WHAT IT IS

A passive electronically scanned array (PESA), also called a passive phased array, is a radar antenna made of many small radiating elements whose beam is steered, with no moving parts, by setting the relative phase of the signal at each element. The word "passive" describes the face. The elements and the phase shifters behind them do not generate power. The radio-frequency energy is made somewhere else, usually in one high-power tube such as a travelling-wave tube (TWT), a klystron or a crossed-field amplifier, or in a small number of them. It is then carried to the array through a feed, delayed or advanced element by element, and radiated. On receive, the echoes come back through the same phase shifters and are added together in a shared receiver.

That makes a PESA a halfway step between two older ideas. A mechanical radar points one large antenna by turning it. A fully "active" array (an AESA) puts a small transmitter and receiver behind every element. A PESA keeps the single central transmitter of the mechanical radar and takes only the pointing electronic. For about thirty years, from the 1960s to the 1990s, this was the only practical way to build an electronically steered radar that could also handle megawatts or kilowatts of power, and it is still in service in a very large number of ground, ship and air systems.

This article explains the physics from the beginning: why many small radiators form a beam, how a phase gradient tilts it, what limits the scan angle and the bandwidth, and what the real hardware looks like. Then it covers the history, a set of real radars, the strengths and weaknesses, and a side-by-side comparison with the AESA. It is a general technology article. Russian and Soviet radars are used as examples where they are the best examples, and the Cyrillic names appear only there.

Related articles on this site: AESA radar for the other half of the comparison; Don-2N, A-135 / A-235 Nudol and Elbrus for the Moscow missile-defence radar and its computers; 91N6E, S-400, S-300 and S-500 for the air-defence radars; Patriot for the American AN/MPQ-53; Voronezh, Dnepr and SPRN at Solnechnogorsk for early warning; MiG-31 for the first PESA in a fighter; and over-the-horizon radar for a different way of using large arrays.

QUICK FACTS

• Full name: passive electronically scanned array (PESA); also "passive phased array" or "passive ESA".
• Steering: electronic, by phase shifters (ferrite or diode) behind each element or small group of elements. Typical beam-switching time is microseconds to about a millisecond depending on the phase-shifter type (see below).
• Power source: one or a few high-power tubes (TWT, klystron, crossed-field amplifier) or a central solid-state amplifier; not one amplifier per element.
• Receive path: phase shifters and a combining network feed one or a few shared receive channels. Low-noise amplification is normally after the combiner.
• Feed types: constrained (corporate or series) waveguide feed, or optical "space" feed with a horn illuminating a lens or reflection array.
• Best known examples: AN/FPS-85 (USA, 1960s, a borderline case), Safeguard MSR (USA, 1970s), Cobra Dane (USA, 1977), AN/SPY-1 Aegis (USA), AN/MPQ-53 Patriot (USA), N007 Zaslon on the MiG-31 (USSR), N011M Bars and N035 Irbis-E (Russia), 91N6E and 92N6E of the S-400 (Russia).
• Main strengths: very high peak power from proven tubes, mature and well-understood, excellent beam agility compared with mechanical scanning, can interleave search, track and missile guidance.
• Main weaknesses: one transmitter is a single point of failure, losses in the feed and phase shifters, one waveform at a time, limited instantaneous bandwidth, heavy and costly ferrite hardware, harder to make low-probability-of-intercept.
• Successor: the AESA, which since about 2000 has been replacing PESAs in new fighters, ships and ground radars. PESAs remain in production in some families.

WHY MANY SMALL ANTENNAS MAKE A BEAM

Every radiating element on its own sends energy out in a very broad pattern, nearly in all directions in front of it. A beam is made by letting the signals from many elements interfere. At a far-away point, the signals from the different elements arrive having travelled slightly different distances. If those paths differ by a whole number of wavelengths, the waves are in step and add up. If they differ by half a wavelength, they are opposite and cancel. Set up properly, the waves add in one narrow direction (the main beam), cancel in most others, and leave small leftovers in between (the sidelobes).

For a straight row of N identical elements spaced a distance d apart and each driven with a phase step of β from one element to the next, the field in a direction θ measured from broadside (straight out of the face) is the sum of N phasors. This sum is called the array factor:

AF(θ) = sum for n = 0 to N-1 of a(n) · exp[ j · n · (2π d / λ · sin θ − β) ]

where a(n) is the amplitude weight on element n and λ is the wavelength. All the terms add exactly in phase, and the pattern peaks, when (2π d / λ) sin θ = β. That is the whole principle. Choose β = 0 and the beam points at broadside. Choose another β and the beam points at

sin θ0 = β λ / (2π d)

For half-wavelength spacing (d = λ/2), this becomes β = π sin θ0. A beam 30 degrees off broadside needs a phase step of 90 degrees from one element to the next; one at 60 degrees needs about 156 degrees. The beam-steering computer calculates one such number for each element, or for each small group, and sends it to the phase shifters. Nothing has to move. A planar array does the same thing in two directions at once, with one phase gradient for azimuth and another for elevation.

