Chapter 155 - Chapter 147: The Eye Above the Sky

Chapter 147: The Eye Above the Sky

Shergill Aviation Advanced Systems Facility, Gorakhpur2 March 1974 — 07:00 Hours

The aircraft in Hangar Seven did not look like anything that had existed in the world before.

To a casual observer walking in from the morning cold, it might have looked like an ordinary passenger aircraft that had been through some unusual modification — the same wide fuselage, the same four propeller engines mounted on the wings, the same general shape of a plane that was designed to carry people from one place to another. Anyone who flew Indian Airlines in the 1960s would have recognised the basic outline of an Ilyushin Il-18, the Soviet-designed turboprop workhorse that had carried passengers between Delhi and Calcutta and Bombay for years before the jet age made it feel dated. The Il-18 was not a glamorous aircraft. It was reliable, capacious, and a little ungainly — a sturdy thing built to work rather than to impress.

But above the fuselage of the aircraft in Hangar Seven, mounted on two thick metal struts rising from the roof like arms holding something precious, was a disc. Six metres across — roughly the width of a country road — perfectly circular, enclosed in white fibreglass, sitting above the aircraft's spine like a flying saucer that had decided to attach itself to this particular machine and go wherever it went. The disc was the thing that made this aircraft unlike any other aircraft in India, and unlike most aircraft anywhere in the world in March of 1974.

Inside that disc was a radar antenna. And that antenna, when the aircraft was flying, would rotate at six full turns per minute — turning continuously, sweeping the sky in every direction, looking outward in all 360 degrees of the compass simultaneously. Every ten seconds it completed one full revolution. Every ten seconds it scanned the entire sky around the aircraft, from the near horizon all the way to the edge of what radar could reach, which at the altitudes this aircraft would fly was an enormous circle of airspace hundreds of kilometres across.

This disc — this rotating radar enclosure — had a name in the engineering world: a rotodome. Roto for rotating, dome for the shape. It was the defining visual feature of a class of aircraft that the military world was only beginning to develop in the early 1970s, aircraft that were not fighters and not bombers and not transport planes but something entirely different: flying radar stations, designed not to carry cargo or drop bombs but to see — to build a complete picture of everything flying in a vast region of sky and to share that picture with commanders who could then direct their fighters and missiles and ships accordingly.

The Americans called their version AWACS — Airborne Warning and Control System. It was still in full-scale development in 1974, not yet operational, a programme that Boeing and the United States Air Force were building on a modified Boeing 707 airliner. The Soviets had their own version, earlier and less capable, on a different airliner. The British were working on yet another version, on a Nimrod maritime patrol aircraft, with the specific accumulation of problems that British defence programmes of the 1970s had a tendency to accumulate.

What was in Hangar Seven was none of these things. It was the Akashganga.

The name had been chosen by Dr. Siddharth Rawat, the lead engineer for the Indian radar programme, at the very first joint meeting of the Indian and Israeli teams in June 1973. Rawat had offered it quietly, almost as an aside, after several other names had been proposed and found inadequate. Akashganga — Sanskrit for the Milky Way, literally the River of the Sky. The great band of stars that swept across the Indian night, that ancient people had imagined as a celestial river flowing overhead, containing and connecting everything. An eye that swept the sky the way the Ganges swept the plain — continuously, patiently, finding everything within its reach.

Dr. Yitzhak Peled, the Israeli radar engineer leading the programme from the ELTA Systems side, had noted that in Hebrew the Milky Way carried a similar poetic name — HaDerech HaHalavit, the Milky Road, or in ancient Mishnaic Hebrew, nahar di-nur, the River of Fire. Two civilisations, separated by thousands of kilometres and millennia of separate history, had looked at the same band of stars and independently imagined a river. The name was adopted unanimously, which was unusual in a programme where unanimous adoption of anything had generally required three rounds of documentation and at least one formal disagreement resolved in writing.

Karan had arrived at the hangar at six-forty-five in the morning.

He had come from Gorakhnath Temple, where he had performed pradakshina — the circumambulation of the shrine, the walking meditation that was his morning practice whenever he was in Gorakhpur — but he had moved more quickly than usual. Not because the temple time was shortened, but because his mind had been awake since four-thirty with a question that could not be answered in a temple or an office or anywhere except in the air above eastern India with the rotodome turning and the radar active and the mission computers running.

The question was simple: does it work?

Eighteen months. Two nations. Four companies. Thousands of engineering hours. Arguments in three languages about things that had never been done before. Solutions to problems that had never been posed before. A rotating disc above an old airliner, full of electronics that were — if the tests and the models and the laboratory results were all correct — going to look out across the Indian sky and see everything.

Does it work?

He stood inside the hangar entrance and looked at the aircraft and let himself simply look at it for a moment before the day's programme consumed everything.

The choice of the Il-18 as the base aircraft had been the programme's first significant argument, back in the planning sessions of April and May 1973. The Israeli team — led by Peled and his IAI colleague Brigadier General (retired) Moshe Carmeli — had wanted a jet aircraft. Jets were faster, they argued. Jets could reach their operating altitude more quickly. Jets had a certain prestige. They had proposed several options: the Caravelle, a French short-to-medium range jet; the Fokker F28, a Dutch regional jet; even the Viscount turboprop, which was British.

To understand why the Il-18 won, it helps to understand what an aircraft like this actually needs to do. Unlike a fighter, which needs to go very fast for short periods, a flying radar platform needs to go to a high altitude — perhaps 25,000 to 30,000 feet, roughly five to six miles up, where the curvature of the earth and the absence of most weather gives the radar the clearest possible view outward — and then stay there for as long as possible. Not rushing. Not manoeuvring. Turning in slow, lazy circles or flying a steady patrol pattern, covering its assigned area of sky for eight, ten, twelve hours at a stretch. This was why endurance mattered more than speed.

The Il-18's four turboprop engines — a turboprop uses a jet engine to turn a propeller, combining the efficiency of a propeller with the reliability of a jet — consumed significantly less fuel than pure jet engines at the patrol speeds and altitudes required. The Il-18 could stay airborne for twelve hours on its internal fuel. Most jets they were considering could not match this without external tanks, which added complexity and drag. And crucially, India already owned Il-18 airframes. Indian Airlines had been flying them. The IAF had examples in its transport fleet. Spare parts were available. Maintenance personnel who knew the aircraft existed.

Carmeli had been the one who settled the argument. He was looking at the wing loading calculations — a comparison of the lift a wing generates against the weight it must carry — and had said, with the flat finality of a man who has found his answer in the numbers rather than in the debate: The turboprop gives us better airtime per kilogram of fuel at the altitudes we need. If the mission is surveillance rather than intercept, this is the correct platform. He had looked at Peled, said something in Hebrew, and the argument was over.

The specific Il-18 in Hangar Seven had been in Indian Airlines service until 1971, retired after a hydraulic system failure — the hydraulic system being the network of fluid-pressure lines that control the aircraft's landing gear, flaps, and other moving parts — that was more economically addressed by retirement than repair. Shergill Industries purchased it from Indian Airlines at a price appropriate to its condition, had it repaired to airworthy standard for a single ferry flight, and flew it to Gorakhpur in September 1972. It had been in the hangar since.

The transformation between September 1972 and March 1974 was substantial enough that the aircraft in the hangar bore only a family resemblance to what had arrived. The fuselage had been stripped to bare metal and structurally reinforced at fourteen specific points — chosen by Dr. Padma Krishnan's structural engineering team at Shergill Aviation, who had done the same calculations for the S-27 and S-35 airframes and whose word on what an airframe could and could not bear carried the weight of demonstrated expertise. The passenger windows had been removed: on the starboard side replaced with smooth aluminium panels flush with the fuselage skin, on the port side replaced with a series of apertures for antennas and observation equipment whose exact positions had required six months of negotiation between the Israeli radar engineers and Krishnan's structures group, each side wanting slightly different things from the fuselage skin and both sides having very good reasons for what they wanted.

