How Does a Septum Polarizer Work?

A septum polarizer works by using a shaped metal partition inside a waveguide to produce two orthogonal electric-field components with nearly equal amplitudes and a 90-degree phase difference. Together, those components form circular polarization at the common port. Exciting one of the two separate ports produces one circular sense; exciting the other produces the opposite sense. In receive, an ideal device routes the two incoming circular senses to different ports. [1]

The septum does two jobs at once: it separates or combines polarization channels and introduces the required differential phase shift. The common waveguide may be square or circular. The separate ports are rectangular in many square-guide designs, but circular-guide designs can use semicircular branches with transitions to rectangular interfaces. The three physical ports therefore do not all carry the same number of polarization modes. [2]

How the Septum Creates the 90-Degree Phase Shift

In a square common guide, the two orthogonal fundamental modes are TE10 and TE01. A circular common guide instead supports two orthogonal polarizations of TE11. These are the two field components that must emerge with the correct amplitude balance and phase relationship. Their horizontal or vertical labels depend on the chosen coordinate system. [1] [2]

The mechanism is easier to understand by considering two excitations of the separate ports. For the in-phase, or even, excitation, the currents on opposite faces of the septum close with relatively little disturbance. For the out-of-phase, or odd, excitation, the septum changes the current path and impedance much more strongly. The resulting ridge-like loading changes the propagation constant and accumulated phase of that field component. [1] [2]

Driving just one port is a superposition of those even and odd excitations. The septum’s step heights and lengths are chosen so that the two contributions reach the common port approximately 90 degrees apart, while keeping reflections low. Several steps distribute the phase correction and impedance transformation across a band rather than making a single-frequency quarter-wave section. This is differential phase delay, not a claim that the signal envelope experiences a constant quarter-cycle time delay at every frequency. [2]

Thickness matters as well as height and length. Dubrovka and colleagues demonstrated a three-section stepped-thickness prototype for 7.70–8.50 GHz, reporting measured VSWR below 1.4 and cross-polar discrimination above 30 dB with the required transitions and bends included. Those are results for that assembly, not universal specifications for a stepped septum. [3]

How Linear Input Becomes Circular Output

At a fixed point in the common guide, equal orthogonal components in phase quadrature make the electric-field vector rotate with constant magnitude. That is circular polarization. Unequal amplitudes or a phase difference other than 90 degrees generally produce an ellipse instead. The 1973 square-guide prototype reported about 0.2 dB amplitude difference between its orthogonal components; the authors noted that near-unity axial ratio also required the intended phase relationship. [1]

Changing the driven branch reverses the relative sign of the components, giving the opposite circular sense. In the ideal reverse process, one incoming circular sense adds at one branch and cancels at the other. A linearly polarized wave at the common port divides equally between the two circular channels; their relative output phase depends on the linear polarization angle. Real devices have finite leakage and imbalance. [2]

How to Identify RHCP and LHCP

Under the IEEE convention, clockwise electric-field rotation viewed in the direction of propagation is right-hand circular polarization, or RHCP. Counterclockwise is left-hand circular polarization, or LHCP. Reversing the viewing direction reverses the apparent rotation, so a port label is incomplete without a propagation arrow and viewing convention. [4]

For a drawing or acceptance test, specify the driven branch, the direction through the common port, and the convention used. Do not infer the receive-port assignment simply by looking back into a transmit aperture. Use the documented reciprocal port mapping and a known polarization reference.

What Limits the Operating Bandwidth?

The differential phase shift, amplitude balance, impedance match, and higher-order modes all vary with frequency. For an empty square guide, the next-mode cutoff is about 1.414 times the fundamental cutoff, giving an approximately 34% interval around the arithmetic center frequency. Circular guide has a narrower corresponding interval, about 26%. These are conventional mode-window estimates, not absolute bandwidth ceilings for every loaded polarizer. Chen and colleagues summarize traditional usable bandwidths around 20% for square and 12% for circular guide, with refinements reaching roughly 25% and 15%. [2]

The same researchers reported a five-step circular-guide design covering 77–118 GHz, or about 42% fractional bandwidth. Their VNA covered 75–115 GHz, and polarization results were derived over 77–115 GHz; the full 77–118 GHz design range should not be described as fully measured. The design controls higher-order-mode resonances, and its performance depends on the boundary conditions presented by the connected horn or test adapter. [2]

What Loss, Return Loss, and Isolation Actually Mean

Insertion loss measures transmission reduction for a defined excitation and output channel. Return loss measures reflection. Isolation measures unwanted coupling between specified ports. Cross-polar discrimination measures the wanted-to-unwanted polarization power ratio. They are related through the device response, but they are not interchangeable specifications.