The last point is that the array factor is multiplied by the pattern of one element. The pattern of the whole antenna is (element pattern) times (array factor). The array factor makes the narrow beam. The element pattern, which is broad, makes the envelope that the beam sits inside. This product is the reason for scan loss, discussed below.

BEAMWIDTH, GAIN AND THE SIZE OF THE FACE

The width of the main beam depends on how big the array is measured in wavelengths. For a uniformly illuminated row, the half-power beamwidth is about

θ3dB ≈ 0.886 λ / (N d cos θ0) radians, which is roughly 51° divided by (L / λ), where L is the length of the aperture,

and, for half-wavelength spacing, about 102° divided by N when the beam is at broadside. Two things are visible immediately. A longer array makes a narrower beam. And the beam gets wider by a factor of 1 / cos θ0 as it is steered off broadside, because the face looks shorter when seen from the side. At 60 degrees the beam is about twice as wide as at broadside.

Gain follows the effective area. The directivity of an aperture is about 4π A / λ², with A the effective area. With half-wavelength spacing on a square lattice, each element accounts for about λ²/4 of area, so the directivity is roughly π times the number of elements. As a worked example only, an array of 5,000 elements would have an ideal directivity of about 15,700, or about 42 dB, before any of the losses described later. Real gain is lower, and the effective area also shrinks with scan angle, as described next.

A narrow beam is the whole point of a high-gain radar antenna. It concentrates power onto a small patch of sky, which raises range, and it improves angular accuracy and resolution. It also means that to cover a sector, the radar has to point the beam at many positions in turn. The speed at which a phased array can do this is what makes it different from a dish.

WHY STEERING HAS LIMITS: SCAN LOSS

A flat face steered away from broadside has two penalties that add up.

• Projected aperture. Seen from an angle θ0, the face has an effective area reduced by cos θ0. Gain falls by that factor, and the beam broadens by 1 / cos θ0.
• Element pattern. The individual elements radiate less strongly toward the side than straight ahead, and mutual coupling (the elements influencing one another) makes this worse and can even cause "scan blindness" at particular angles.

Together these are commonly modelled as a gain loss proportional to cos^1.5 of the scan angle, though the exponent depends on the element design; wide-pattern elements reduce the loss. This is a model and not a law. Textbooks give figures between cos^1 and cos^2.5 depending on whether one counts one way or both ways and on the element. Using cos^1.5 for one-way gain as an illustration: at 45 degrees the loss is about 2.3 dB, and at 60 degrees about 4.5 dB. A radar loses this on transmit and again on receive, so the two-way effect on signal-to-noise ratio is about double, around 9 dB at 60 degrees, in addition to the doubled beamwidth. Typical limits for a flat array are therefore about plus or minus 60 degrees, sometimes quoted as 50 to 70.

That limit explains many designs. A radar that has to see the whole sky is built with several flat faces: four fixed faces on the Aegis ships, four on the building of Don-2N, two on the rotating head of the 91N6E. A fighter nose array covers a cone of about plus or minus 60 to 70 degrees, which is why designers use gimbals or "repositioners" on some arrays to turn the face and extend coverage. The Irbis-E uses exactly this approach: electronic scan of about plus or minus 60 degrees, plus a mechanical drive that moves the face to extend the field of view to about 120 degrees each side (published figures).

GRATING LOBES AND ELEMENT SPACING

The array factor has a repeating nature. For a given spacing, beams of equal strength appear in other directions where the path difference between neighbouring elements is a whole wavelength. These are grating lobes. They waste power, create ambiguous angles and let jammers or clutter in from unwanted directions. They appear when the element spacing is too large for the scan angle.

The condition for no grating lobe to appear anywhere in real space, for a beam scanned to θmax, is

d / λ ≤ 1 / (1 + sin θmax)

For a scan of plus or minus 60 degrees the spacing must be at most about 0.54 wavelength; for plus or minus 90 degrees it must be 0.5 wavelength; for plus or minus 45 degrees about 0.59. This is why "half a wavelength" is the standard number. Triangular (hexagonal) lattices can use somewhat larger spacing for the same scan limit and need about 14 percent fewer elements than a square lattice, which is why they are common.

The rule has to hold at the highest frequency in the band, since the spacing in wavelengths grows as the wavelength shrinks. That is a design limit on the whole bandwidth of the radar. The reverse, the element being too close together, makes mutual coupling worse and increases cost, so designers cluster near half a wavelength.

Quantisation of the phase matters too. A digital phase shifter with n bits steps in increments of 360° / 2^n: 22.5 degrees for a 4-bit shifter, 5.6 degrees for a 6-bit shifter, 2.8 degrees for a 7-bit shifter. Rounding to the nearest step produces small phase errors across the aperture. These raise the average sidelobe level and, in the worst case with a regular pattern of errors, produce "quantisation lobes" of power about 6 dB lower for each additional bit (about -6n dB relative to the main beam). Early radars with 3-bit or 4-bit phase shifters had to live with higher sidelobes; the Patriot AN/MPQ-53 is described with 4-bit ferrite shifters, and AN/SPY-1 moved from 4-bit shifters to 7-bit shifters in later versions (the latter from a Navy technical history). Some arrays randomise the errors across the face to turn quantisation lobes into a low, even floor.

