Electronics Guide

Cavity Resonant Modes

Consider a rectangular cavity whose dimensions are a, b, and c, with a the largest dimension and c the smallest. The resonant frequencies of its TE and TM modes are:

f_mnp = (c₀/2)√((m/a)² + (n/b)² + (p/c)²)

Here c₀ is the speed of light and m, n, and p are non-negative integers that index the field variation along each axis. With the p index taken along the shortest dimension, TM modes require m and n to be nonzero while p may be zero, and TE modes require p to be nonzero with at least one of m and n nonzero.

Each mode has a unique field distribution within the cavity. At resonance, the fields form standing-wave patterns with nulls and maxima at specific locations. The spacing between modes in frequency is not uniform. It tends to be larger at lower frequencies and decreases as frequency increases.

Mode Density Calculation

The number of modes with resonant frequencies below a given frequency f, in a cavity of volume V, is approximated by the leading term of Weyl's formula:

N(f) ≈ 8πVf³/(3c₀³)

Taking the derivative with respect to frequency gives the smoothed mode density:

dN/df ≈ 8πVf²/c₀³

Mode density therefore grows with the square of frequency. At low frequencies, modes are sparse and well separated. As frequency rises, the number of modes per unit bandwidth grows rapidly, eventually producing a quasi-continuous distribution. Weyl's expression is an asymptotic count that smooths over the discrete modal structure, so it is reliable only once the chamber holds many modes.

For practical chambers, the mode density determines the lowest usable frequency (LUF). A commonly cited criterion requires roughly sixty to one hundred modes to exist below the operating frequency, and an equivalent rule of thumb places the LUF at three to six times the chamber's lowest resonant frequency. Below the LUF, the statistical assumptions underlying reverberation chamber theory break down, and field uniformity cannot be achieved. IEC 61000-4-21 added guidance in its second edition on operating near the low-frequency limit and on the effect that such operation has on the estimated field uncertainty.

Modal Overlap Factor

Each resonant mode has a finite bandwidth set by the losses in the cavity. The 3 dB modal bandwidth is related to the quality factor by:

Δf = f/Q

The modal overlap factor (MOF) is the ratio of modal bandwidth to average mode spacing, which is the product of mode density and modal bandwidth:

MOF = (dN/df)Δf = 8πVf³/(c₀³Q)

When the modal overlap factor is well above one, modes overlap strongly and the chamber behaves as a smoothly varying random medium rather than as a set of resolvable resonances. When it falls to one or below, individual modes can be resolved, and the field statistics depart from the ideal model.

The factor rises with the cube of frequency, so a chamber crosses this threshold somewhere inside its working range. Take an empty chamber of 15 cubic meters operating at 1 GHz with a quality factor of 20,000. The mode density is about 14 modes per megahertz, giving a mean mode spacing near 72 kHz, while the modal bandwidth is 50 kHz. The modal overlap factor is therefore about 0.7. Many chambers work for part of their range with overlap near or below unity and rely on stirring, rather than on overlap alone, to deliver the required statistics.

Lowest Usable Frequency

The lowest usable frequency is arguably the most important parameter in reverberation chamber specifications. Several methods exist for determining it:

• Mode counting method: requires a minimum number of modes, commonly taken as sixty to one hundred, below the operating frequency
• First mode multiple: places the LUF at three to six times the frequency of the first resonant mode
• Statistical uniformity criterion: takes the LUF as the frequency above which measured field uniformity meets the required tolerance
• Modal overlap criterion: takes the LUF as the frequency above which the modal bandwidth spans more than one mode spacing, so that several modes are excited together

Consider a chamber measuring 3 meters by 2.5 meters by 2 meters. Its lowest resonance, set by the two largest dimensions, falls near 78 MHz, so the three-times rule places the LUF near 234 MHz. In practice the LUF is confirmed empirically through field uniformity measurements, because the mode-counting and multiple-of-resonance rules take no account of stirrer effectiveness or chamber loading.

Ideal Chamber Assumptions

The statistical theory of reverberation chambers rests on several idealized assumptions:

• Uniform mode excitation: all modes within the measurement bandwidth are excited equally
• Random phase: the phases of individual mode contributions are uniformly distributed and statistically independent
• Sufficient stirring: the stirrer creates enough statistically independent field configurations
• Negligible direct coupling: energy transfer between transmit and receive antennas occurs only through the reverberant field

When these conditions are satisfied, the field at any location is the sum of many random contributions, and the central limit theorem fixes the distributions that follow.