The wings had been strengthened at their attachment points — the joints where wing meets fuselage — to handle the additional weight and aerodynamic forces generated by the rotodome mounted above. The rotodome was not just heavy, though at over two tonnes it was certainly that. It was also a disc sitting six metres above the aircraft's centreline, which meant that in any kind of banking turn or turbulence the disc wanted to swing out in one direction while the aircraft swung in another. This created what engineers called a moment — a rotating force applied to the airframe. Managing that moment without the aircraft's structure failing was one of the more demanding engineering problems of the whole programme.

Krishnan had found the solution at eleven-thirty on a Tuesday evening in January 1974 and had called Peled at his hotel immediately. The solution involved a specific pattern of reinforcing ribs inside the fuselage along the rotodome's attachment struts, combined with a revised strut geometry that redistributed the moment forces more evenly across the airframe. Peled had said: I will review it in the morning. He had reviewed it at six-thirty and called back at seven: It works.

Wing Commander Prabhat Anand arrived at seven-twenty, which was the schedule, and said good morning to Karan with the compressed quality of a man who had slept adequately and was entirely focused on the next six hours. Anand was the test pilot who had been embedded in the Akashganga programme since November 1973, and he carried the specific alert quality of someone who had been preparing for today for months and had chosen to experience that preparation as readiness rather than anxiety. He had a copy of the demonstration profile in his hand.

"Two concerns before the briefing," he said.

"Go ahead," Karan said.

"The low-level scenario. The demonstration profile specifies one target aircraft at low altitude — three hundred feet AGL." AGL meant above ground level, the height of the aircraft above the actual ground beneath it, as distinct from altitude above sea level. Three hundred feet AGL was very low — lower than most of Gorakhpur's buildings were tall. It was the kind of altitude that a cruise missile or a fast jet conducting a surprise attack might fly, skimming below radar coverage by staying close to the earth. "I want to run two simultaneous targets at low altitude, different headings. The radar can track both simultaneously. I want to show the delegation that it actually does."

"The computer load?" Karan said. The mission computer — the airborne digital brain that processed all the radar data — had limits on how much it could handle at once, like any computer.

"Rawat and I checked it yesterday. At two low-level targets plus the normal high-altitude traffic, the computer is running at seventy-one percent of peak capacity. We have headroom."

"Approved," Karan said. "The second concern."

"The EW demonstration — the passive detection test." The EW suite, the Electronic Warfare suite, was the system that listened for the radar signals of other aircraft and ships. Unlike the main radar — which sent out radio waves and listened for their reflections — the EW suite only listened, passively, without transmitting anything itself. The profile called for demonstrating it against one radar station on the ground. "I want to use two stations simultaneously. The radar at Allahabad and the IAF station at Varanasi. Tadiran's EW system can handle both. The profile doesn't ask it to."

Karan looked at Peled, who had been standing nearby and listening.

"The signal processing for two simultaneous EW tracks," Peled said. "Tadiran's team ran it on the bench with the ISMC processor boards last week. No problems. The air test is the next step and today is the air test." He looked at Anand. "Use the two stations."

Anand wrote it in his notebook.

The IAF — the Indian Air Force — delegation arrived at eight o'clock with the unhurried precision of people who had been told to arrive at eight and had treated this as a serious instruction.

Air Vice Marshal Denzil Keelor led it. The same Keelor who had been at the S-35 presentation in Vizag. The same Keelor who had shot down a Pakistani Sabre over Halwara in 1965 in twelve seconds. Today he was not here as a pilot — or not primarily as one — but as the IAF's senior representative to the programme, with the authority to assess what he saw and recommend to Air Headquarters whether the IAF should formally commit to this system as its future airborne radar platform.

With him: Air Commodore Vijay Gowda from Air Staff Requirements, the branch of the IAF responsible for deciding what equipment the service needed and writing the specifications for it. Gowda's job put him permanently in the productive discomfort of the gap between what equipment could do and what doctrine required it to do — the gap between the engineer's answer and the commander's need. He had a thick folder of questions. Wing Commander Rajan Mahajan was the IAF's signals and electronics specialist, a man whose function on this delegation was to understand the radar at its technical level — not to fly it, not to command it, but to evaluate whether it was what it claimed to be. He had a thinner folder than Gowda but each item in it was more precisely targeted. Squadron Leader Deepak Nair was the IAF's air traffic control specialist — a man who spent his professional life managing the complex picture of aircraft moving through shared airspace, coordinating their movements, and ensuring they did not collide or interfere with each other. He had been selected for this delegation specifically because the Akashganga's operational role would require managing a complex air picture and directing friendly aircraft based on it, which was not entirely unlike the ATC function he already performed, except the environment would be a combat zone rather than a peacetime airspace, the aircraft would be carrying weapons rather than passengers, and the consequences of error were of a different order entirely.

Nair was thirty-four years old, lean, with the quiet quality of someone who had spent his career in a job that required absolute accuracy — because the difference between a correctly issued ATC instruction and an incorrect one was sometimes measured in aircraft and human lives. He carried a notebook and a pencil. He had already read the programme documents and circled, in the radar performance specification, the line that claimed the system could simultaneously track two hundred and fifty airborne targets. Beside it he had written: at what minimum target size? false alarm rate? degradation in clutter? These were not hostile questions. They were the questions of someone who understood the gap between a specification claim and an operational reality.

The two defence ministry observers — senior civil servants from the Ministry of Defence, whose presence was not entirely comfortable for anyone in the room but who were there because programmes at this stage required witness from people with the authority to approve what came next — settled into chairs at the back and opened their own notebooks. They had the manner of men attending a recital: interested, evaluating, but conscious that their role was to observe rather than to direct the music.

The technical briefing began at eight-thirty in the hangar's briefing room — a converted workshop space with a projector, a screen, and enough chairs for the forty-two people who needed to be in it. Karan sat in the third row, which was the row he always sat in at technical briefings: close enough to see the details, far enough to see the room.

Rawat opened.

Dr. Siddharth Rawat had been building radar systems since the early days of the Netra programme — the multi-mode radar fitted in the S-27 Pinaka, which was itself the foundation of the Indian side's contribution to Akashganga. He was a compact man who wore glasses and had the quality of someone for whom the world's most interesting features were invisible to the naked eye — expressed in radio waves and Doppler shifts and signal-to-noise ratios. When he talked about radar, he was not explaining something he had learned. He was describing something he inhabited.

"Let me begin," Rawat said, "with the question the whole programme rests on. Why do you need a flying radar?"

He paused.

"In the simple answer: because the earth is round."

He clicked to a diagram — a cross-section of the earth's curvature, with a radar on the ground at one edge and an aircraft flying low over the horizon.

"A radar on the ground is powerful. A radar on the ground at a major air base can see hundreds of kilometres in every direction. But it cannot see below the horizon. And because the earth is curved, anything flying low enough — close enough to the surface — disappears below the radar's view before it gets close enough for the ground radar to detect it. A fast jet at sea level, two hundred kilometres away, may be completely invisible to the ground radar. A cruise missile at fifty feet altitude can approach undetected."

He clicked to the next diagram. The same cross-section, but now the radar was not on the ground — it was in an aircraft flying at 25,000 feet.

"An aircraft at 25,000 feet — roughly five miles up — has a radar horizon that extends approximately 480 kilometres to the surface. It can look down and forward and to the sides at everything in an enormous area, including the things that are too low for any ground radar to see. The curvature of the earth that hides targets from ground radar is no obstacle to a radar looking down from above."

He paused and let this settle.

"This is the fundamental value of what the Akashganga does. It fills the gap. The ground radars watch the medium and high altitudes. Akashganga watches everything — high, medium, and low. Together they give the commander a complete picture of what is happening in the sky."

Gowda: "How far can the Akashganga actually see?"

"At high altitude — against an aircraft flying above 10,000 feet — the detection range is 440 kilometres. The entire subcontinent can be covered by three or four aircraft operating simultaneously. Against a target at very low altitude — 300 feet above the ground — the detection range reduces to approximately 280 kilometres, because the geometry becomes less favourable and the background clutter from the ground interferes with the radar return. We will discuss this more in a moment."

He clicked to the rotodome rotation diagram.

"The radar antenna inside the rotodome rotates at six revolutions per minute. One full circle every ten seconds. This is important to understand because it shapes everything else about how the system works."