PHASE SHIFTERS: WHERE A PESA SPENDS ITS MONEY

The phase shifter is the central component. Two families matter.

Ferrite phase shifters. A ferrite (garnet) toroid, or a rod, sits in a waveguide. A pulse of current through a wire that threads the ferrite magnetises it to a chosen level, and the magnetisation changes the speed at which the microwave signal travels through it, which sets the phase. Latching types hold the setting with no power until the next pulse. Ferrite shifters handle very high peak power, which is why they sit behind the tube of a PESA. Typical figures from the engineering literature: insertion loss of roughly 0.4 to 1.6 dB over the full range, and switching times of microseconds, perhaps 1 to 100 microseconds depending on design (modern high-current designs under 10 microseconds; older ones 20 to 300). Their weaknesses are weight, cost and the drive electronics. The AN/SPY-1 development team at the Johns Hopkins Applied Physics Laboratory made a particular advance with a garnet ferrite that was far less sensitive to temperature, which mattered for keeping thousands of shifters calibrated.

Diode phase shifters. PIN diodes switch short sections of line in or out. They switch in nanoseconds (roughly 1 to 100 ns in modern designs), are light and cheap, and need only a low-voltage drive. However, they handle much lower power and have more loss per bit, around 0.5 to 1 dB per bit, so a 4-bit device may cost 1 to 2 dB and a 5-bit device 2.5 to 5 dB. Diode shifters therefore appear in lower-power arrays, in the low-power part of an AESA module, and in some PESAs where the power is distributed to subarrays.

In both cases the phase shift is a fixed setting. It does not change with frequency the way a real time delay would. That has a large effect on bandwidth, described in the next section.

Because every element needs its own shifter and drive, a PESA has thousands of them. A Patriot AN/MPQ-53 array is described with about 5,000 ferrite phase shifters (5,161 elements in one source). The AN/SPY-1B/D has 4,350 radiating elements per array, in 2,175 two-element subarrays, with a garnet phase shifter at each element; there are four arrays on a ship. The 91N6E is described in published material as having on the order of 2,700 phase shifters per face; that figure is a published one, as listed in this site's 91N6E article.

TIME DELAY, PHASE AND BANDWIDTH

A phase shifter sets phase, and a time delay sets time. They are equal only at one frequency. To steer a beam properly for a wide-band signal, the element-to-element delay must be a fixed time (about d sin θ / c), which is a different phase at each frequency. With plain phase shifters the phase is fixed, so when the frequency moves away from the design value the beam points somewhere else. This is called beam squint. The relation is f · sin θ0 = (f + Δf) · sin(θ0 + Δθ). For example, at a scan angle of 45 degrees and a frequency change of 5 percent, the beam moves by about 2.7 degrees, which is more than the beamwidth of a narrow-beam radar.

Another way to see it is aperture fill time. A signal sent along the face takes time T = L sin θ / c to cross the aperture; for a 1 metre face at 60 degrees that is about 3 nanoseconds. A signal with a bandwidth wider than about 1/T would be distorted because different parts of the aperture see different parts of the pulse. A common rule is that at wide scan angles the usable fractional bandwidth is about two times the beamwidth in degrees; for a 2 degree beam this is about 4 percent, or about 120 MHz at 3 GHz.

Phase-shifter arrays therefore have limited "instantaneous bandwidth": the width of the signal they can send or receive at once with the beam pointing correctly. The "tunable" or "operating" bandwidth (the range over which the centre frequency can be moved when the shifters are reset) is wider, and the reason a PESA can still frequency-hop pulse to pulse. In the AN/SPY-1, public descriptions mention an instantaneous bandwidth of the order of tens of megahertz in the 3.1 to 3.5 GHz S-band, with measurements over much wider bands.

The cure for squint is true time delay (TTD): switched lengths of line that give a real delay that is the same at all frequencies. TTD units are larger and lossier, so they are used on the sub-arrays of very wide band arrays instead of at every element. Many modern wide-band radars combine phase shifters at each element with TTD at the subarray level. A PESA built on ferrite shifters is, by its nature, a narrow-band to moderate-band device. This was acceptable for tracking and surveillance radars, and it matters more for wide-band imaging and electronic warfare, where AESAs and digital arrays have the advantage.

THE ARCHITECTURE OF A PESA, FROM EXCITER TO ELEMENT

The signal path of a typical PESA on transmit runs: exciter (a stable low-power source and waveform generator) → driver stages → high-power amplifier (tube) → duplexer or circulator → feed network → phase shifters → radiating elements. On receive it runs in reverse: element → phase shifter → combining network (which also forms the sum and difference channels used for monopulse angle measurement) → duplexer and receiver protector → low-noise amplifier → down-conversion and digital processing.