Rectangular Component Distribution

Write the electric field as three complex rectangular components E_x, E_y, and E_z. Under ideal conditions the real part and the imaginary part of each component are independent, zero-mean Gaussian variables with a common variance σ²:

p(E_xr) = (1/√(2πσ²))exp(−E_xr²/(2σ²))

The same density applies to the imaginary part E_xi and to both parts of the other two components. That one variance is common to all six quantities is a consequence of isotropy in an ideal chamber. The magnetic-field components H_x, H_y, and H_z follow the same form with a variance smaller by the square of the free-space wave impedance, so that the electric and magnetic energy densities are equal.

Field Magnitude Distributions

Because each component combines an independent in-phase and quadrature Gaussian part, the magnitude of a single component is Rayleigh distributed, equivalently chi distributed with two degrees of freedom:

p(|E_x|) = (|E_x|/σ²)exp(−|E_x|²/(2σ²))

The total electric field magnitude combines all six Gaussian quadratures:

|E| = √(|E_x|² + |E_y|² + |E_z|²)

It therefore follows a chi distribution with six degrees of freedom. The squared magnitude of a single component, which is proportional to the power in that component, is exponentially distributed, while the squared magnitude of the total field is chi-square distributed with six degrees of freedom. Measurements reported by NIST in the 1990s confirmed both densities in mechanically stirred chambers.

These distributions have practical implications:

• The average power in the total field is three times that carried by any one rectangular component
• Peak-to-average ratios follow from the distributions, which allows test margins to be set
• Confidence intervals can be established for field level estimates
• The exponential density has a long tail, so an individual stirrer position can deliver a field far above the average

Statistical Uniformity

Spatial uniformity in a reverberation chamber is described statistically rather than deterministically. For any single stirrer position, the field varies strongly with location. Averaged over many stirrer positions, however, the statistical properties of the field become uniform throughout the working volume.

Field uniformity is characterized by the standard deviation, in decibels, of the maximum field values recorded at several locations. IEC 61000-4-21 sets a frequency-dependent tolerance: the standard deviation may reach 4 dB from 80 MHz to 100 MHz, must fall linearly from 4 dB to 3 dB between 100 MHz and 400 MHz, and must not exceed 3 dB above 400 MHz. The wider allowance at the bottom of the range reflects the sparser, less overlapped modes found there.

The working volume is the region in which the uniformity requirement is met. Practice keeps its boundary at least a quarter wavelength, evaluated at the lowest usable frequency, away from the walls, the stirrer, and the antennas. For a chamber that starts near 80 MHz that separation is roughly 0.9 meter, which is why published working volumes sit well inside the physical shell.

Isotropy

Isotropy means the statistical field properties are independent of direction. In an ideal chamber:

• The average power in each polarization is equal
• The field arrives equally from all directions when averaged over stirrer positions
• Equipment orientation does not affect the average received power

This is a significant advantage over anechoic testing, where equipment must be rotated and the antenna repositioned to cover all angles of incidence. In a reverberation chamber, isotropy ensures thorough exposure regardless of equipment orientation.

Definition and Physical Meaning

The quality factor is defined as:

Q = 2π × (energy stored in the chamber)/(energy dissipated per cycle)

Equivalently, in terms of the angular frequency ω:

Q = ωU/P_d = 2πfU/P_d

Here U is the total stored electromagnetic energy and P_d is the dissipated power, which in steady state equals the power delivered to the chamber. A high Q means energy accumulates efficiently, producing higher field levels for a given input power.

Loss Mechanisms

Energy leaves a reverberation chamber through four principal mechanisms.

Wall losses: the metallic walls have finite conductivity, so the surface currents they carry dissipate energy. For nonmagnetic walls the corresponding quality factor is:

Q_wall = 3V/(2Sδ)

Here V is the chamber volume, S is the total interior surface area, and δ is the skin depth in the wall material. Because skin depth falls as the inverse square root of frequency, the wall Q rises roughly as the square root of frequency, and it also rises with chamber size through the favorable volume-to-surface ratio. Wall loss usually dominates in an empty chamber.

Aperture leakage: energy escapes through imperfect door seals, cable penetrations, and ventilation openings. Apertures are characterized by an average transmission cross section σ_t:

Q_aperture = 4πV/(λσ_t)

The transmission cross section, not the geometric opening area, governs the loss, and a resonant slot can present a cross section far larger than its physical size. Careful shielding of every penetration is therefore essential.