He looked at the room.

"The radar can only look in one direction at a time. But it needs to see in all directions. The solution is to rotate the antenna and build the complete 360-degree picture from sequential slices — the radar looks north, then northeast, then east, continuing around the compass, completing the full circle in ten seconds. Every ten seconds the picture is refreshed. Every target in the coverage area gets a new radar measurement once every ten seconds."

Gowda: "That means a fast jet could move considerably between measurements."

"Correct. In ten seconds, an aircraft flying at 900 kilometres per hour moves two and a half kilometres. If we simply plotted each new measurement as a dot, the display would show a series of dots jumping around the sky rather than a smooth track. The operator would see chaos rather than a picture."

He clicked to the tracking algorithm diagram.

"The solution is the tracking software. The mission computer — the aircraft's airborne digital brain — doesn't just display raw measurements. It maintains a continuous track for every target it has seen. Think of a track as the radar's best current knowledge of where a target is, where it is going, and how fast it is moving. Between measurements, the computer predicts where the target will be based on its last known velocity and heading. When the next measurement arrives, the computer updates the prediction with the actual measurement and refines the track accordingly."

He looked at the room.

"What the operator sees on their display is not raw measurements. It is a smooth, continuously updated picture of every tracked object in the coverage area, with the computer filling in the gaps between ten-second measurements using physics — the simple fact that an aircraft going north at 500 knots will be, ten seconds later, approximately 2.5 kilometres further north."

Mahajan: "And if the target turns?"

"The computer detects the deviation from the predicted position on the next measurement. If the target turned right, its actual position is to the right of where the prediction said it would be. The computer notes the deviation and updates the track — recalculates the heading and the velocity based on the new position. Within one or two measurement cycles after a turn, the track is accurate again."

"The delay is unavoidable," Rawat added. "If a target turns sharply and accelerates simultaneously, the track may be momentarily confused — the computer has to update its understanding of what the target is doing, and this takes one or two measurement cycles. During those five to twenty seconds, the track quality degrades. We flag degraded tracks for the operator. But this is a fundamental limitation of any rotating radar: the update rate is determined by the rotation speed, and ten seconds is as fast as the physical mechanism can practically rotate a six-metre disc."

This was the kind of honest limitation that made technically educated people in the room sit up slightly — the acknowledgment that the system had a real constraint, stated plainly, without wrapping it in qualification. Keelor made a note.

The ground clutter question came next, and it was the technically deepest part of the morning's briefing. To understand it requires understanding how radar actually works when looking down.

When the Akashganga's radar sends out its radio pulses and they hit the ground, the ground reflects them back. This reflected energy — called ground clutter — arrives at the antenna mixed in with the reflections from actual aircraft. Distinguishing an aircraft from the ground return beneath it was, in the early 1970s, one of the most technically demanding problems in radar engineering.

The solution used by sophisticated radars was called pulse-Doppler processing. The Doppler effect — the same physics that makes a passing train's horn seem to change pitch — applies to radio waves as well as sound. Radio waves bounced off a stationary object return at the same frequency they were transmitted. Radio waves bounced off a moving object return at a slightly shifted frequency — higher frequency if the object is moving toward the radar, lower if moving away. By measuring this frequency shift, the radar can determine the speed of everything it sees. Stationary ground has zero Doppler shift. Moving aircraft have significant Doppler shift. The radar can therefore separate moving returns — aircraft — from stationary returns — ground. This is the principle.

The problem is that at very low altitude, the geometry becomes difficult. A fast jet at 300 feet is not very far above the trees, buildings, and terrain beneath it. The radar beam, looking down at the jet, also illuminates a great deal of the ground near the jet. Some of that ground — moving water, trees swaying in wind, vehicles on roads — also has small Doppler shifts. And the Doppler shift from the jet, viewed from a specific angle, might be smaller than expected if the jet is moving partially across the beam rather than directly toward it.

This overlap between the aircraft's Doppler signature and the clutter's Doppler signature was where conventional radar systems of the era struggled. Their signal processing — the electronics that translated raw radar returns into useful information — was not fast enough or fine enough to separate signals that were close together in Doppler frequency.

This was where ISMC came in.

The Indian Semiconductor Manufacturing Corporation's Gorakhpur facility — Karan's semiconductor division, which had been producing chips since 1971 for the aerospace and defence programmes — had developed what the engineers called 3-millimetre processing. The 3 millimetres referred not to physical size but to the scale of the semiconductor devices being made: transistors and circuit elements fabricated at dimensions an order of magnitude smaller than conventional electronics of the era. These tiny devices operated much faster — they could switch on and off at a rate far exceeding what larger conventional semiconductors could achieve.

For radar signal processing, speed meant resolution. The finer the semiconductor device, the faster the electronic circuits, and the more finely the Doppler spectrum could be divided into measurement channels. With ISMC's 3-millimetre devices, the radar's signal processing could divide the Doppler spectrum into frequency bins approximately forty times narrower than those achievable with conventional mid-1970s semiconductor technology.

Forty times narrower meant forty times finer discrimination. Where a conventional radar system would see two signals blurred together in the same broad frequency channel — unable to tell the aircraft from the clutter — the ISMC-powered system would see them in separate narrow channels, cleanly separated, distinguishable.

Arvind Sethuraman — twenty-eight years old, ISMC's computing team leader, a man who still seemed slightly surprised by what he had helped build — stood to explain the practical result.

"Against ground clutter with conventional processing," he said, "the minimum detectable speed of a target is approximately 70 knots. Below that speed, the target's Doppler signature falls within the clutter region and the radar cannot reliably separate it from the ground return." Seventy knots is roughly 130 kilometres per hour — fast for a car, slow for an aircraft. "With ISMC 3-millimetre processing, the minimum detectable speed drops to approximately 22 knots. Forty kilometres per hour."

He paused.

"Consider a military helicopter operating in search mode, flying slowly over the terrain. At 70 knots minimum detectable speed, a helicopter at 30 knots is invisible to the radar. It is indistinguishable from the background. At 22 knots minimum detectable speed, we see it. We track it. We can direct a response."

Keelor, quietly: "And a hovering helicopter? Zero speed."

"A hovering helicopter has zero translational velocity — it is not moving across the ground. But its rotor blades are rotating rapidly, and those rotating blades each have a velocity relative to the radar. As each blade sweeps toward the radar and then away, it generates a distinctive pattern of Doppler returns — high positive, then high negative, cycling with the rotor speed. This pattern — called the blade flash signature — is quite different from any ground clutter pattern and appears clearly in the processed signal. The system detects hovering helicopters through their rotor signature rather than through their body movement."

Keelor made a note. The note, from the quality of the writing, was going into a mental file labelled something like: Pakistani helicopter operations — this matters.

Peled stood for the EW suite briefing.

Electronic Warfare — EW — is one of those terms that sounds technical and mysterious but describes something conceptually straightforward: using the radio-frequency electromagnetic spectrum as a battlefield. Every radar, every communications transmitter, every electronic system that sends out radio waves is both a capability and a vulnerability — a capability because it performs its function, and a vulnerability because its transmissions can be detected, located, and characterised by anyone with the right equipment who happens to be listening.

The Akashganga's EW suite did three things with this principle.

First: it listened. Constantly, across a wide range of frequencies, for any radio-frequency transmission in the surrounding airspace. A warship's search radar transmitting. A fighter's airborne radar scanning. An early-warning station on the ground sending out its pulses. Any of these, and many others, would be detected by the EW suite even at ranges far beyond what the Akashganga's own active radar could see. Think of it like the difference between seeing and hearing — you can hear a sound from much further away than you can see its source. The active radar was sight: powerful but limited by the physics of the radar horizon. The EW suite was hearing: passive, with no transmissions of its own to reveal its presence, extending the Akashganga's awareness far beyond the visual range of its radar.

