From a voltage to a wave and back
A phone transmitting at a fraction of a watt is heard by a mast several kilometres away. A satellite 38,000 km above the equator delivers a hundred television channels to a dish the size of a bicycle wheel. NASA's Deep Space Network still receives data from the two Voyager probes, launched in 1977 and now more than 20 billion kilometres away, each transmitting with about 20 watts, the power of a refrigerator's light bulb. In every case, a few watts in a conductor become an electromagnetic wave, the wave spreads over an enormous sphere, and a receiver catches a tiny part of it and separates it from the background hiss of thermal noise.
This lesson follows that chain: the antenna, which converts between guided waves on a transmission line and free waves in space; the link budget, which says how much power arrives; noise, which says how much power is needed; the receiver, which brings the signal down to where it can be digitised; and arrays, which shape and steer beams electronically, as the base stations and satellite terminals of 2026 do.
How antennas radiate
A charge at rest has an electric field; a charge moving steadily drags its field along. Only when a charge accelerates does a kink appear in its field lines, and that kink travels outwards at the speed of light: radiation. An alternating current is charge accelerating back and forth, so any conductor carrying a high-frequency current radiates a little. An antenna is a conductor shaped so that it radiates a lot.
Picture a two-wire transmission line ending in an open circuit. The currents in the two wires flow in opposite directions, so their fields almost cancel at a distance, and the energy is reflected back. Now bend the last quarter-wavelength of each wire outwards until they form a straight line. The currents in the two arms now flow in the same direction, their fields add instead of cancelling, and the energy leaves as a wave. That is a half-wave dipole, the reference antenna of radio engineering.
Close to an antenna the fields are complicated: there is a reactive near field, within about , where energy sloshes back and forth between antenna and space without leaving (this is the field an NFC card or a wireless charger uses). Far away, in the far field, the wave has become a spherical wave that looks locally like the plane wave of the previous lessons: and perpendicular, in phase, , and both falling as , so the power density falls as . For an antenna much larger than a wavelength, such as a dish of diameter , the far field starts at about (28 m for a 60 cm satellite dish at 11.7 GHz). All the formulas below are far-field formulas.
Pattern, directivity and gain
Definition— radiation pattern and directivity
The radiation intensity is the power radiated per unit solid angle in a given direction. An isotropic antenna, which exists only as a reference, would radiate equally in all directions, . The directivity of a real antenna is the ratio of its peak intensity to the isotropic value: . It is usually quoted in dBi, decibels relative to isotropic.
Definition— gain, efficiency and EIRP
The gain includes the antenna's radiation efficiency , the fraction of the power delivered to it that is radiated rather than lost as heat. The effective isotropic radiated power is the power an isotropic antenna would need to produce the same intensity in the direction of the beam.
Definition— effective aperture
As a receiver, an antenna with gain collects power as if it were a perfect absorbing area facing the wave. For a dish of physical area , is typically 55–70 % of .
The half-power beamwidth is the angle between the directions in which the intensity has fallen to half its peak (−3 dB). Directivity and beamwidth trade against each other: concentrating power into a narrow beam is what makes an antenna directive. A rule of thumb for a dish is a beamwidth of about degrees.
Analogy, not a formal construction
An antenna's gain is like the reflector of a torch. The bulb does not get brighter when you add a reflector; the same light is gathered into a beam, so the spot in front is brighter and everywhere else is darker. A 30 dBi dish makes a spot a thousand times brighter than a bare bulb would, over a correspondingly small patch of sky.
The half-wave dipole, computed
The current on a thin half-wave dipole is very nearly a half-cycle of a sine wave, largest at the feed and zero at the tips. Adding up the fields of all its small current elements in the far field gives a radiation intensity proportional to
where is measured from the antenna's axis. A very short dipole (much shorter than ) has the simpler pattern . The program integrates each pattern over the whole sphere to get the directivity, finds the half-power beamwidth, computes the half-wave dipole's radiation resistance (the resistance that would dissipate the radiated power, ), and then looks at short antennas.
import numpy as np
eta0 = 376.73
theta = np.linspace(1e-6, np.pi - 1e-6, 200001) # angle from the antenna's axis
dtheta = theta[1] - theta[0]
def analyse(name, F):
"""F is the radiation intensity pattern (power per steradian, up to a constant)."""