Antenna losses: the receiving antennas draw energy from the chamber into their terminating loads. For an antenna of efficiency η, including mismatch:

Q_antenna = 16π²V/(ηλ³)

This term falls as the cube of wavelength, so antenna loading dominates at low frequencies, where the wavelength is a substantial fraction of the chamber dimensions. Note that η here is an efficiency, distinct from the wave impedance η₀ used below.

Object loading: the equipment under test and any absorbing material inside the chamber take energy out of the field. This is frequently the dominant loss during testing and is treated in detail under loading effects below.

Composite Q Factor

Because the loss powers add, their reciprocal quality factors add:

1/Q_total = 1/Q_wall + 1/Q_aperture + 1/Q_antenna + 1/Q_load

The smallest individual term dominates the composite value. Since the wall and antenna terms carry opposite frequency dependences, the composite Q of an empty chamber typically rises with frequency at the low end, where antenna loading limits it, and then flattens as wall loss takes over.

Reported empty-chamber values for metal-walled chambers commonly span roughly 10³ to 10⁵, rising with chamber size and frequency; the loaded value during a test is lower, often by an order of magnitude. Because Q enters the field-strength calculation directly, measuring it is a routine diagnostic of chamber health.

Field Level and Power Relationship

The average power density in the chamber follows from the stored energy and the quality factor:

S_avg = QP_inλ/(2πV)

The root-mean-square total electric field is the corresponding field quantity:

E_rms = √(QP_inη₀λ/(2πV))

Here η₀ is the free-space wave impedance, approximately 377 ohms. A higher Q and a smaller volume both raise the field obtained from a given input power. The two effects work against each other in practice, because Q generally improves with chamber size, so a large chamber does not lose as much field strength as the volume term alone would suggest. The explicit wavelength factor also means that, at fixed Q, the achievable field falls as frequency rises.

Time Constant and Settling Time

Energy in the chamber decays exponentially when the source is switched off, with a time constant set by the quality factor:

τ = Q/ω = Q/(2πf)

The same time constant governs the buildup when a source is switched on. For swept-frequency or stepped-frequency measurements, the dwell time at each frequency must be several time constants so that the field reaches steady state before a reading is taken.

At 1 GHz with Q = 10,000, the time constant is about 1.6 microseconds. For pulsed immunity testing, the pulse must be long compared with this figure if the field is to reach its full level. The relationship also runs the other way: fitting an exponential to the measured decay of received power yields Q = ωτ without any knowledge of the chamber volume or the antenna efficiency, which makes the time-domain method an attractive alternative to the power-ratio method.

Purpose of Stirring

Stirring serves several critical functions:

• Mode mixing: changes the relative amplitudes and phases of the excited modes
• Statistical sampling: creates independent field configurations for averaging
• Uniformity improvement: reduces spatial variation when averaged over stirrer positions
• Polarization randomization: ensures all polarizations are represented in the averaged field

Without stirring, the field would exhibit fixed standing-wave patterns with nulls and maxima at specific locations, which would make the chamber unsuitable for standardized testing.

Mechanical Stirring

Mechanical stirring uses rotating metal paddles or tuners to change the boundary conditions. As the stirrer rotates, it scatters the field differently at each angle, producing a new field configuration at every position.

Effective mechanical stirrers share several characteristics:

• Size: the stirrer should be electrically large, spanning at least one wavelength at the lowest frequency of operation
• Asymmetry: non-symmetric designs mix modes better than symmetric paddles, which can repeat a boundary condition partway through a rotation
• Multiple axes: stirrers rotating about different axes improve uniformity
• Continuous variation: the boundary conditions should change throughout the rotation rather than in a few abrupt steps

Stirrer designs include flat paddles, Z-fold configurations, irregular asymmetric shapes, and corrugated surfaces. Large chambers often use more than one stirrer, driven at unrelated rates so that their combined effect does not repeat.

Frequency Stirring

Frequency stirring, also called electronic stirring, varies the excitation frequency across a band to create effectively different mode configurations. Because each mode has its own resonant frequency, a change in excitation frequency changes which modes are strongly excited and how their contributions add.

The stirring bandwidth must be wide enough to span several mode spacings yet narrow enough that the response of the device under test stays essentially constant across it. In practice this places the bandwidth at a fraction of a percent to a few percent of the center frequency, and the acceptable width narrows for a device with a sharp resonance.