"The detection range of the RWGS — the Radar Warning and Geolocation System — against a typical airborne radar operating at standard power is over 600 kilometres," Peled said. The RWGS was the first component of the EW suite, the one responsible for detecting and locating emitting radars. "Six hundred kilometres. The Akashganga flying at 25,000 feet over the Bay of Bengal can detect a radar-equipped aircraft approaching from Pakistan before that aircraft has crossed the Indian border. We have awareness at distances where we have no direct capability to respond — but awareness is the foundation of response. You cannot intercept a threat you do not know exists."

He paused.

"Compare this to the active radar. The active radar detects everything — transmitting or not — but only to 440 kilometres at altitude. The RWGS detects only transmitting emitters — only things that have their radar or communications switched on — but to 600 kilometres and beyond. The two systems together cover the full threat environment: things that are transmitting, seen by both active radar and RWGS at long range; things that have gone emissions-silent, seen only by the active radar at shorter range."

Gowda: "An aircraft that turns off all its electronics — radar silent, communications silent. What do we see?"

"The active radar still sees it — it is a physical object, it reflects radio waves regardless of whether it is transmitting anything. We see the radar return. We do not know from the radar return alone what type of aircraft it is. The RWGS tells us the type — because the RWGS knows what radar signature that aircraft type carries when it is transmitting, and we can recognise the type from its emissions history. But a truly silent aircraft, observed only by radar return, appears as an unknown contact."

"Then the correct adversary tactic," Gowda said, "is to approach in complete emissions silence."

"Yes," Peled said. "This is a known limitation and a known adversary tactic. The system is more powerful against adversaries who are operating their radars than against those who have gone silent. Against a completely silent adversary, the system provides detection but not identification."

The second capability of the EW suite was geolocation — using the detected signals not just to know that an emitter existed, but to calculate where it was. The physics of this relied on timing. The EW suite's antennas were placed at multiple positions around the aircraft — forward fuselage, aft fuselage, wingtips. A signal from a distant emitter would arrive at each antenna at a slightly different moment, because each antenna was at a slightly different distance from the source. These tiny timing differences — measured in millionths of a second — encoded geometric information about the direction to the emitter. By computing these differences across multiple antennas and combining the results with the aircraft's known position and movement, the RWGS could calculate the emitter's position.

"Position accuracy at 400 kilometres: within three kilometres," Peled said. "Accurate enough to vector strike aircraft to a search area containing the emitter. Accurate enough to build a picture of the opposing force's radar deployment — where their ground-based air defence radars are, where their picket ships are positioned — from a safe distance, before any hostile action has been taken, simply by listening."

The third component of the EW suite was COMINT — Communications Intelligence. If RWGS listened for radar signals, COMINT listened for communications signals: voice radio between aircraft and ground controllers, data transmissions between ships and headquarters, the radio traffic through which a military force coordinated itself. COMINT did not need to understand the content of intercepted communications to extract value. The frequency used, the direction the transmission came from, the pattern of who was transmitting when — all of this built a picture of what the opposing force was doing, where its assets were, how its command structure was organised.

Peled handled this section briefly, because the COMINT capability was the most politically sensitive of the three — its implications for peacetime intelligence collection were not lost on anyone in the room — and because the demonstration of it during the flight would be limited for the same reason. But its existence, and the fact that it was integrated into the same aircraft as the radar and the RWGS, meant that the Akashganga's operators would be looking at a battlefield picture that combined radar surveillance, radar emitter intelligence, and communications intelligence simultaneously, on a single display.

The defence ministry observer in the back row wrote something in his notebook. Whatever he wrote, it was long.

Arvind Sethuraman and Ruth Ben-David — the latter representing Tadiran's mission computer and communications programme — presented the operator interface section together, alternating between them with the smooth efficiency of people who had done this presentation before.

The Akashganga carried six operator stations in the main cabin. To understand what these stations were for, it helps to understand what information they were handling and what decisions had to be made with it.

Picture the following situation. The Akashganga is flying its patrol orbit — a slow oval pattern at 25,000 feet above the Bay of Bengal. Its active radar is tracking 180 objects simultaneously: commercial airliners on their scheduled routes, Indian Air Force transport flights, patrol aircraft, and several unknown contacts whose speed and altitude suggest they are military. Its RWGS has detected five radar emitters and located four of them — one is a Pakistani Navy ship 380 kilometres to the west, one is an unknown airborne radar 290 kilometres to the northwest, and two are Indian ground stations positively identified. Its COMINT receiver is monitoring three active communications channels.

All of this information is arriving simultaneously. It changes every ten seconds with each radar sweep. The unknown contact northwest is now heading east and accelerating. The Pakistani ship has changed its radar frequency. One of the communications channels has gone silent. A new radar contact has appeared at low altitude to the south.

Six people are responsible for making sense of this. Each one manages a portion of the picture. The two radar picture operators maintain the overall track database — they manage the list of all tracked contacts, update track statuses, and flag anything that needs the mission commander's attention. The RWGS operator manages the EW picture — tracking emitter locations, updating the identification database, alerting when a new emitter appears or a known one changes behaviour. The COMINT operator manages the communications monitoring. The fighter control officer is the link between the Akashganga and the friendly fighters it is directing — they speak directly to the fighter pilots, giving them vectors toward targets, keeping them informed of the tactical picture, coordinating their movements so they do not interfere with each other. And the mission commander oversees all of this, making the decisions that require human judgment: designating a contact as hostile, authorising a fighter to close on an unknown, determining when to alert the ground command post.

Each of these six stations had a fourteen-inch display screen — the largest practical CRT screen that 1974 technology could fit in the space and weight available, roughly the size of a large television set of the era — plus a keyboard for data entry and a track-ball for moving a cursor around the display. The displays showed synthetic pictures: clean, processed information rendered as symbols and data blocks on a map background, rather than the raw, noisy radar returns that only a trained radar operator could interpret.

Sethuraman explained the display format at length. Each tracked aircraft appeared as a small pointed symbol indicating its current position and heading. Beside the symbol, a data block showed its altitude, speed, and track number. The system automatically identified any aircraft whose transponder code — the electronic identification signal that commercial aircraft are required to transmit — matched the known friendly list. These appeared in one colour. Unknown contacts appeared in another. Contacts designated as potentially hostile by the mission commander appeared in a third.

Squadron Leader Nair had been listening through all of this with the quiet intensity of a man building a mental model. He raised his hand when Sethuraman reached the display format section.

"The data block shows altitude, speed, heading, track number, identification status," Nair said. "Does it show anything derived from those figures? Time to reach a specified point, for instance?"

Sethuraman paused. "Not in the current format."

"It should," Nair said. He said it without accusation, in the flat tone of someone identifying a gap rather than criticising a failure. "In ATC — air traffic control — we work with derived metrics constantly. Time to closest point of approach between two aircraft. Predicted conflict point. Descent rate needed to reach a cleared altitude. These are not things the controller computes in their head while simultaneously managing six other communications. The system computes them and displays them. The controller sees the implication directly rather than deriving it from the raw data."

He looked at Sethuraman.

"An operator on this aircraft is managing potentially dozens of tracks simultaneously. The cognitive load of computing time-to-closest-approach from raw speed and position data — doing that calculation mentally, for each relevant pair of tracks, while simultaneously monitoring all the others — is unreasonable. The system has all the data needed to perform that computation. It should perform it and display the result."

Sethuraman looked at Rawat. Rawat wrote something in his notebook.

"That is an accurate analysis," Rawat said. "The capability exists in the mission computer. Adding the display element is a software modification. I will specify it."

Nair nodded and wrote something in his own notebook.

The interaction produced a quality of silence in the room that was different from the normal silence of a briefing. It was the silence of people recognising that a good question had identified a real gap — and more specifically, recognising that the gap existed because the people who built the system had not had the right kind of expertise in the room when they were designing the operator interface. Aviation engineers built what they understood. An ATC specialist saw it from a different angle.

At ten-thirty, Anand and his crew boarded the aircraft.

The crew that walked across the hangar floor in the particular unhurried way of people who are about to do demanding work and have decided that urgency is not the same as readiness consisted of: Anand and his co-pilot Wing Commander Suresh Tiwari in the flight deck; the flight engineer, responsible for monitoring the aircraft's systems; the navigator; and in the main cabin, six mission operators — three from the Indian side of the programme and three from ELTA's team, who had been training together on the ground simulator for the previous six weeks. Among the mission operators was Flight Lieutenant Bhaskar, an IAF signals officer who had been embedded in the programme since August 1973 and who had spent so many hours on the ground simulator that he could, by his own account, operate the radar picture station in his sleep. He had not tested this claim.