total = 2 * np.pi * np.sum(F * np.sin(theta)) * dtheta # integrate over the sphere
D = 4 * np.pi * F.max() / total # directivity: peak / average
half = theta[F >= F.max() / 2] # half-power beamwidth
hpbw = np.degrees(half[-1] - half[0])
print(f"{name:20} directivity {D:5.3f} = {10 * np.log10(D):4.2f} dBi, beamwidth {hpbw:5.1f} deg")
return total
analyse("short dipole", np.sin(theta) ** 2)
F = (np.cos(np.pi / 2 * np.cos(theta)) / np.sin(theta)) ** 2
total = analyse("half-wave dipole", F)
# Radiated power with a peak current I0 is (eta0 I0^2 / 8 pi^2) * total, so R = 2P / I0^2:
print(f"half-wave dipole radiation resistance: {eta0 * total / (4 * np.pi**2):.1f} ohm")
print("\nangle from axis relative field (half-wave dipole)")
for deg in [0, 15, 30, 45, 60, 75, 90]:
t = np.radians(max(deg, 1e-6))
field = np.cos(np.pi / 2 * np.cos(t)) / np.sin(t)
print(f"{deg:9d} deg {field:5.3f} " + "#" * int(round(40 * field)))
# A short centre-fed dipole (current falling linearly to zero at the tips) has
# R_rad = 20 pi^2 (l / lambda)^2. With 0.5 ohm of loss in the wire and matching coil:
for length, f in [(0.10, 100e6), (0.03, 433e6), (1.0, 1e6)]:
lam = 299_792_458 / f
r_rad = 20 * np.pi**2 * (length / lam) ** 2
print(f"{length * 100:5.0f} cm dipole at {f / 1e6:4.0f} MHz: l/lambda = {length / lam:.4f}, "
f"R_rad = {r_rad:.3f} ohm, efficiency = {100 * r_rad / (r_rad + 0.5):4.1f} %")The half-wave dipole has a directivity of 1.64 (2.15 dBi), a beamwidth of 78°, and a radiation resistance of 73 Ω, which is why a dipole fed directly from a 75 Ω cable, or from 50 Ω with a small mismatch, works well. Engineers quote antenna gains relative to a dipole too, in dBd: 0 dBd = 2.15 dBi.
The short dipoles show the real difficulty of small antennas. Radiation resistance falls as the square of the length in wavelengths, so a 10 cm whip at 100 MHz radiates through a fraction of an ohm, comparable with the resistance of its own wire and matching coil, and only about 30 % of the power gets out. A 1 m wire at 1 MHz is hopeless as a transmitter, which is why medium-wave broadcasting stations use masts a hundred metres or more tall.
Practical antennas are variations on these themes. A monopole over a ground plane (a car's whip, a mast on a roof) is half a dipole with its mirror image in the ground, with half the radiation resistance (about 36 Ω). A patch antenna, a square of copper on a circuit board, radiates from its edges and has a gain of 6–8 dBi. Phone antennas are folded, meandered, multi-band shapes designed by simulation, sharing a tiny volume with metal, battery and hand. A Yagi–Uda antenna (invented by Shintaro Uda and Hidetsugu Yagi at Tohoku University in the 1920s) adds parasitic rods in front of and behind a dipole to make a beam, and still sits on rooftops for television reception. A parabolic dish focuses a plane wave onto a small feed, and its gain grows as .
The Friis equation
How much of the transmitted power reaches the receiver? Harald Friis of Bell Laboratories put the answer into one line in 1946.
Theorem— Friis transmission equation
In free space, with transmitter and receiver in each other's far field, polarisations matched, and both antennas matched to their lines, the received power is
In decibels: , where the free-space path loss is dB.
Proof
An isotropic transmitter spreads evenly over a sphere of area ; a transmitting antenna with gain multiplies the power density in its beam by , giving . The receiving antenna captures the power falling on its effective aperture . Multiplying, .
∎
Two features are worth noticing. The received power falls as , so doubling the distance costs only 6 dB; radio links span enormous ranges because their loss grows only logarithmically in decibels. And the frequency enters only through , and only because we fixed the antennas' gains. A wave does not "weaken" faster at high frequency in free space: an antenna of a given gain simply has a smaller capture area at shorter wavelengths. (Real links do lose more at high frequency to rain, foliage, walls and the oxygen and water vapour of the atmosphere, which the ITU's propagation models describe.)