Frequency stirring combines with mechanical stirring to increase the number of independent samples. The combination is most valuable at the low end of the range, where mode density is limited and mechanical stirring alone yields few independent samples.

Source Stirring

Source stirring uses several transmit antennas at different locations, exciting the chamber from each in turn or from several at once with controlled phase relationships. Each source position creates a different mode distribution and so provides additional independent samples.

Source position stirring works because the coupling from the source into each mode depends strongly on where the source sits in the modal field pattern. Moving the source by a fraction of a wavelength can therefore change the excited configuration substantially, which also makes source placement a repeatability concern when it is not being stirred deliberately.

Number of Independent Samples

The number of independent samples available from stirring determines the measurement uncertainty. Independence is judged from the autocorrelation of the measured quantity as a function of stirrer angle θ. Throughout this article, mean[ ] denotes an average taken over the stirring sequence:

ρ(Δθ) = mean[(x(θ) − x_avg)(x(θ + Δθ) − x_avg)]/mean[(x(θ) − x_avg)²]

Here x(θ) is the sampled quantity, usually the received power or the complex field, and x_avg is its mean over a full rotation. Samples are treated as independent once the correlation falls below a threshold, commonly taken as 1/e, or about 0.37. The number of independent samples N_ind then sets the standard error of any averaged quantity:

σ_estimate = σ_true/√N_ind

The count of independent samples is generally far smaller than the number of stirrer positions actually sampled, particularly near the LUF, where adjacent positions remain strongly correlated. It improves with frequency and with stirrer size and asymmetry, and it approaches the number of positions only at the high end of a chamber's range. Sampling more positions than the stirrer can decorrelate adds measurement time without reducing uncertainty.

Absorption Cross Section

The absorption cross section quantifies how much power an object takes from the reverberant field:

P_abs = σ_aS_avg

Here σ_a is the absorption cross section in square meters and S_avg is the average power density. The cross section depends on the object's size, shape, material properties, and frequency, and it is averaged over all angles of incidence and both polarizations, which is precisely the average the chamber itself performs.

The cross section is measured by comparing the chamber Q with and without the object:

σ_a = (2πV/λ)(1/Q_loaded − 1/Q_empty)

The quantity is useful for characterizing test objects, for predicting their effect on chamber performance, and for deliberately setting chamber Q with calibrated absorber when a test calls for a shorter time constant.

Loading Factor

The loading factor compares the loaded and empty quality factors:

LF = Q_empty/Q_loaded

A loading factor of one indicates negligible loading. A factor of two or more indicates substantial absorption that must be accounted for when field levels are calculated, since the field for a given input power falls as the square root of Q.

Standards require the loading to be verified and bounded so that the reference calibration stays valid and the chamber stays overmoded. The bound is expressed through insertion loss, which measures Q indirectly: a loading factor of 1.4, for instance, corresponds to about 1.5 dB of additional insertion loss relative to the reference condition. Loading beyond the permitted bound calls for recharacterizing the chamber with the load in place rather than correcting the earlier calibration.

Effects on Field Statistics

Moderate loading lowers the field level without disturbing the field statistics or the uniformity. Heavy loading, however, can:

• Reduce Q below the value needed to keep the chamber overmoded
• Create local field perturbations in the shadow of the absorbing object
• Shorten the correlation distance in stirrer angle and change the number of independent samples
• Introduce direct coupling paths between antennas as the reverberant field weakens

Large equipment under test must therefore be positioned with care, keeping it inside the working volume while leaving the stirrer and the transmit antenna unobstructed. Heavy loading also shortens the chamber time constant, which is beneficial for pulsed testing even as it lowers the achievable field.

Loading Compensation

When loading lowers Q, the input power must rise to restore the desired field level:

P_loaded = P_empty(Q_empty/Q_loaded)

Some test standards require the chamber loss to be measured at each test frequency with the equipment under test in place, and the drive power to be adjusted accordingly. This keeps field exposure consistent regardless of how heavily a particular setup loads the chamber.