Keelor boarded as an observer on the flight deck jump seat. Peled took a position in the forward cabin with a radio connection to the main mission operator area. Both would observe rather than operate — today was a demonstration, not a crew evaluation.

The aircraft taxied out at ten-fifty-eight. The rotodome, which had been stationary in the hangar, would begin rotating once the aircraft was airborne and the radar was activated. From outside, watching from the apron, the sight of the rotodome beginning to turn as the aircraft climbed away was something that Carmeli noted afterward as the moment the whole programme became real to him in a physical way it had not been in all the months of engineering documents and laboratory tests. The disc turning above the climbing aircraft, catching the morning light, rotating steadily — six RPM, one full revolution every ten seconds — had a quality of purpose that abstractions and schematics did not carry.

The aircraft took off at eleven-oh-three. The demonstration profile covered, in sequence: detection and tracking of commercial aircraft at high altitude, using the known schedules of Indian Airlines flights as ground truth to check accuracy; detection and tracking of two military aircraft at very low altitude; passive detection and location of two ground-based radar stations simultaneously; and the communications intelligence demonstration. The ground control station — a converted trailer parked on the apron, filled with displays connected to the aircraft by the data link — would show the delegations watching from the ground the same picture the crew was seeing in the air.

The ground control station was more crowded than its designers had intended. Thirty-seven people — the programme teams, the IAF delegation, the ministry observers — occupied a space built for twenty comfortable viewers. Nobody was thinking about comfort.

The data link — a frequency-hopping UHF radio connection that transmitted the compressed mission data from the aircraft to the ground at the speed of radio waves — was the technology that made the ground station possible. Frequency-hopping meant the transmitter and receiver were synchronised to jump between radio frequencies many times per second, making the transmission extremely difficult to jam or intercept, because an adversary who found one frequency would lose the signal almost immediately as it hopped to the next. The cost of this security was a slight transmission delay: the ground station's picture was approximately four seconds behind the aircraft's own displays. Four seconds was the time the data compression and transmission required. In most operational contexts, four seconds was acceptable. It would not be used for situations requiring split-second reaction, but command oversight and mission management did not require split-second reaction — they required accurate situational awareness, and four seconds old was still very current.

The first radar returns appeared on the display at eleven-seventeen.

They appeared first as blips — single points of light on the display, raw detections without any further information attached. The radar had found something but had not yet determined what. Within the next two rotations — twenty seconds — the mission computer had associated the blips with tracks: each contact now had a small symbol indicating its position and heading, and a data block beginning to populate with altitude, speed, and identification information.

Commercial aircraft. Indian Airlines flights on the Delhi-Calcutta corridor, the Bombay-Patna route, a freight aircraft on the Lucknow-Madras leg. The radar was finding exactly what the airspace information predicted it would find, in approximately the positions where it would be expected to be at that time.

But finding them was only the beginning of the verification. The question was not whether the radar detected them — of course it would detect an airliner, which was a large, slow-moving object specifically designed to be navigated accurately. The question was whether the radar's measurements matched the reality.

Rawat was on the radio to the Varanasi ATC centre, which had radar surveillance of its own and was tracking the same commercial flights with its ground-based equipment and with the Mode C transponder data — the altitude information that commercial aircraft automatically broadcast. This transponder data, combined with the ATC radar range-and-bearing measurement, gave a ground truth for each aircraft: its actual altitude, actual speed, actual heading, as close to the truth as current measurement technology provided.

The correlation for Indian Airlines Flight 147 — a Delhi-Calcutta service, at 24,000 feet — was: altitude difference 180 feet, speed difference seven knots, heading difference less than one degree.

One hundred and eighty feet altitude error. Seven knots speed error. Less than one degree of heading error.

The specification required accuracy better than 500 feet in altitude, better than fifteen knots in speed, better than two degrees in heading. The actual performance was significantly better than the specification on all three metrics. This happened when a system was built by people who treated the specification as a minimum rather than a ceiling.

Mahajan, looking at the numbers: "The track quality at the edge of the coverage area. You said the detection range was 440 kilometres. What does the track quality look like at 430 kilometres?"

Rawat spoke to the aircraft, received a response, and said: "The tracks at the edge of the coverage area have a track quality flag — the system marks them as marginal. The update interval becomes irregular: sometimes one missed update in three, sometimes one missed in two. The position accuracy degrades roughly proportionally to the range — at 430 kilometres, the position error is approximately twice what it is at 220 kilometres. The system shows the operator which tracks are marginal and which are reliable."

"If I am directing an intercept based on a marginal-quality track," Mahajan said, "what wider search area do I need to give my fighter?"

"The track position error at maximum range with marginal quality is approximately five kilometres," Rawat said. "I would recommend directing the fighter to a twenty-kilometre search area centred on the track position, and instructing the fighter's own radar to scan that area broadly until it acquires its own track on the target."

Mahajan wrote. The answer was not perfect — a five-kilometre positional error was a significant uncertainty for an intercept. But it was an honest answer, and it described a situation at the system's extreme range in the worst-case quality conditions. At normal operational ranges, the accuracy was considerably better.

The low-altitude demonstration began at eleven-forty-two.

In the air, Anand had coordinated with two IAF aircraft that had been prearranged for the demonstration. A MiG-21 from the Gorakhpur fighter squadron was flying a low-level ingress profile — an approach simulating a surprise attack run: 300 feet above the ground, 480 knots — coming from the northeast. An old Hawker Hunter from the training unit at Sulur, slower and with a different radar cross-section profile, was flying at 500 feet from the east. Different altitudes, different speeds, different directions, different aircraft types.

The word cross-section in radar engineering refers to how much of a radar beam an object reflects back. It is not literally a cross-section of the physical object but rather a measure of its radar reflectivity. A large commercial airliner has a very large radar cross-section — it returns a strong signal. A small fighter jet has a smaller radar cross-section. The cross-section can also vary with the angle from which the radar is looking at the aircraft. The Hunter and the MiG-21 had meaningfully different cross-sections from each other, which made them useful for demonstrating the system's ability to track different types of targets simultaneously.

On the ground display, the coverage circle showed the border of the Akashganga's radar reach. The two aircraft were approaching that boundary, not yet inside it.

At eleven-forty-four, the MiG crossed into coverage.

The display showed nothing.

Thirty-seven people looking at the display. Nothing.

Two seconds.

"Track acquisition in progress," Bhaskar's voice from the aircraft over the radio. "Processing low-level clutter rejection." The clutter rejection — the computational process of separating the aircraft's return from the background noise of the ground — was running, filtering, working.

Four seconds. Still nothing visible.

Then the track appeared.

Not a tentative hint — a clean, full track symbol with a complete data block. Altitude: 290 feet AGL. Speed: 482 knots. Heading: consistent with the northeast ingress vector.

The ground station was quiet in the specific way of rooms where thirty-seven people have all stopped breathing simultaneously.

Eight seconds after the MiG appeared, the Hunter crossed the coverage boundary. It appeared in four seconds — slightly faster than the MiG, because its higher altitude put it slightly above the densest ground clutter layer.

Two targets. Low altitude. Simultaneously tracked.

Rawat checked the correlation against the pilots' own instruments. MiG: 210-foot altitude error, eleven-knot speed error. Hunter: 290-foot altitude error, fourteen-knot speed error.

Worse than the commercial aircraft at high altitude — expected, because the clutter environment at low altitude degraded measurement accuracy. But both within specification. Both tracked simultaneously.

Mahajan: "At those error levels — 210 feet altitude error at 300 feet AGL — the radar could place the MiG at 90 feet above the ground. In some terrain that is below tree-top level."

"This is true," Rawat said. "The altitude accuracy at very low altitude with the current system is not precise enough to distinguish a target at 300 feet from one at 90 feet. What we can say with confidence is that the target is low — below 600 feet. The classification is low-level threat, not a precise altitude measurement."