Noise: kTB and the noise figure
A received power means nothing until it is compared with noise. The electrons in any resistor at temperature jiggle randomly and produce a noise voltage (Johnson–Nyquist noise), and the noise power a matched receiver picks up in bandwidth is
with Boltzmann's constant J/K. At the standard reference temperature K this is dBm per hertz of bandwidth: dBm in a 20 MHz channel, dBm in 80 MHz once a typical receiver's own noise is added. A wider channel carries more data but lets in more noise; that trade is the subject of information theory.
Every amplifier, mixer and lossy cable adds noise of its own. The noise factor of a stage is the ratio of the signal-to-noise ratio at its input to that at its output, when the input noise is ; in decibels it is the noise figure NF. A perfect stage has (0 dB). Equivalently, a stage behaves like a noiseless one with an extra noise temperature at its input. A passive loss of dB at room temperature has a noise figure of dB.
Theorem— Friis formula for noise
A chain of stages with noise factors and power gains has total noise factor
Each stage's noise is divided by all the gain in front of it, so the first stage dominates: put a high-gain, low-noise amplifier (LNA) right at the antenna and the rest of the chain hardly matters. The program compares three orderings of the same parts.
import math
def cascade(stages, T0=290.0):
"""Friis's formula for noise: stages are (name, gain_dB, noise_figure_dB), antenna first."""
F_total, G_before = 0.0, 1.0
for i, (name, g_db, nf_db) in enumerate(stages):
F = 10 ** (nf_db / 10)
F_total += F if i == 0 else (F - 1) / G_before
G_before *= 10 ** (g_db / 10)
return 10 * math.log10(F_total), T0 * (F_total - 1)
lna = ("low-noise amplifier", 18.0, 1.0)
cable = ("cable and filter", -3.0, 3.0) # a passive loss of L dB has a noise figure of L dB
mixer = ("mixer", -7.0, 8.0)
if_amp = ("IF amplifier", 30.0, 4.0)
for label, chain in [("LNA at the antenna", [lna, cable, mixer, if_amp]),
("cable first, then LNA", [cable, lna, mixer, if_amp]),
("no LNA at all", [cable, mixer, if_amp])]:
nf, te = cascade(chain)
sensitivity = -174 + 10 * math.log10(20e6) + nf + 10 # 20 MHz, 10 dB SNR needed
print(f"{label:24} NF = {nf:5.2f} dB, Te = {te:6.0f} K, sensitivity = {sensitivity:6.1f} dBm")With the LNA first, the whole chain has a noise figure of 2.3 dB, only 1.3 dB worse than the LNA alone. Put 3 dB of cable in front of the LNA and the cable's loss adds directly to the noise figure, giving 4.7 dB; leave the LNA out and it is 14.4 dB. The last column turns this into sensitivity, the weakest signal the receiver can use (here: noise in 20 MHz plus the noise figure plus a 10 dB signal-to-noise ratio needed by the demodulator): the good design hears signals 12 dB weaker than the one without an LNA, which in free space is four times the range. This is why satellite dishes carry their LNA at the focus, in the "LNB" box on the arm, rather than at the end of the cable indoors.
Link budgets
A link budget is a table that adds up, in decibels, everything between the transmitter's output and the receiver's detector: powers and gains as positive entries, losses as negative ones. The result is compared with the noise to give the signal-to-noise ratio, and that is compared with what the modulation needs to give a margin. Here are two, with typical (illustrative, not vendor-specific) values: a Wi-Fi link through two walls in a home, and a direct-to-home television downlink from a geostationary satellite.
import math
c, k = 299_792_458.0, 1.380649e-23
def db(x):
return 10 * math.log10(x)
def fspl_db(d, f):
"""Free-space path loss 20 log10(4 pi d / lambda)."""