Keep the sign convention straight: an S-parameter magnitude of −20 dB corresponds to 20 dB return loss or isolation, depending on the ports involved. For return loss, 20 dB means 1% reflected power and VSWR about 1.22; 10 dB means 10% reflected power and VSWR about 1.92. These are calculated from reflected power = 10^−RL/10 and VSWR = (1 + |Γ|)/(1 − |Γ|), where |Γ| = 10^−RL/20.

For a defined matched-port measurement, 30 dB isolation means a coupled-power ratio of 0.001, or 0.1%. It does not by itself prove that a receiver can tolerate the transmitter. The published 225 GHz duplexer reported 30 dB isolation, but receiver protection still requires a system-level leakage budget. [6]

Measurement boundaries also explain apparently different loss figures. Chen et al. reported about 0.2–0.3 dB additional transmission loss relative to ideal simulation in their W-band test, partly attributable to the preceding adapter and following splitter. That is not a de-embedded bare-polarizer loss specification across the full design range. [2]

How Axial Ratio Reveals Conversion Quality

Small Mismatch Loss Can Still Hide Poor Polarization Purity

Axial ratio is the major-to-minor axis ratio of the polarization ellipse. An ideal circle has a linear ratio of 1 and an axial ratio of 0 dB. [4] For a fully polarized elliptical wave, its decomposition into the two circular senses gives the following calculation. Let r = 10^ARdB/20 and q be the unwanted-to-wanted circular power ratio:

q = ((r − 1)/(r + 1))²
XPD = −10 log10(q)
η = 1/(1 + q) = (r + 1)² / (2(r² + 1))
Mismatch loss = −10 log10(η) = 10 log10(1 + q)

Here η is the fraction coupled into an ideal, matching-sense circular receiver. The calculation follows circular-component decomposition and polarization-vector overlap; it excludes conductor loss, reflections, and all other link losses. Because one polarization is perfectly circular, ellipse tilt does not change this result. Two imperfect elliptical antennas require their full polarization relationship instead. [8]

A calculated 3 dB axial ratio loses only about 0.125 dB into an ideal matching circular receiver, yet its unwanted circular component is about 2.9% of the wanted component. A modest link-budget penalty does not guarantee enough polarization isolation for frequency reuse.

Phase and Amplitude Errors Must Both Be Checked

For equal component amplitudes and a phase difference of 90° − ε, with 0° ≤ ε < 90°, ARdB = 20 log10(cot((90° − ε)/2)). At exact quadrature, amplitude imbalance alone sets axial ratio in dB. The general ellipse relation includes both amplitude and phase. [8]

As a combined-error example, 0.5 dB amplitude imbalance and a 10° departure from quadrature produce about 1.60 dB axial ratio. That is a calculation, not a measured part specification. It explains why equal transmission magnitudes alone cannot establish good circular polarization.

Why Geometry Also Affects Power Handling

Sharp step edges can concentrate the electric field. Kim, Kovitz, and Rahmat-Samii reported a 40% reduction in peak electric field for an optimized smooth sigmoid profile compared with their stepped reference in full-wave simulation. This was a profile-specific simulation result, not a measured breakdown rating or a universal 40% power improvement. [9]

A low-power S-parameter sweep does not demonstrate high-power survival. Request a rating with frequency, peak and average power, duty cycle, pressure or vacuum conditions, materials, and the supporting analysis or test method. A smaller gap does not imply a universally lower breakdown voltage; breakdown and multipactor require their own operating-condition assessment.