The transmitter. Early PESAs used one big tube or a few. The travelling-wave tube is an amplifier with a wide bandwidth, good gain and the ability to transmit a programmed waveform, which is why it became the standard. Klystrons are more powerful and efficient at a fixed frequency; crossed-field amplifiers (amplitrons) were used where lots of peak power and efficiency were more important than gain. Some designs gang several tubes together. As an example, the AN/MPQ-53 uses a TWT-based transmitter with upgrades adding a second TWT (published description), the Irbis-E uses a pair of 10 kW tubes for 20 kW peak (published description), and the AN/FPS-108 Cobra Dane is described with 96 TWTs in 12 groups of eight, each rated roughly 175 kW peak. At the other end of the range, the Moscow-area radars described on this site used much larger devices for megawatt pulses.

A useful way to compare architectures is to divide the peak power by the number of elements. For the Irbis-E, 20,000 W over 1,772 elements is about 11 W per element, before any losses. This is about the same order as the per-module power of a modern X-band AESA. The two designs do not differ because one has more peak power; they differ in where the power is made, how much of it survives to the antenna, and how flexibly it can be controlled.

The feed. The feed carries the transmitter's power to every element and collects the received signal. Two families exist.

• Constrained (corporate or series) feed. A branching tree of waveguide or coaxial lines, each branch dividing the power, leads from the transmitter to each phase shifter. A corporate tree keeps all paths the same length, which keeps the beam from squinting with frequency. A series (travelling-wave) feed taps off the elements one after another along a line and is simple, but its beam squints with frequency. Constrained feeds are efficient when well made and were common in ship and airborne PESAs.
• Space (optical) feed. A horn or a small feed array at some distance illuminates the back of the main array in the same way as a lamp illuminating a lens. The back of each element picks up a small part of the wave, passes it through a phase shifter, and the front of the element radiates it. A "transmissive" version is a lens; a "reflective" version is a mirror with a phase shifter in front. The space feed avoids a huge branching waveguide tree across a very large face and is particularly attractive on weight-sensitive ground radars. The AN/MPQ-53 of Patriot and the Soviet 30N6 engagement radars of the S-300 family have been described as using this lens approach with ferrite shifters, and the 91N6E as a space-fed reflection-type dual-face array.

A space-fed array is still passive. Nothing in the lens produces power. It only imposes phase on a wave that the central tube already made.

Subarrays. Between a single tube and a tube per element there is a middle path. In arrays of subarrays, each group of 32 or 64 elements gets its own tube or feed, and the phase shifters work inside the group while time delay or phase control works between the groups. The AN/SPY-1 was developed from the APL AMFAR demonstrator, which proved the idea of a subarray (say 64 elements) fed by a high-power microwave tube, so that very high total power could be made without any single tube producing it all. Public descriptions of how many transmitters the SPY-1A has differ: a 1977 US Naval Institute article describes two transmitters, forward and aft, while a later reference describes eight transmitters each with four 132 kW crossed-field amplifiers driving transmit subarrays. I have left the numbers out of the comparison because the sources disagree and the design changed between versions.

Receive. On receive there is generally one sum channel and two difference channels (azimuth and elevation) for monopulse measurement, and sometimes a guard or sidelobe-cancelling channel. The small number of receive channels is a defining PESA feature. It means the radar has one view of the world at a time, and a beam is formed by the analogue feed. Some hybrid radars add receive channels or low-noise amplifiers at the subarray level; the Bars and Irbis-E families are described as hybrid designs with several receive channels, but their transmit side is still a passive tube-fed array.

LOSSES: WHY THE FEED MATTERS TWICE

In a PESA all the power travels through the phase shifters and the feed on the way out, and all the weak echo travels back through them on the way in, before it reaches a low-noise amplifier. Every decibel of loss is paid twice.

A worked example with round numbers (illustrative, not any specific radar): suppose the phase shifter, feed and duplexer have a total of 3 dB loss each way. On transmit, only half the power reaches the sky. On receive, a 3 dB loss ahead of the amplifier is the same as raising the system noise by 3 dB. Both effects reduce the signal-to-noise ratio, a total of 6 dB. Because radar range varies as the fourth root of the signal-to-noise ratio, a 6 dB loss reduces range by a factor of about 0.71, around 29 percent shorter. The published ferrite figures of 0.4 to 1.6 dB per shifter show that this is not an exaggeration, especially once the long waveguide runs, the combiner and the duplexer are added.

An AESA avoids most of that. It places the power amplifier and the low-noise amplifier right behind the element so the long lossy path has little effect. Carlo Kopp, writing in 2002, estimated that having the low-noise receiver within the antenna cuts receiver thermal noise by a factor of two to four, and noted that the ferrite shifters of earlier passive arrays made them less sensitive than the mechanically steered antennas they replaced, which was compensated by more powerful tubes. The same loss is the reason that the PESA keeps the high-power tubes: it needs the power to overcome its own feed.