return 20 * math.log10(4 * math.pi * d * f / c)
def dish_gain_dbi(diameter, f, efficiency=0.65):
return db(efficiency * (math.pi * diameter * f / c) ** 2)
def budget(title, rows, noise_dbm):
print(title)
total = 0.0
for i, (name, value) in enumerate(rows):
total += value
print(f" {name:34}{value:+8.1f} {'dBm' if i == 0 else 'dB'}")
print(f" {'received power':34}{total:8.1f} dBm")
print(f" {'noise power':34}{noise_dbm:8.1f} dBm")
print(f" {'signal-to-noise ratio':34}{total - noise_dbm:8.1f} dB\n")
# 1. Wi-Fi at 5.5 GHz, 30 m away through two interior walls, in an 80 MHz channel
f, d, B, nf = 5.5e9, 30.0, 80e6, 6.0
noise = db(k * 290 * B / 1e-3) + nf # kTB in dBm, plus the noise figure
budget("Wi-Fi link, 5.5 GHz, 30 m", [
("transmit power (100 mW)", 20.0),
("access-point antenna gain", 3.0),
("free-space path loss", -fspl_db(d, f)),
("two interior walls (assumed)", -2 * 6.0),
("laptop antenna gain", 2.0),
], noise)
# 2. Satellite TV from geostationary orbit at 11.7 GHz into a 60 cm dish
f, d, B, T_sys = 11.7e9, 38_000e3, 27e6, 120.0
noise = db(k * T_sys * B / 1e-3)
budget("Geostationary TV downlink, 11.7 GHz", [
("satellite EIRP (52 dBW = 82 dBm)", 82.0),
("free-space path loss, 38,000 km", -fspl_db(d, f)),
("atmosphere, clear sky (assumed)", -0.5),
("60 cm dish gain", dish_gain_dbi(0.60, f)),
], noise)
print(f"path loss to a satellite 550 km overhead instead: {fspl_db(550e3, 11.7e9):.1f} dB "
f"({fspl_db(38_000e3, 11.7e9) - fspl_db(550e3, 11.7e9):.1f} dB less)")The Wi-Fi link loses 77 dB to free space in 30 m and another 12 dB in the (assumed) walls, yet arrives with an SNR of 25 dB, enough for a dense modulation such as 64-QAM or 256-QAM (see modulation). The satellite link loses 205 dB, a factor of , and still arrives with 15 dB to spare, because the satellite concentrates its power into a beam aimed at one region (an EIRP of 52 dBW is what 160 kW would give from an isotropic antenna) and the 60 cm dish adds 35 dB. The satellite's noise budget is written with a system noise temperature of 120 K rather than a noise figure, because the cold sky behind the satellite is much colder than 290 K; satellite engineers summarise a receiving station by its G/T in dB/K.
Receiver architectures
The received signal may be at 2.4 GHz or 11.7 GHz and as weak as W. It must be filtered from neighbouring signals that may be a million times stronger, amplified, and brought down to a frequency where it can be demodulated or digitised. Three architectures do this.
The superheterodyne. Edwin Armstrong's 1918 invention mixes the incoming signal at with a local oscillator at ; the mixer (a multiplier) produces the sum and difference frequencies, and the difference is kept. Tuning the receiver means tuning only the LO, while the selective filters and the gain sit at one fixed IF (10.7 MHz in FM radios, for example). The catch is the image frequency: a signal at (if the LO is above the wanted signal) produces the same IF and cannot be separated after the mixer, so the RF filter in front must remove it. Superhets dominated radio for most of the twentieth century and still appear in demanding receivers, often with two or three IFs.
Direct conversion (zero-IF). Choose , so the IF is zero. A single mixer would fold the upper and lower halves of the signal's spectrum on top of each other, so the receiver uses two mixers with LOs 90° apart and keeps both outputs, the I and Q components: together they are the real and imaginary parts of the complex envelope of phasors, which preserves everything. There is no image to filter and the channel filters are simple low-pass filters that can be integrated on a chip, which is why almost every phone and Wi-Fi radio is direct-conversion (or "low-IF", a close relative). The price is sensitivity to DC offsets, LO leakage and I/Q imbalance, which are corrected digitally.
Software-defined radio (SDR). Move the analog-to-digital converter as close to the antenna as possible and do the rest in software. Low-cost SDR receivers digitise a few megahertz of I/Q samples after a direct-conversion or low-IF front end; base stations and test equipment use RF-sampling converters that digitise gigahertz of spectrum directly, as sampling and the FFT allows. Filtering, tuning and demodulation are then FFTs and digital filters, and a new standard can be supported by new software. Free toolkits such as GNU Radio let anyone build a receiver this way.