Where Septum Polarizers Fit in Feed Systems

In an antenna feed assembly, the common port connects toward the horn and the separate branches connect toward the polarization channels. A septum junction can combine the jobs otherwise performed by an orthomode transducer and a separate polarizer. It is therefore misleading to assume that every septum installation needs another OMT behind it. Béhé and Brachat describe this combined duplexer-polarizer function in circular guide. [5]

A documented instrument example appears in Hunter and Napier’s 2016 review: the VLA Q-band receiver, covering 40–50 GHz, uses a septum polarization splitter. This is a historical receiver configuration described by that review, not a statement that every current radio-astronomy receiver uses the same architecture. [10]

The 225 GHz duplexer is another example: opposite circular channels enable ferrite-free duplexing in the reported design. That does not make a septum a nonreciprocal circulator, or guarantee isolation from arbitrary reflected signals. [6]

How to Test the Working Principle Without Misreading the Fixture

The common guide supports two polarization modes, so a complete test must account for both. With a calibrated VNA and suitable mode-launching transitions, measure the complex response for each orthogonal common-port excitation, recording amplitude and phase at both branches. State the reference planes and terminate unused ports appropriately. For interbranch isolation, document the common-port termination or radiating horn and its higher-order-mode behavior. Chen et al. used separate configurations for these measurements. [2]

For an ideal linearly polarized common-port input, each circular branch receives half the input power: 10 log10(1/2) = −3.01 dB. That calculated split is not 3 dB of dissipative loss. Likewise, similar branch magnitudes are not enough to prove the correct differential phase. For an assembled horn, a calibrated rotating linear probe can determine axial ratio from 20 log10(Vmax/Vmin), but that amplitude-only measurement does not identify handedness.

A Published Adapter-Length Check

Chen et al. tested common-port adapters with circular sections of 0.2 mm, 5 mm, and 10 mm. The observed resonances changed with adapter length, and simulations of the complete measurement configurations reproduced them. Together, those checks supported the interpretation that the resonances came from the external cavity formed by the adapter rather than a fixed internal cavity mode. A moving dip is evidence of fixture interaction, not automatic proof that the polarizer is fault-free. [2]

If a narrow feature appears during acceptance testing, preserve the original trace, change a controlled fixture parameter, and compare against a model of the complete setup before blaming the blade. This is a diagnostic approach grounded in the published experiment, not a substitute for the part’s acceptance limits.

Why Small Dimensional Errors Matter

Geometric scaling illustrates the manufacturing challenge. A 0.1 mm feature at 3 GHz scales to 0.025 mm at 12 GHz and 0.6 µm at 500 GHz if every dimension scales inversely with frequency. Those are calculated feature sizes, not tolerance requirements. Hasegawa et al.’s actual 500 GHz-band device achieved flatness within 3 µm and horizontal alignment within ±3 µm. At 500 GHz, 3 µm is about 0.5% of the 0.600 mm free-space wavelength. [7]

The practical question is how the attainable edge radii, septum thickness, surface finish, and alignment change phase balance and matching. Include those features in the electromagnetic model instead of assuming that a scaled drawing preserves the original performance.

A 12 GHz Calculation: Useful Starting Point, Not a Finished Design

Consider an ideal, air-filled square guide with side a = 15.8 mm and a proposed 10.8–13.2 GHz band. Using c = 299,792,458 m/s, the fundamental cutoff is fc10 = c/(2a) = 9.487 GHz. TE01 has the same cutoff. The next-mode cutoff is fc11 = √2 × fc10 = 13.417 GHz.

That next-mode cutoff is only 0.217 GHz above the proposed upper band edge, about 1.64% of 13.2 GHz. This is a warning about limited modal margin, not proof that the design will work. Septum loading, transitions, and corner geometry still need full-wave analysis.

At 12 GHz, λ0 = 24.983 mm and λg = λ0/√(1 − (fc10/f)²) = 40.80 mm. Thus λg/4 is about 10.20 mm for the empty guide. It is not a prescribed step length: the septum changes the modal propagation constants, and the required quantity is the relative phase accumulated by the two components.

A septum polarizer works when the geometry delivers the required amplitude balance, phase quadrature, matching, and channel isolation over the intended band. Verify those together, with the actual interfaces and operating conditions, rather than treating a 90-degree center-frequency result as the whole specification.