Nair spoke: "Should the system alert the operator when a tracked contact is this low? If the altitude error band means the target could be at ground level, the operator needs to know this is a special case, not a normal track."

Rawat looked at him. This was not in the demonstration script. This was the same kind of insight as the time-to-closest-approach question in the briefing — an operator-perspective question about what the human needed to see, not just what the system was technically providing.

"A terrain proximity alert," Rawat said. "The system knows the terrain elevation from its pre-loaded terrain database. Track altitude minus terrain elevation gives terrain clearance. If clearance is below a threshold — say, 500 feet — the track is flagged."

"Yes," Nair said. "That is what I mean."

Rawat wrote it in his notebook.

The two items — time-to-closest-approach and terrain proximity alert — were now items one and two on the modification list. The system had worked exactly as specified. The specifications, it turned out, had not fully captured what the operators needed. These were different problems.

The RWGS demonstration began at twelve-fifteen.

The aircraft was now over the eastern Gangetic plain, at altitude, with the active radar maintaining its normal track picture. On the ground station display, the picture changed. The active radar picture moved to the right side of the screen; the left side now showed a different kind of picture — not aircraft tracks, but a map of detected radar emitters and their locations.

The physics of the RWGS geolocation had been explained in the briefing. In practice, what it produced on the display was this: a small circle at the estimated position of each detected emitter, with a data block showing the radar type and the identification confidence. The circle, rather than a point, was important — it communicated the uncertainty honestly. The position was not precisely known. It was known to be within the circle. The size of the circle at a given range reflected the measurement accuracy.

The Allahabad ground radar appeared first — 380 kilometres from the aircraft. Its position circle had a radius of approximately four kilometres at the scale of the display. Beside it, the data block: frequency band, pulse pattern, scan rate. And the identification: P-12 Spoon Rest search radar, Soviet design, 94 percent confidence.

To explain the 94 percent: the RWGS maintained a library of radar signatures — a catalogue of the electronic fingerprints of 180 different radar types, built from intelligence gathering across multiple sources and updated periodically. When the system detected a radar, it compared the detected signal's characteristics against every entry in this library and found the closest match. Ninety-four percent confidence meant that 94 percent of the numerical similarity score pointed to the P-12, six percent to something else in the library or to a type not yet in it. It was not certainty. It was an informed probability assessment, displayed honestly.

The Varanasi IAF radar appeared eleven seconds later. Different frequency. Different pulse pattern. Different identification: RL-2 approach surveillance radar, British design, 96 percent confidence.

Two emitters. Different types. Different nations' designs. Simultaneously tracked and identified. Position circles displayed with explicit uncertainty indicators.

Keelor, from the aircraft, spoke over the radio.

"If I am an adversary carrier battle group," he said — thinking out loud in the way he sometimes did, particularly when he was genuinely working through an operational problem rather than asking a rhetorical question — "and I know the Akashganga exists and has this passive detection capability, the obvious defensive response is to operate in emissions silence. Turn off my radars. Transmit as little as possible."

A pause.

"What does the Akashganga see from a carrier battle group operating in emissions silence?"

Rawat: "The active radar still sees the ships. They are physical objects, they reflect radio waves, they appear as surface contacts on the radar picture regardless of whether they are transmitting anything. We see them — we know something is there. The RWGS tells us what they are by their emissions. Without emissions, we know something is there at the radar-detected position but we cannot identify it from the EW picture alone."

"So a carrier group that has gone completely silent appears as a cluster of radar contacts of unknown character."

"Correct. We know it is large — a carrier and its escorts make a distinctive pattern of surface contacts. We know its position and its course and speed from the radar returns. We do not know for certain from electronic intelligence that it is a carrier group rather than a civilian convoy, though the pattern of the contacts and the tactical behaviour would usually make this assessment clear."

"But we cannot put a weapons system onto it based solely on the radar picture," Gowda said, from the ground station. "The identification needs to meet a threshold before weapons can be employed."

"Correct," Rawat said. "The identification threshold for weapons employment is a doctrinal question, not a technical one. The system provides the best available information. The decision to act on that information rests with the commander."

"This is an honest answer," Keelor said again, with the quality of someone who appreciated it precisely because it was unusual.

The COMINT demonstration at thirteen-thirty was brief and its results were not displayed publicly. The test signal was transmitted by the prearranged ground station. The receiver detected it. The direction-finding calculation placed it within 1.8 degrees of the true bearing. The frequency measurement was accurate. The signal type was correctly characterised.

What was demonstrated was function. What the function implied was left, by mutual silent agreement, for conversations that would not happen in a converted trailer with thirty-seven people in it.

Gowda, very quietly, to Karan: "This changes things."

Karan said nothing, which was a form of agreement.

"We cannot discuss the specifics here."

"No," Karan said.

The aircraft landed at fifteen-forty-five. Four hours and forty-two minutes. The full demonstration profile was complete. Three issues had been noted during the flight — a brief spike in display update latency during the low-altitude scenario, a seven-second data link dropout when the aircraft was at maximum range from the ground station, and a vibration in the rotodome rotation mechanism at specific combinations of bank angle and rotation speed. None of these were surprises. All were on the known issues list. None were fundamental problems. All were fixable.

The post-flight debrief in the hangar had the quality of a room where people know something has been proven and are moving on to the question of what it means. The aircraft sat visible through the open hangar doors, its turboprops cooling, the rotodome now stationary after hours of turning.

Anand delivered the flight crew's assessment in the flat, precise language of a test pilot reporting.

"Flight deck integration complete and functional. Four-crew workload appropriate. The data link management function — the additional task of monitoring the link quality and data rate — is not covered by any existing IAF procedure for this aircraft type. New procedures are required. The navigator's workload increase from the mission area management function is also not covered. Both items go on the procedure development list."

He moved to the operator stations. The radar picture operators had found the track management interface — the process of maintaining the track database, flagging new contacts, resolving track conflicts when the computer associated one contact with another incorrectly — intuitive for operators who had previous radar experience, requiring approximately forty hours of simulator training for operators without it. The RWGS station was the most demanding: managing sixteen simultaneously tracked emitters with different confidence levels and different tactical priorities was a cognitive challenge that the training programme had not fully prepared people for. Anand's assessment was that the RWGS operator needed to be a dedicated specialist, not a general signals operator adapted from another role.

Flight Lieutenant Deepak Nair — the fighter control officer during the demonstration, not related to Squadron Leader Deepak Nair the ATC specialist, a coincidence of names that had caused mild confusion throughout the day — gave his assessment through Anand. The task of directing intercept aircraft from the Akashganga to targets was substantially different from any existing IAF function. It required simultaneous management of multiple radio communications channels, maintaining awareness of the full air picture, performing intercept geometry calculations, and coordinating fighter movements to avoid deconfliction conflicts. None of the IAF's existing training covered this. None of the existing doctrine described how it should be done. The function was real and the hardware supported it. The procedures, the training, and the doctrine to use it did not yet exist.

"How long to develop them?" Keelor said.

Flight Lieutenant Nair: "Six months for a first-draft doctrine document based on what we learned today and in the preceding months of ground operation. One year to have that doctrine tested in exercises and revised based on the exercise outcomes. Two years to have a trained cadre of fighter control officers sufficient for operational deployment."

"We start immediately," Keelor said. He said it without qualification or delay, which was Keelor's manner when he had reached a conclusion.

He looked at Karan.

"The system works," he said. "The capabilities demonstrated today are real. The limitations that were honestly documented during the demonstration are real and are addressable over time. The performance on the commercial aircraft tracks was significantly better than specification. The low-altitude simultaneous dual-track was the demonstration I was most uncertain about before today, and it performed as described. The RWGS geolocation was within specification at both test ranges with both emitters simultaneously."

He paused.

"I am going to recommend to Air Headquarters that the IAF formally join the Akashganga programme as an operational partner — committing to the crew training programme, the doctrine development process, and the operational trials that will be needed before the system can be declared combat-ready. I have two conditions."

"Tell me," Karan said.