Note
Listening with an inexpensive SDR receiver and a home-made wire dipole is safe and is a superb way to learn this subject, though some countries restrict listening to certain services, so check the rules where you live. Transmitting is different: outside licence-exempt devices such as Wi-Fi routers and toy walkie-talkies, it needs a licence, because a careless transmitter can block emergency and aviation radio. Amateur radio licences, available in most countries through national societies (see the IARU), are the legal way to experiment with transmitters. Never put up an antenna near overhead power lines, and keep away from the antennas of broadcast and mobile-network transmitters, which are fenced for good reason.
Arrays and beam steering
A dish points its beam by moving. An array of small antennas can point it electronically, in microseconds, and form several beams at once. The idea is interference: feed identical elements spaced apart along a line, with element given a phase shift . In a direction at angle from broadside, the wave from element has travelled an extra , a phase of with . The total field is the element's own pattern times the array factor
All terms line up, and reaches its maximum , in the direction where . So choosing steers the beam to . The sum is a geometric series, which gives with : the same function as the frequency response of a moving-average filter in digital filters, because an array samples the wave in space just as a filter samples a signal in time.
import numpy as np
def array_factor(theta_deg, n, d_over_lambda, steer_deg):
"""|AF| / n for n isotropic elements spaced d apart along a line, phased to point at steer_deg
(angles measured from broadside, i.e. from the direction perpendicular to the line)."""
theta = np.radians(theta_deg)[:, None]
k_d = 2 * np.pi * d_over_lambda
phase = -k_d * np.sin(np.radians(steer_deg)) # progressive phase shift per element
m = np.arange(n)[None, :]
return np.abs(np.exp(1j * m * (k_d * np.sin(theta) + phase)).sum(axis=1)) / n
angles = np.arange(-90, 91, 0.1)
for n, d, steer in [(8, 0.5, 0), (8, 0.5, 30), (16, 0.5, 30), (8, 1.0, 30)]:
af = array_factor(angles, n, d, steer)
main = angles[(af >= 1 / np.sqrt(2)) & (np.abs(angles - steer) < 20)] # half-power points
lobes = [round(float(a)) + 0 for i, a in enumerate(angles[1:-1], 1)
if af[i] > af[i - 1] and af[i] >= af[i + 1] and af[i] > 0.9]
print(f"N = {n:2d}, d = {d} lambda, steered to {steer:2d} deg: half-power beamwidth "
f"{main[-1] - main[0]:4.1f} deg, full-strength beams at {lobes} deg")
print("\nN = 8, d = lambda/2, steered to 30 deg (|AF| in dB, 10 deg steps)")
for a in range(-90, 91, 10):
v = array_factor(np.array([a]), 8, 0.5, 30)[0]
level = max(20 * np.log10(max(v, 1e-6)), -30)
print(f"{a:4d} deg {level:6.1f} dB |" + "#" * int(round(level + 30)))Eight elements half a wavelength apart give a beam about 13° wide; sixteen halve it, since the array is twice as long. Steering to 30° widens the beam slightly, to about , because the array looks shorter seen from an angle. Between the nulls are sidelobes, the highest about 13 dB down, which designers lower by tapering (feeding the edge elements less strongly) at the price of a wider main beam. The last case is a warning: with elements a full wavelength apart, steering to +30° also produces a full-strength grating lobe at −30°, sending half the power the wrong way. To steer anywhere without grating lobes, elements must be at most half a wavelength apart.
Phased arrays are no longer exotic. A 5G base station's "massive MIMO" panel holds 32 to 64 transceivers behind hundreds of elements and forms separate beams towards different users; low-Earth-orbit satellite user terminals are flat phased arrays tracking satellites across the sky; and radars, from weather radars to car radars at 77 GHz, steer electronically. Modern arrays often apply phases and amplitudes digitally, one transceiver per element or group, which is digital beamforming, and the MIMO capacity it unlocks is computed in wireless systems.
The professional view
- Design and measurement. Antennas are designed with electromagnetic simulators (FDTD, finite elements and the method of moments from the first lesson of this level; the public-domain NEC-2 code from Lawrence Livermore is a classic for wire antennas) and measured in anechoic chambers lined with radio-absorbing foam, or with near-field scanners that measure close in and compute the far field by a Fourier transform.
- Link budgets in practice. Professional budgets add fade margins for rain and multipath, use statistical propagation models from ITU-R recommendations or 3GPP channel models rather than free space, and are run across thousands of locations in network-planning software.