Power handling and cooling. The feed also has to carry the tube's full peak power without arcing, which requires large waveguide, sometimes pressurised, and careful design of every joint. The heat, though, is concentrated at one or a few points (the tubes and their power supplies), which is simpler to cool than thousands of small amplifiers spread over the face but means that the tubes need high-voltage supplies, which are heavy and a reliability concern.

HOW A PESA SEARCHES, TRACKS AND GUIDES AT ONCE

The main operational gain of a PESA over a dish is beam agility. A computer decides, pulse by pulse or dwell by dwell, where the next beam goes, how long it dwells and what waveform it carries. The beam can jump from one end of the sector to the other in the time it takes to reset the phase shifters: a few microseconds for modern ferrite shifters, tens to hundreds of microseconds in older ones. Published figures for the radars below are of that order: 1.2 milliseconds for the beam positioning of the Zaslon, 400 microseconds for the N011M Bars. A dish takes seconds.

Agility allows time-sharing. The radar divides its time into short slices. In one it searches a sector, in another it fires a confirmation beam at a new detection, in a third it tracks a known target, and in a fourth it sends an uplink to a missile in flight. All of this is interleaved so quickly that every function seems to run at once. Patriot's AN/MPQ-53 is a standard example: it is described as performing surveillance, tracking, missile guidance and electronic-protection functions in a time-shared way, with the computer building action cycles lasting milliseconds, with up to 32 radar configurations and 54 operational modes (published descriptions). AN/SPY-1 does the same at sea: its 1977 description states that instead of painting a target once per rotation, the radar uses hundreds of pencil beams, and when a detection occurs the computer schedules several more dwells on that target within a second, creating a track in less time than it takes a rotating radar to complete one turn.

Notice the limit, though. At any instant a PESA transmits one beam with one waveform. Time-sharing hides this, but it takes real time from the radar's budget. If a radar spends 40 percent of its time tracking and guiding, that time is not spent searching. An AESA, which can split its face into groups and send different beams with different waveforms at the same instant, can use the same face for several jobs concurrently; see the AESA article.

HISTORY: FROM BRAUN TO AEGIS

• 1905. Ferdinand Braun demonstrates directional transmission with three antennas whose phases are controlled, the first demonstration of the beam-forming principle.
• World War II. Phased arrays are built in several countries. The German GEMA FuMG 41 "Mammut" was a large VHF phased-array radar with phase-controlled elements, produced late in the war. In the United States, Luis Alvarez developed an electronically steered microwave phased array for ground-controlled approach; the US Navy's Mk-VIII from Bell Laboratories/Western Electric was described by the US Naval Institute as the Navy's first operational phased-array radar. Phase-control in these systems was largely mechanical or by frequency, not the digital shifters of later radars.
• Late 1950s to early 1960s. Ferrite and computer technology made large electronically steered arrays practical. The Hughes AN/SPS-32 and AN/SPS-33 "SCANFAR" were installed on USS Long Beach and USS Enterprise in 1961 and 1962; SPS-33 used frequency scanning in elevation and phase scanning in azimuth.
• 1960s: AN/FPS-85 at Eglin. Construction began in 1962 at Eglin Air Force Base, Florida, and was delayed by a fire; the Air Force took ownership in 1968. It operates around 442 MHz (UHF) and has a transmit array of 5,184 modules in a 72-by-72 matrix, each rated for up to 10 kW peak, and a separate receive array of about 19,500 crossed-dipole elements with about 4,660 active receive modules, with a published peak power of about 32 MW (some sources 30 to 35 MW). It is a borderline case for this article because each transmit element has its own tube amplifier, which sets it apart from a single-tube PESA. I describe it as an early distributed-transmitter array and not a classic PESA; sources do not use a standard label for it.
• 1970s: Safeguard. The Missile Site Radar at Nekoma, North Dakota, had four faces of an S-band phased array, each about 13 feet across with about 5,000 elements, and reached initial operational capability on 1 April 1975. The separate Perimeter Acquisition Radar was a single-face UHF phased array with 6,888 elements (published figures from the Army's Safeguard history).
• 1977: Cobra Dane (AN/FPS-108). At Shemya in the Aleutians, an L-band (1,175 to 1,375 MHz) array about 29 m across with 15,360 active elements, fed by 96 TWTs, with a reported peak power of about 15.4 MW. It is a clear example of a big, tube-fed, subarray-style PESA, used for missile-test observation and space tracking.
• 1970s to 1980s: AN/SPY-1 Aegis. Four fixed S-band faces, each with about 4,350 elements, 3.1 to 3.5 GHz, entered US Navy service in the early 1980s. Variants SPY-1A, B, D and later improved the sidelobes, shifter resolution and processing.
• 1970s to 1980s: Patriot AN/MPQ-53. C-band space-fed ferrite lens with about 5,000 shifters; one face that performs search, track and missile guidance (see Patriot).
• 1980s: first airborne PESAs. Soviet N007 Zaslon (developed between 1975 and 1980) in the MiG-31. Carlo Kopp's 2002 survey also lists the electronically steered AN/APQ-164 radar of the US B-1B bomber (in service from the 1980s) as an early airborne passive array.
• 1980s to 2000s: Soviet and Russian ground radars. The 30N6 engagement radar of the S-300 family and the 9S32 of the S-300V were PESAs with an optical space feed, TWTs and ferrite shifters, as the Microwave Journal's history lists them. The 91N6E and 92N6E followed for the S-400.
• 2000 onward. The AESA takes over in new fighters, starting with the AN/APG-63(V)2 on US F-15Cs in 2000, but PESAs continue in production and upgrades, such as the Irbis-E on the Su-35.

