Correlative Interferometer
How Manifold Correlation Replaces Phase Inversion
The defining idea of the correlative interferometer is that direction finding is treated as a pattern-matching problem rather than a trigonometric inversion. During a calibration campaign, the installed array is illuminated by a known reference source at hundreds of azimuth angles (typically every 1 to 5 degrees) and across the operating band. For each angle the receiver records the complex response of every element relative to a reference channel, producing a manifold: a lookup table of expected phase and amplitude vectors indexed by azimuth, elevation, and frequency. In operation, an incoming pulse or continuous-wave emission produces one measured vector, and the processor computes a normalized correlation between that vector and every manifold entry. The peak of the correlation surface is the estimated bearing.
This structure gives the technique its two signature strengths. First, ambiguity resistance: a long baseline whose raw phase wraps through several cycles is not a problem, because the wrapped pattern itself is what was stored during calibration, so non-uniform and widely spaced apertures become an asset rather than a liability. Second, fidelity to the real installation: mutual coupling between elements, dielectric radome effects, and scattering off a ship mast or aircraft fuselage are all baked into the manifold, so the system reports the bearing that physically corresponds to the measured response rather than an idealized free-space prediction. The cost is a calibration effort that must be repeated whenever the antenna geometry or nearby structure changes.
Compared with amplitude-comparison direction finders, which infer bearing from the relative power between squinted beams, the correlative interferometer extracts far more information per measurement because it uses the full complex response of every element. Compared with single-baseline phase interferometers, it trades a small amount of additional hardware and the calibration burden for substantially better accuracy and robustness, which is why it became a workhorse of wideband electronic-support and emitter-location systems.
Correlation and Bearing-Estimate Equations
Δφmn(θ) = (2π / λ) × dmn × cos(θ − αmn)
Normalized correlation against manifold vector a(θ):
C(θ) = |xH · a(θ)|2 / [ (xHx) × (a(θ)Ha(θ)) ]
Estimated angle of arrival:
θ̂ = arg maxθ C(θ)
Where λ = wavelength, dmn = baseline length between elements m and n, αmn = baseline orientation, x = measured complex response vector, a(θ) = stored manifold vector, and superscript H = Hermitian (conjugate) transpose. C(θ) ranges 0 to 1; values ≈ 1 indicate a clean single-emitter match while a depressed peak flags multipath or co-channel interference.
Direction-Finding Method Comparison
| DF Technique | Typical Accuracy (RMS) | Antenna Count | Ambiguity Handling | Calibration Need | Best Use |
|---|---|---|---|---|---|
| Correlative interferometer | 1 to 3° | 5 to 9 (UCA) | Implicit via manifold | High (full manifold) | Wideband ESM / emitter location |
| Phase interferometer | 1 to 5° | 3 to 5 | Multi-baseline integer search | Moderate | Narrowband, fast DF |
| Amplitude comparison (Watson-Watt) | 3 to 10° | 4 to 6 squinted | Not applicable | Low | Compact, low-cost warning |
| MUSIC superresolution | < 1° | 6 to 16 | Eigen-subspace search | High (manifold + SNR) | Multi-emitter, dense signals |
| Monopulse (single beam) | 0.05 to 0.5° | 2 to 4 (sum/diff) | Single-lobe only | Moderate | Tracking a known target |
Frequently Asked Questions
How does a correlative interferometer resolve phase ambiguities that defeat a simple phase interferometer?
It never inverts the phase equation. Instead it correlates the full measured vector of inter-element phase differences against a stored manifold of phase vectors recorded at every calibrated azimuth (typically 1 to 5° steps). The wrapped phase pattern itself becomes the fingerprint identifying the true bearing, so long or non-uniform baselines that would be ambiguous for a two-element interferometer are handled naturally. The estimate is the manifold entry that maximizes the normalized correlation C(θ).
What antenna spacing and how many elements does a correlative interferometer DF array need?
Practical systems use 5 to 9 elements in a circular or non-uniform circular array. A 0.5 to 18 GHz electronic-support array might be a 5-element UCA about 0.3 m across, which is electrically small (about half a wavelength) at the bottom of the band and many wavelengths across at the top, so the aperture spans a fraction of a wavelength to tens of wavelengths over a decade-plus band. Half-wavelength spacing is not required because the manifold absorbs the true element pattern, mutual coupling, and platform scattering. Well-calibrated 5 to 7 element arrays reach 1 to 3° RMS, and larger arrays do better than 1°.
How is the array manifold calibration table generated and maintained?
The array is illuminated by a calibrated source at many azimuth angles, often several elevations and frequencies, while the complex response of each element is recorded relative to a reference. For platform-mounted arrays this is done with the full vehicle, aircraft, or ship present so scattering and coupling are captured. The result is a lookup table indexed by azimuth, elevation, and frequency, with interpolation between calibrated frequency points. It must be regenerated if antenna placement, the radome, or nearby structure changes.
What degrades correlative interferometer accuracy in the field?
The main culprits are multipath and co-channel interference, which distort the measured vector so its correlation peak broadens or shifts; low signal-to-noise ratio, which adds variance roughly inversely with SNR and aperture; manifold staleness after a mechanical or structural change; and frequency-interpolation error between sparse calibration points. A depressed peak correlation coefficient (well below 1) is the practical health indicator that flags these conditions in real time.