"The terrain proximity alert and the time-to-closest-approach display metric that Squadron Leader Nair identified this morning. Both of these are operator safety requirements, not optional improvements. They must be incorporated in the software before the system is declared operational."

"They will be," Rawat said.

"Second condition: a full-fidelity ground simulator. A simulator that replicates all six operator stations with the same software as the operational aircraft. You cannot develop doctrine or train operators by removing the operational aircraft from its flying programme. The simulator is the training device that allows the doctrine to be developed and tested without tying up the aircraft."

"Six months from the existing software foundation," Sethuraman said. "If the specification is finalised within two weeks."

"Finalise it within two weeks," Keelor said. To Karan: "The two conditions are not negotiable."

"They were never going to be," Karan said.

Carmeli and Peled sat with Rawat and Karan in the small office at the back of the hangar after the main debrief had finished and the delegations had dispersed to their own conversations. The Israeli team's designated driver had been waiting for an hour, and the flight back to Tel Aviv was scheduled but flexible.

"Eighteen months," Carmeli said. It was the kind of statement that did not require a verb to be complete.

"The Indian preliminary work started before that," Rawat said. "The radar architecture, the ISMC semiconductor development — those go back to August 1972. The joint programme starting in June 1973 accelerated the integration because you brought the EW suite and the mission computer framework that we would have needed to develop independently."

"The semiconductor component," Peled said. He said it with the quality of something he had been thinking about for some time. "The 3-millimetre devices. We had not encountered this level of miniaturisation and switching speed at this production cost outside of American defence contractors. Not from any European vendor. Not from any other source."

"ISMC has been developing since 1971," Karan said. "The Gorakhpur facility was built for the aerospace and defence semiconductor requirements from the beginning. The Akashganga's signal processing was the most demanding application we had run through it."

Carmeli looked at the aircraft through the partition. The post-flight inspection was still in progress — ground crew working methodically through the checklist, the rotodome being inspected visually and by instrument, the engine nacelles examined.

"What the aircraft did today," Carmeli said, "is what we specified it to do. This is not always how it goes."

"No," Karan said. "It is not always how it goes."

"In my experience," Carmeli said, with the quality of someone consulting a long professional memory, "programmes succeed when the engineering teams understand the problem clearly before they design the solution, and when the programme management is honest about what the data shows rather than what it would be convenient for the data to show. Both were true here."

Rawat said nothing. He had the expression of someone who was glad to have the assessment confirmed but had already moved on in his mind to the twenty-seven-item modification list that was sitting in his notebook.

Peled: "The IAF evaluation. Keelor's recommendation to Air HQ. If it is accepted — and I believe it will be — what is the timeline to the bilateral production agreement?"

"The ministry intends April for the formalisation meeting," Karan said. "Both governments need to be at that meeting. It is not a commercial contract between two companies. It is a programme framework between two states."

"April gives us one month from today," Peled said.

"I know," Karan said. "The documents are being prepared."

Carmeli: "IAI's position is clear. We want the production programme. The technology demonstrated today — the integration of the ELTA radar, the Tadiran EW suite, and the ISMC signal processing — is something that IAI cannot replicate using only domestic Israeli industrial capacity. The ISMC component is irreplaceable in the near term. IAI's interest in the production programme is genuine."

"And IAI's interest in the programme extends to both countries' needs," Karan said. "India's requirement and whatever the Israeli Air Force evaluation produces."

"Subject to the IAF evaluation completing," Carmeli said. He said it carefully. "The IAF evaluation has not yet been formally initiated. If it produces a favourable result — and today's demonstration makes the case for initiating it — the specific configuration for Israeli service will need to be defined in a separate specification. Some elements will be common. Some will differ."

"We will manage the common and the different elements through the joint programme structure," Karan said. "The programme structure is designed to accommodate this."

Carmeli looked at him. He had been in defence programmes for thirty years and he understood what designed to accommodate this usually meant in practice versus what it meant in theory. He also understood that the man across the table had built an aircraft from nothing in eighteen months and had demonstrated it to specification today, which was a form of evidence about how he managed programmes that was more reliable than any claim in a meeting.

"Very well," Carmeli said.

They sat for a moment in the hangar office. The ground crew was finishing the post-flight inspection. The aircraft's turboprops had cooled. The rotodome sat still above the fuselage in the late-afternoon light coming through the hangar's high windows — its six-metre diameter no longer seeming outsized but simply what it was: the right size for the antenna inside it, for the radar that antenna housed, for the mission the radar performed.

"One thing I want to say," Peled said. He said it in the register of someone who has been waiting for the right moment and has found it. "The ISMC contribution. The clutter rejection at 22 knots minimum detectable speed. The dual-emitter RWGS processing. These are capabilities that the American AWACS programme does not have at this stage of their development. I reviewed their published technical specifications. Their minimum detectable speed against ground clutter at low altitude is approximately sixty knots. Their RWGS geolocation accuracy at this range is wider than what was demonstrated today."

He paused.

"I am not saying the Akashganga is better than the E-3 Sentry. The E-3 is a larger aircraft with a larger antenna and more processing power overall. It has advantages we do not have. I am saying that in specific performance areas — particularly the low-altitude clutter rejection — the ISMC processing provides a capability that the Americans, at this stage of their programme, do not match."

He looked at Karan.

"This is worth knowing clearly. Not for boasting. Because it tells you where the programme's most significant technical contribution lies. If you are thinking about how to develop this further — what to invest in, where the leverage is — the semiconductor signal processing is the answer."

Karan absorbed this.

"We have been thinking about the semiconductor programme as infrastructure," he said — meaning that it had been thought of as the foundation that made other things possible rather than as a capability in its own right. "What you are describing is a case for treating it as a strategic capability in itself."

"Yes," Peled said. "That is what I am describing."

Squadron Leader Deepak Nair stayed in the hangar after the debrief ended and the visitors began to leave. He had asked permission to sit in the Akashganga's cabin — at the operator stations, with the displays dark and the systems powered down — and to write. He had a notebook. He planned to use it.

What he was writing was not a technical evaluation report. He was a specialist in air traffic control, not in radar engineering, and the technical specifics of what he had seen today were already being covered by Mahajan's evaluation. What Nair was writing was something more operationally specific: a set of questions about how the system would actually be used by real crews in real combat situations, and what the doctrine for that use needed to address.

He sat at the fighter control officer station. He ran through the scenario he had been thinking about throughout the day: an Akashganga on patrol over the Bay of Bengal, maintaining a track picture of forty-seven contacts simultaneously, when a new contact appears to the south, low altitude, high speed, heading northeast toward an Indian naval task force. The track appears on the display. The mission commander designates it as a potential threat. The fighter control officer — sitting at this station, in this seat — has to direct the two intercept aircraft currently on the IAF's eastern combat air patrol toward the contact.

What does the fighter control officer say? In what order? To which aircraft? Using what radio procedure? How does the fighter control officer know which of the two intercept aircraft can reach the threat contact most quickly, given their current positions and fuel states? How does the fighter control officer avoid directing both aircraft onto the same target, leaving other parts of the task force undefended?

These were not questions the aircraft's engineering could answer. They were questions that doctrine had to answer — written procedures, trained responses, standardised communications that allowed the crew aboard the Akashganga and the pilots in the fighters to work together without confusion or contradiction.

None of this doctrine existed. What existed was the hardware. The hardware was real. The doctrine was a blank page.

Nair wrote the blank page's first words. He wrote: The Akashganga fighter control function requires the following doctrinal elements to be specified before operational deployment. And then he listed them. The command authority chain. The communications procedures. The intercept geometry calculation standards. The rules of engagement integration. The emergency procedures for loss of data link. The handling of degraded track quality. The coordination with ground-based air defence. The training standards for fighter control officers and the qualification requirements.

Seventeen items. He numbered them.

He wrote for two hours. When he was done, he had the outline of a doctrine framework — not the framework itself, which would take months of exercise and revision, but the skeleton of questions that needed to be answered before the framework could exist.

When he left the aircraft he found Rawat in the office adjacent to the hangar, also writing.

"Your modification list?" Nair said.