Also worth knowing: the AN/FPS-115 PAVE PAWS, from the late 1970s, used solid-state transmit-receive modules (1,792 active per face) and so is an early active array, not a PESA. The "PESA era" and the "AESA era" overlapped for decades.

HARDWARE FAMILY TREE: WHERE "PASSIVE" ENDS AND "ACTIVE" BEGINS

It is helpful to see the designs as a spectrum of how far the power is distributed.

• One tube, one feed (Patriot AN/MPQ-53, Zaslon): the classic PESA. All power is in one place.
• A few tubes, each feeding a subarray (AN/SPY-1, Cobra Dane with 96 TWTs): still passive in the sense that the phase shifters are passive and the tubes are shared, but there is partial redundancy because one tube failing takes out only a part of the face.
• One tube or large module per element or per big group, with phase control (AN/FPS-85 with 5,184 transmitters; Don-2N with a small number of megawatt-class modules according to Russian sources): this is the grey area. Whether you call these "active" depends on the definition.
• Low-power solid-state amplifier at every element (PAVE PAWS, then fighter AESAs from the 1990s): active, with hundreds or thousands of small transmitters.

The usual dividing line is whether there is a transmit amplifier at or immediately behind the radiating element, along with a receive amplifier, so that the power is made in the aperture and not fed to it. By that test, tube-fed subarrays count as passive and a transmitter per element counts as active, and the big-module arrays fall in between. A design can be argued either way, and an honest article should say so. The Don-2N article on this site does so explicitly.

SOVIET AND RUSSIAN PESA RADARS (Радары с пассивной фазированной антенной решёткой)

Russia and the Soviet Union built many PESAs, because they had mature tube and ferrite technology before solid-state modules were available at scale, and some of the systems are still in production.

Заслон (Zaslon, N007; NATO Flash Dance). The radar of the MiG-31 interceptor, developed from 1975 to 1980 at the Tikhomirov research institute (НИИП, NIIP). It is a pulse-Doppler X-band radar with a PESA antenna, a diameter of about 1.1 m (some sources say 1.4 m), a fixed face with a published scan sector of about plus or minus 70 degrees in azimuth and +70/-60 degrees in elevation. The published description lists about 1,700 X-band emitters and 64 L-band emitters (for the identification transponder) in the same antenna, ferrite phase shifters with a beam-positioning time of about 1.2 ms, detection of a 16 square metre target at 200 km, and the ability to track 10 targets and engage four (Rosoboronexport catalogue, via Wikipedia). It was the first PESA fitted to a jet fighter. The MiG-31 is described in this site's MiG-31 article.

Барс (Bars, N011M). A hybrid PESA for the Su-30 family. Published figures: a slotted-array antenna about 960 mm in diameter, peak power of 4 to 5 kW and average about 1.2 kW, a beam-positioning time of 400 microseconds, and a fixed-face scan sector of about plus or minus 70 degrees in azimuth and plus or minus 45 in elevation, extended to about 90 degrees by mounting the antenna on a mechanical drive. It builds on the Zaslon experience.

Ирбис-Э (Irbis-E, N035). An X-band hybrid PESA for the Su-35, developed from the Bars. Published figures: a 900 mm antenna with 1,772 radiating elements, two 10 kW tubes for 20 kW peak and about 5 kW average, electronic scanning of about plus or minus 60 degrees plus a mechanical drive extending the total field of view to about 120 degrees, detection of a 3 square metre target at up to 400 km in a narrow search sector (200 km otherwise), tracking of up to 30 targets and attack on up to 8 (NIIP product data as reported). Whether the 400 km figure holds in service is not independently verified, and it is a manufacturer's number.

91Н6Е and 92Н6Е (91N6E, 92N6E). The S-400's acquisition radar and engagement radar. The 91N6E is a rotating, space-fed, dual-face reflection-type PESA in the S band (the "Big Bird") and the 92N6E is an X-band phased-array engagement radar. Details are in the 91N6E and S-400 articles. Both are tube-fed passive arrays by the descriptions on this site.

30Н6 (30N6) and the S-300 family. The engagement radars of the S-300 are described by the Microwave Journal's history as PESAs with passive optical space feed, transmissive ferrite arrays and TWT transmitters, similar in design approach to Patriot's AN/MPQ-53. See S-300.