- Over-the-air testing. Phones and base stations with integrated arrays have no connector to plug a cable into at millimetre-wave frequencies, so their performance is certified over the air in chambers, following test plans written by industry bodies.
- Receivers on a chip. A phone's radio front end is a few chips containing direct-conversion transceivers for many bands, with filters (acoustic-wave devices) and switches in a front-end module; the digital back end is part of the modem.
Common misconceptions
Misconception
"A high-gain antenna amplifies the signal." An antenna is passive. Its gain is directivity times efficiency, always relative to an isotropic antenna radiating the same total power, and it is bought by radiating less in other directions. A 20 dBi antenna pointed the wrong way receives less than a 0 dBi one.
Misconception
"Higher frequencies suffer more free-space loss because the air absorbs them." Free-space path loss involves no air at all. The term appears only because an antenna of fixed gain has a smaller effective aperture at higher frequency. With fixed-size dishes at both ends, the gains grow as each and a higher frequency gives a stronger link. Atmospheric absorption is real but separate, and matters mostly above about 20 GHz and in rain.
Misconception
"Turning up the receiver's gain lets it hear weaker signals." Gain amplifies signal and noise together. Sensitivity is set by the noise figure of the first stages and the bandwidth, not by how much gain follows.
Misconception
"The antenna should be a quarter or half wavelength long, or it will not work at all." Resonant lengths make matching easy and efficiency high, but any conductor radiates. Electrically small antennas work, with low radiation resistance, narrow bandwidth and careful matching, as every phone and smartwatch shows.
Exercises
1. Antennas and radio links
Warm-upHumanWhat is the free-space path loss at 2.4 GHz over 100 m, in dB, to the nearest whole number? Use m/s.
2. Antennas and radio links
Warm-upHumanWhat is the thermal noise power at K in a 20 MHz channel, in dBm, to the nearest whole number? (Recall that is about dBm per hertz.)
3. Antennas and radio links
CoreHumanA Wi-Fi access point's datasheet says its antenna has a gain of 6 dBi. What does this mean?
4. Antennas and radio links
CoreHumanA soil-moisture sensor on a farm transmits at 868 MHz with 14 dBm into an antenna of 2 dBi gain. The gateway, 10 km away in clear line of sight, also has a 2 dBi antenna. Using the Friis equation (free space, no other losses), what is the received power in dBm, to the nearest whole number?
5. Antennas and radio links
CoreHumanAn FM superheterodyne receiver is tuned to a station at 100.0 MHz. It uses an intermediate frequency of 10.7 MHz with the local oscillator above the wanted signal. At what frequency is the image, the other frequency that the mixer also converts to 10.7 MHz? Give it in kHz.
6. Antennas and radio links
CoreComputationalWrite cascade_nf(stages), where stages is a list of (gain_dB, noise_figure_dB) pairs in order from the antenna, returning the noise figure of the whole chain in dB by Friis's formula for noise:
Remember to convert decibels to ratios before using the formula, and back at the end. The tests include the three chains from the lesson and two passive losses in a row.
7. Antennas and radio links
StretchComputationalWrite array_pattern(n, d_over_lambda, steer_deg, theta_deg), returning the normalised array factor of n isotropic elements in a line, spaced d_over_lambda wavelengths apart, with progressive phase that steers the beam to steer_deg. theta_deg is a NumPy array of angles from broadside, and the function returns an array of the same shape.
The tests check the broadside beam, the first null of an 8-element array, steering to several angles, and the grating lobe that appears with one-wavelength spacing.
Further reading and resources
- GNU RadioThe free, open-source toolkit for software-defined radio, with tutorials that start from a first receiver.
- International Amateur Radio UnionThe federation of national amateur radio societies, which explain how to obtain a licence to transmit in each country.
- ITU Radiocommunication Sector (ITU-R)Publishes the Radio Regulations and the propagation models used in professional link budgets.
- A Note on a Simple Transmission Formula · Harald T. Friis · 1946Proceedings of the IRE, vol. 34. The two-page paper that introduced the transmission equation in this lesson.
- Noise Figures of Radio Receivers · Harald T. Friis · 1944Proceedings of the IRE, vol. 32. Where the noise figure and the cascade formula were defined.
- NEC-2: Numerical Electromagnetics Code · G. J. Burke and A. J. Poggio, Lawrence Livermore National Laboratory · 1981A long-established method-of-moments program for wire antennas; free and open-source versions are widely available.