every three-element array, element type, height, ground, combining network and frequency range.

Simple array-factor code can reveal many dangerous combinations when the geometry and complex weights are correct. A full-wave model becomes important when embedded element patterns, coupling, ground and nearby structures materially affect those weights or responses. Neither model replaces an installed pattern and SNR check.

Design for a Useful Rejection Region

The deepest possible mathematical point is rarely the best operational target. A slightly shallower but wider and more stable rejection region may suppress real interference better across a band, across changing arrival angles and after normal component drift.

Choose the design around the operating question:

• For one stable local source, prioritize repeatable attenuation at its measured arrival direction.
• For changing skywave interference, compare several steering states or use constrained adaptive combining.
• For general low-band noise discrimination, evaluate RDF, beamwidth and sidelobes rather than celebrating one null.
• For multiband use, qualify every intended band and steering state independently.
• When wanted and unwanted energy share frequency, direction and polarization, spatial processing cannot cleanly separate them.

Measure the Installed Array

1. Declare the geometry. Record element coordinates, height, orientation, feedline routes and the forward-direction convention.
2. Calibrate the channels. Measure amplitude, phase or group delay through every path at declared reference planes over the operating band.
3. Check element equality. Swap elements and channels to expose embedded-pattern, amplifier, cable and termination differences.
4. Map more than one cut. Measure azimuth at relevant elevations and repeat across frequency; record null direction, depth, width, main lobe, sidelobes and F/B.
5. Test the actual interference. Compare the wanted signal and the unwanted-plus-noise level using simultaneous channels or rapid A/B/A switching with unchanged receiver settings.
6. Repeat over time. Recheck temperature, weather, propagation and installation changes. A null that cannot be found again is not an operational specification.

NIST's over-the-air phased-array calibration work is at much higher frequency, but it demonstrates a transferable measurement lesson: calibration can be steering-state dependent. The HF hardware is different; the need to characterize the complete signal path for each useful state is not.

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