Дон-2Н (Don-2N). The Moscow ABM radar at Sofrino. Here the sources disagree, and the disagreement is instructive. Some Western sources, and an earlier version of this very article, listed it as a PESA. Russian developers describe it as a large-aperture active phased array in which the transmit power is generated within the antenna: a memoir by the head engineer of the manufacturing plant at Dnepropetrovsk states that Don-2N adopted the "active phased array with power distributed in the aperture" model, with megawatt-class transmit modules of about 300 kg each, tens of them per face. That is neither a single-tube PESA nor a modern AESA with thousands of small modules. The module count is not public. The Don-2N article gives the full account and takes the reading "active, large-module". This article follows that reading and does not list Don-2N as a PESA. The real lesson is that "passive" and "active" are the ends of a range and that big-module arrays sit between.

Воронеж (Voronezh). The current family of Russian early-warning radars. Descriptions in sources differ: some call the arrays active phased arrays, while a technical reference for the Voronezh-DM describes a passive phased array with transmit power from air-cooled disk triodes at around 470 MHz. This article does not settle it. The Voronezh article on this site uses a hedged description. The older Dnepr radars are a different principle again: they steer in azimuth by changing the frequency of the transmitted signal, not by phase shifters. The SPRN at Solnechnogorsk article describes how their data are used.

Резонанс-Н (Rezonans-N). A metre-band (VHF) surveillance radar with a phased antenna array designed to detect low-observable aircraft; the sources I found do not state whether the array is passive or active, so it is mentioned here only to show that the choice of band matters as well as the choice of architecture.

Chinese and Japanese examples. China's KLJ-3 (Type 1473) fighter radar on the J-10B is listed in the sources I found as a PESA, while the J-20's Type 1475 and the Type 346 naval radar are AESAs. Japan's OPS-24 is listed as the first serially produced shipborne AESA, and Japan's F-2 fighter carried the J/APG-1, described as the first AESA on a combat aircraft (1995). I did not find reliable, detailed public specifications for Chinese or Japanese PESAs, so I give no numbers for them. Both countries moved to active arrays early compared with Russia.

WHAT A PESA IS USED FOR

• Ground-based air and missile defence. Patriot AN/MPQ-53, S-300 and S-400 radars: search, track and missile uplink from one array with a tube transmitter.
• Ship radars. AN/SPY-1 and similar: four fixed faces for all-round, horizon-to-zenith coverage with a very short time to establish tracks.
• Ballistic-missile early warning and space surveillance. Cobra Dane, Safeguard's PAR, AN/FPS-85, Don-2N and the Voronezh family: very large apertures, UHF to S-band, high power and quick beam switching. See A-135 / A-235 Nudol for how one of them is used to guide interceptors.
• Fighter radars of the 1980s to 2010s. Zaslon, Bars, Irbis-E: electronic scan of limited angle, often combined with a mechanical gimbal to widen coverage.
• Bomber and attack radars. Terrain-following and mapping radars such as the AN/APQ-164 (B-1B) used passive arrays for fast beam repositioning.

ADVANTAGES OF A PESA

• Very high peak power from proven tubes. Tubes are mature and can deliver kilowatts to megawatts. For long-range surveillance and for radars where peak power matters more than flexibility, this is a real advantage.
• Beam agility. Beam switching in microseconds to about a millisecond permits interleaving search, track and guidance, and allows track-while-scan on many targets.
• Lower number of active parts. One transmitter and a few receivers are much simpler than thousands of transistor modules, and in the 1970s and 1980s they were far cheaper than T/R modules.
• Mature and well understood. Maintenance practices, spare parts and manufacturing for tube-fed arrays are established, which is part of why they stay in production.
• Frequency agility between pulses. A PESA can hop its carrier frequency from pulse to pulse within the tube's band, while keeping beam direction by resetting the shifters, which gives ECCM resistance, subject to the squint limits above.
• Lower required dissipation at the face. The heat is made in the tube and its supplies, not in the array, which simplifies the aperture structure.

LIMITS OF A PESA

• Single point of failure. If the tube or its high-voltage supply fails, the whole face (or the whole group of elements it feeds) is down. A subarray design softens this but does not remove it. An AESA, in contrast, loses a few modules at a time.
• Losses in the feed and phase shifters. Paid twice, as the worked example shows, and they raise system noise on receive.
• One waveform per beam at a time. All elements radiate the same signal. Different beams with different waveforms at the same instant require splitting the face (with a gain loss) or time-slicing.
• Limited instantaneous bandwidth. Ferrite phase shifters give beam squint for wide signals, which limits wideband imaging and wideband electronic warfare.
• Heavy, costly ferrite hardware and drive electronics. Thousands of precisely made shifters with their drive circuits, calibrated and kept in tolerance.
• Slower reset than diodes. Ferrite shifters need microseconds, whereas diode shifters need nanoseconds.
• Lower reliability of the transmitter. High-voltage tubes and supplies have been the weak link of tube-fed radars. Kopp's 2002 figures for fighter radars, which compare typical conventional radar MTBF of 60 to 300 hours with 1,000 hours or better for an AESA, show the scale of the difference (a general comparison, not a measurement of any one radar).
• Harder to make low-probability-of-intercept. A big tube emits a strong, concentrated, predictable signal, and it cannot easily turn its power down or spread it over many weak beams. Passive arrays can still use pulse compression, frequency hopping and power management, but the AESA is better suited.
• Cannot be upgraded into an AESA. Swapping the shifters does not change the feed, the power distribution or the cooling. A conversion is a new radar.

