🛰 SAR · Satellite Imagery
Foundations — the physics of signal & image

Complex signals & phase math

I/QEuler's formulaM×S*wiki/sar-complex-signal-phase-math
TL;DR

A single SLC pixel is one complex signal, I + jQ, whose amplitude holds the backscatter strength and whose phase holds the round-trip timing of the wave (its rotation angle). The real reason for treating phase as a complex number is computation, not storage — thanks to Euler's formula e^(jθ)=cosθ+jsinθ, a rotation reduces to a single multiplication. An interferogram is Master × Slave* (the conjugate of the slave), and this multiplication automatically extracts the phase difference θ₁−θ₂, which equals the path-length difference and thus converts into elevation and displacement. That θ₁−θ₂ is exactly what SNAPHU solves for.

An SLC pixel is a complex signal (I/Q)

  • A radar wave is amplitude plus a rotating oscillation, so a received pixel can be written as coordinates (I, Q) or as the complex number I + jQ — two notations for the same information.
  • The amplitude A = √(I² + Q²) is the backscatter strength, i.e. how much of the transmitted wave came back.
  • The phase θ = atan2(Q, I) holds the round-trip timing, i.e. the rotation angle, and this phase is the heart of InSAR.
  • This is precisely where GRD diverges as a product that keeps only amplitude and discards the phase.
📡Received waveamplitude + phase
SLC pixelI + jQ
Amplitude Abackscatter → GRD
Phase θround-trip → InSAR
From received wave to the complex SLC pixel, then the amplitude/phase split

Why complex numbers — rotation is multiplication

  • The key is not that complex numbers are convenient for storage, but that they reduce a rotation to a single multiplication.
  • The x/y coordinate approach must run trigonometry per pixel — x' = x·cosθ − y·sinθ, y' = x·sinθ + y·cosθ — exploding the cost over millions of pixels.
  • The complex approach finishes with one multiplication by e^(jθ) and needs no trig calls at all.
  • For example, rotating the signal (1,0) by 90° requires the rotation formula in coordinates, whereas in complex form you simply multiply by j to get (0,1).
x/y coordinatesComplex numbers
Rotation oprotation formula x·cosθ − y·sinθ …one multiply by e^(jθ)
Trig functionssin/cos explode per pixelnone
Handling a rotation: coordinates vs complex numbers

Euler's formula — a rotation as a single number

  • A point on the unit circle is (cosθ, sinθ), or cosθ + j·sinθ as a complex number, and Euler's formula states that this remarkably equals e^(jθ).
  • So a wave is written compactly as A·e^(jθ) — a single vector of length A and angle θ on the I/Q plane.
  • SAR engineers use Euler's formula not for its proof but because it expresses a rotating wave in one line and reduces phase operations to multiplication.
amplitude (vector length)
phase (rotation angle)
imaginary unit (a 90° rotation)
Euler's formula: a rotating wave as the single term A·e^(jθ)

The conjugate and the interferogram = M × S*

  • InSAR wants not the absolute phase but the phase difference θ₁ − θ₂ between two acquisitions, yet a plain product adds the phases (θ₁+θ₂) and becomes meaningless.
  • So multiplying by the slave's conjugate — e^(−jθ₂) with the phase sign flipped — yields AB·e^(j(θ₁−θ₂)), automatically extracting the phase difference.
  • By the clock analogy, with Master 30° and Slave 20°, plain addition gives a meaningless 50°, but flipping the slave to −20° gives the desired relative angle of 10°.
  • SNAP's Interferogram Formation step is effectively a per-pixel phase-difference calculator running this M × S*, and what SNAPHU solves is neither the master nor slave phase but exactly this θ₁ − θ₂, which becomes a path-length difference ΔR and thus elevation and displacement.
MasterA·e^(jθ₁)
Slave conjugate S*B·e^(−jθ₂)
✳️M × S*AB·e^(j(θ₁−θ₂))
Path diff ΔR→ elevation / displacement
The M × S* product yields the phase difference, leading to path difference and displacement
Pitfalls & gotchas

If you memorize complex numbers as merely a storage format, M×S* never clicks — you must grasp the computational motive that rotation equals multiplication, which makes the conjugate's sign flip feel natural. The sole reason it is M×S* rather than M×S is that we want subtraction (θ₁−θ₂), not addition (θ₁+θ₂), and a sign mistake flips the entire direction of the phase difference.

The topic cards on this page are compiled from the Brain Trinity wiki. The original wiki can be demoed live in an interview.Back to study log