Calibration Objectives

Chamber calibration serves to:

• Determine the power required to achieve specified field levels
• Verify statistical uniformity throughout the working volume
• Characterize the lowest usable frequency
• Assess stirrer effectiveness and the number of independent samples
• Establish measurement uncertainty budgets

Field Uniformity Verification

Field uniformity is verified by measuring the field or the received power at several locations with a calibrated probe or reference antenna. The IEC 61000-4-21 procedure specifies:

• Measurements at the eight corners of a volume enclosing the working area
• Three orthogonal probe or antenna orientations at each location
• A full stirrer rotation sampled at a fixed set of positions for each configuration
• Recording of the maximum value observed over the stirrer cycle for each configuration

More stirrer positions are sampled at lower frequencies, where independent samples are scarce, and fewer at higher frequencies, where mode density is high. The standard deviation of the resulting twenty-four maxima, from eight locations and three orientations, characterizes the uniformity and is judged against the frequency-dependent tolerance described above. Because the calibration also records the received power referenced to the input power, the same data set yields both the uniformity result and the chamber Q.

Chamber Calibration Factor

The chamber calibration factor (CCF) relates input power to the mean-square electric field:

CCF = mean[|E|²]/P_in

It carries units of volts squared per meter squared per watt. The factor is obtained by measuring received power with a reference antenna of known efficiency and applying the statistical relationships given above. Once known, it tells the test engineer what forward power will produce a specified field, without a probe having to remain in the chamber during the test.

Reference Antenna Requirements

Calibration antennas should have:

• Known, stable efficiency across the frequency range
• An omnidirectional or well-characterized directional pattern
• A good impedance match, to limit mismatch uncertainty
• A size appropriate to the working volume

Typical reference antennas include log-periodic antennas, horn antennas, and broadband probes designed for reverberation work. Antenna factors and efficiencies must be known and traceable to national standards, since they enter the calibration factor directly.

Spatial Correlation

The spatial correlation function describes how the fields at two points separated by a distance d are related:

ρ(d) = mean[E(r) · E^*(r + d)]/mean[|E(r)|²]

The superscript asterisk denotes the complex conjugate. In an ideal reverberation chamber this correlation evaluates to a sinc function:

ρ(d) = sin(kd)/(kd)

Here k = 2π/λ is the wave number. The correlation first reaches zero at d = λ/2, a result derived independently from plane-wave integral theory, cavity mode theory, and radiative transfer theory, and confirmed by measurement. Measurement points should therefore be separated by at least half a wavelength if they are to be treated as independent, and the same length scale explains why an extended test object averages over the field rather than sampling it at a point.

Frequency Correlation

Field values at nearby frequencies are correlated as well. The magnitude of the frequency correlation falls off approximately as:

ρ(Δf) ≈ 1/√(1 + (Δf/Δf_c)²)

The correlation bandwidth Δf_c is of the order of the modal bandwidth f/Q, which is the same quantity that sets modal overlap. Frequencies separated by more than a few correlation bandwidths give essentially independent field values.

This behavior is what frequency stirring exploits: sweeping across a band that spans many correlation bandwidths yields samples that can be averaged like independent stirrer positions. It also sets a floor on how finely a swept measurement needs to be stepped, because samples closer together than Δf_c add little new information.

Stirrer Position Correlation

The correlation between field values at different stirrer positions depends on stirrer design and on frequency. Well-designed stirrers produce low correlation between adjacent positions, which maximizes the number of independent samples per rotation.

Given the measured autocorrelation, the effective number of independent samples in a sequence of N positions follows a Bartlett-type estimate:

N_ind ≈ N/(1 + 2∑_k=1^N−1(1 − k/N)ρ(k))

Here ρ(k) is the autocorrelation of the averaged quantity at a lag of k stirrer steps. For a complex Gaussian field the power autocorrelation equals the squared magnitude of the field autocorrelation, so the two descriptions are consistent. In practice the autocorrelation is computed directly from recorded power values against stirrer position, which makes N_ind a measured property of the chamber and stirrer rather than an assumed one.

Implications for Measurement Uncertainty

Averaged power in an ideal chamber is the mean of exponentially distributed samples, so its relative standard error falls with the square root of the number of independent samples:

u(P)/P = 1/√N

Because power is proportional to the square of field amplitude, the corresponding relative uncertainty in amplitude is about half as large:

u(E)/E ≈ 1/(2√N)

One hundred independent samples therefore leave about 10 percent relative uncertainty in mean power, and reducing that to 1 percent would demand ten thousand. Practical measurements use on the order of fifty to two hundred independent samples and accept uncertainties of roughly 7 to 14 percent in power.

Combining mechanical stirring, frequency stirring, and source position stirring raises the number of independent samples without a proportional increase in test time, which is the main lever available for trading measurement time against uncertainty.