PESA VERSUS AESA, SIDE BY SIDE

The full comparison is in the AESA article. The main points in short form:

• Where the power is made. PESA: in one or a few tubes, then fed to the face. AESA: in hundreds or thousands of small solid-state amplifiers, one for each element or small group, at the face.
• What is behind each element. PESA: a phase shifter (and sometimes a small attenuator). AESA: a complete transmit/receive module with a power amplifier, low-noise amplifier, phase shifter, gain control and switch.
• Failure behaviour. PESA: sudden loss when the tube fails. AESA: gradual loss of gain and a rise in sidelobes as modules fail (a loss of about 10 percent of modules costs less than 1 dB of one-way gain, though sidelobes can rise sooner).
• Beams. PESA: one beam, one waveform at a time, fast switching. AESA: can form several beams at once and split the face between functions.
• Bandwidth. PESA: limited instantaneous bandwidth because of ferrite shifters and the feed. AESA: wider, though phase-shifter squint still applies unless true time delay is added.
• Sensitivity. PESA: feed losses ahead of the first amplifier. AESA: the amplifier is at the element, so the noise figure is set by the module.
• Peak power. PESA: very high from a tube. AESA: modest per module but large in total, with high average power and flexible duty cycle.
• Cost and maintenance. PESA: cheaper to build in the days before cheap modules, and heavy on tube maintenance. AESA: costly in modules but lower maintenance in service.
• Detectability. PESA: harder to make low-probability-of-intercept. AESA: well suited to it.
• Geometry. Scan loss, grating lobes and the limit of about plus or minus 60 degrees are the same for both, because they come from the physics of a flat array, not from the power source.

WHICH IS BETTER, AND WHEN

An AESA is better for almost any new fighter, ship or air-defence radar. It is more reliable, can do more at once, and its costs have come down. A PESA still makes sense in a few cases: when an existing design is mature and cheap, when very high peak power at lower frequency is needed and a tube is the simplest way to get it, and when the cost of a new radar is not justified by the gain. Many of today's PESAs were designed in the 1970s to 1990s and have been upgraded in processing and software to stay useful.

The decision is not just a technology one. A radar that meets the requirement and can be built and maintained with the country's industry has value. A technically superior radar that cannot be made in quantity does not.

COMMON MISTAKES AND MYTHS

• "PESA means old and useless." The Aegis, Patriot and S-400 families show otherwise. They work, and many of their sensors have received new processing.
• "If it is a flat panel it is an AESA." The face of a PESA and an AESA look the same from the outside. The difference is behind the face.
• "A bigger tube makes a better radar." Not always. Feed loss, bandwidth and agility count as well as peak power.
• "Phased arrays have no moving parts so they never need mechanical scanning." Many fighter arrays use a mechanical gimbal to widen the field of view, such as the Irbis-E.
• "Active means more power." Not necessarily. Total power is about the same order in many designs. What changes is where the power is made and what survives.
• "Passive means the radar doesn't emit." It does. "Passive" here refers to the array face and not to the radar's emissions. A passive radar (one that listens to other transmitters) is a different thing altogether.
• "Don-2N is a PESA." See the Soviet section above and the Don-2N article: Russian designers describe it as an active large-module array.

WHAT IS UNCERTAIN IN THIS ARTICLE

• Scan loss exponent. cos^1.5 is a common model, not a universal law, and the correct exponent depends on the element design.
• Published radar specifications. Element counts, power, range and tracking numbers for Zaslon, Bars, Irbis-E, 91N6E and Cobra Dane are manufacturer or reference-work figures, and not independent measurements. They are marked "published" in the text.
• Zaslon antenna diameter. Sources give both 1.1 m and 1.4 m.
• AN/SPY-1 transmitter arrangement. Sources differ between two transmitters and eight transmitters with four amplifiers each; the design changed between variants.
• AN/FPS-85 element counts. Sources distinguish 5,928 transmit antennas (a 78-by-76 pattern) from 5,184 powered transmit modules; I use 5,184 for powered transmit modules.
• Voronezh, Don-2N and Rezonans-N array types. Not settled in the sources I found.
• Irbis-E range. The 400 km narrow-sector figure is a manufacturer's claim.
• Some historical details (such as the extent of early wartime phase control) are summarised from secondary sources.

IN ONE SENTENCE

A PESA radar steers its beam by setting a computer-controlled phase at each element of a flat array fed from one tube transmitter, so it gets microsecond beam agility and very high peak power without moving the face, at the price of feed losses, one waveform at a time, limited bandwidth and a single transmitter that can take the whole radar down.