"Twenty-seven items," Rawat said. "The terrain proximity alert and the time-to-closest-approach display are items one and two. The rest are engineering — the vibration in the rotation mechanism, the data link dropout at maximum range, the display latency spike. Nothing fundamental."

"Twenty-seven items," Nair said. "For the system that worked today."

"A system that works perfectly and has no further development items is a system whose engineers were not paying close enough attention," Rawat said. "Twenty-seven items means we were paying attention."

He looked at Nair's notebook.

"How many items do you have?"

"Seventeen doctrine questions. None of which are fixable in software."

"Seventeen doctrine questions," Rawat said. "For the system that worked today."

"Doctrine is always the harder part," Nair said. "The hardware does what it does. Doctrine is what people do with it. People are more complicated than hardware."

Rawat looked at his modification list, then at Nair's notebook.

"It is the same kind of problem," he said. "You identify what is not yet defined. You define it. You test whether the definition works. You revise. The process is not different because one side is engineering and the other is operational doctrine. Both are specifications of what should happen in a complex situation. Both need to be correct."

Nair considered this.

"Who will develop the doctrine?" he said.

"Air Headquarters will assign someone," Rawat said. "Keelor said start immediately. Starting immediately means someone needs to be assigned immediately."

"The person assigned needs to have been here today," Nair said. "You cannot develop doctrine for a capability you have not operated."

"The programme could use someone embedded from Air Headquarters," Rawat said carefully. "An officer who understands ATC, understands the fighter control integration problem, and can work with the engineering team to make sure the doctrine and the system evolve together rather than separately."

Nair was quiet for a moment. The conversation had moved toward a question he had not expected today to raise, and he was being careful.

"I will speak to Air Commodore Gowda," he said. "The doctrine development needs to be his tasking, not something I volunteer for."

"Yes," Rawat said. "But Gowda will ask who should be tasked."

"Then he will ask," Nair said. He closed his notebook. "The programme needs the doctrine as much as it needs the hardware. That is what today showed."

"Today showed that the hardware works," Rawat said. "It also showed that hardware without doctrine is a capable machine nobody knows how to use."

Nair stood.

"I will speak to Gowda," he said again.

He left. Rawat returned to his modification list. Twenty-seven items. Each one solvable. Each one more specific than before today. Each one a statement of something that needed to be true and was not yet true, which was the definition of useful engineering work.

Karan left the hangar at seven-thirty.

He walked across the apron slowly, in the manner of a man with nowhere specific to be and who has decided to use the walk as the transition between the day and the evening rather than merely as the physical movement between two locations.

March in Gorakhpur had the quality of the early season finding itself — the thick winter cold gone, the heat of April not yet arrived, the air having the specific freshness of a climate in between. The hangar lights were still on behind him, visible as a yellow rectangle in the complex's darkening mass. The apron was quiet except for the ground crew finishing the post-flight check on the Akashganga, their voices carrying lightly across the concrete.

He thought about what had happened today in the way he thought about things that mattered — not immediately, not with the analytical urgency of someone who needed to decide something right now, but with the patient attention he brought to things worth understanding properly. The system had worked. Every subsystem, every integration point between the Israeli electronics and the Indian airframe and the ISMC semiconductor processing and the Netra-derived radar — all of it had performed at specification or better. The low-altitude simultaneous tracking, which was the most demanding and the most novel element of the demonstration, had worked. The two-emitter RWGS geolocation had worked. The data link had transmitted the picture reliably enough that the ground station's display was useful rather than merely indicative.

It works.

He had been carrying the question since May 1973, when the joint programme was formally initiated at Ben-Gurion airport in Tel Aviv, in the IAI conference room, with Carmeli and Peled and the full technical teams of both sides present. He had known, walking into that meeting, that the programme was asking for something that had never been done in this specific combination — Israeli electronic warfare expertise, Indian semiconductor processing, a jointly developed radar architecture, all integrated into a modified Soviet-designed airframe, managed across ten time zones and two languages by two organisations that had known each other for less than a year.

The meeting had been the kind of technical conversation that happened when both sides knew what they were doing. No performance. No persuasion. Simply: here is what we have, here is what we need, here is what a joint programme could build. The Israeli side had been clear about what ELTA's EW expertise could contribute and where the gaps were — the signal processing, the semiconductor speed, the radar architecture that ELTA's own radar team had not developed in this specific configuration. The Indian side had been equally clear. The clarity on both sides was what made the conversation move quickly and what made the programme possible.

He reached the edge of the apron. The temple was not visible from here — it was on the other side of the city — but he thought about it anyway. The morning pradakshina had been rushed, the awareness of today pulling him forward. He would go again this evening, properly, without the urgency.

He thought about the men who had built this. Rawat, whose radar architecture had been the Indian spine of the programme — the man who had been developing the Netra radar since the S-27's early days and who had understood, when the joint programme was proposed, exactly what the Israeli EW expertise would make possible if combined with the signal processing he was building at ISMC. Peled, who had spent his career building the EW systems that were the programme's other spine and who had recognised in the ISMC semiconductor work a capability that neither Israel nor any Western nation outside the United States had available. Carmeli, who had managed the IAI side with the patient authority of someone who had integrated complex systems before and who knew that integration was not primarily an engineering problem but a human one — the problem of getting different teams with different working cultures and different technical languages to build something together that they could not have built separately.

Sethuraman, twenty-eight years old, who had helped design the 3-millimetre semiconductor devices and who had stood in the briefing room this morning and explained what they did with the contained excitement of a young man who has built something new and knows it is new. Krishnan, who had solved the rotodome structural mounting problem at eleven-thirty in the evening in January. Anand, who had flown the aircraft with the undemonstrative precision of a pilot who understood that the best flight test was the flight where nothing dramatic happened, because the absence of drama was the demonstration.

All of them had given a year and a half to this. In some cases — the ISMC semiconductor work, the radar architecture development — they had given longer. The question that had been carrying since May 1973 had been answered today by the work they had done.

He was quiet with this for a moment.

Then he walked to the temple.

The evening aarti was already underway when he arrived. The bells, the chanting, the specific quality of a hundred people's directed attention gathered in the main hall — the sound of it came through the temple's outer corridor before the hall was visible. He stood at the back of the hall for the remaining fifteen minutes of the prayers, in the way he always stood at the back rather than presenting himself to be seen: not as a religious performance but as something quieter, the presence of a person who needed, occasionally, to stand in a place that reminded him what things were for.

The industrial work — the aircraft in the hangar, the semiconductor facility, the radar team, the defence ministry's interest in what all of this produced — was for something. It was not for its own sake. It was not to be admired as a demonstration of what a company could do. It was because the country needed it, because the decade's geopolitics had arrived at a shape that required certain things to be built and deployed and operational, because the specific gap that the Akashganga filled — the gap between India's ground radar coverage and the complete air picture that a genuine air defence required — was a gap that had real consequences in a world where that gap could be exploited.

He stood in the aarti's light and sound and let the day's significance settle below the level of analysis, where it would stay.

When the prayers ended and the crowd began to move, he did not stay. He walked out through the temple's outer gate and back toward the guest house, through the Gorakhpur streets, through the specific evening of an early March in eastern UP that was neither winter nor summer but its own particular thing, carrying the quality of a day completed and the quality of something real now existing that had not existed before it.

The Akashganga was in the hangar. The rotodome was still. The aircraft that had spent four and a half hours looking at the Indian sky would look at it again, many times, through trials and modifications and exercises and adjustments, until the twenty-seven items on Rawat's list were resolved and the seventeen items on Nair's doctrine framework were answered and the bilateral production programme was formalised and the first production aircraft was built and the IAF's first Akashganga crew had completed their training and flew the system over the Bay of Bengal on their first real patrol.

That was still ahead. What was behind, as of today, was the demonstration that it could be done.

The programme was real.

He went inside.

End of Chapter 147

Akashganga Joint Programme — Summary Record

Full Name: Akashganga Airborne Early Warning and Control System Name Origin: Sanskrit — Akasha (sky) + Ganga (the Ganges river) = the River of the Sky, the Indian name for the Milky Way. Proposed by Dr. S. Rawat, June 1973. Unanimously adopted. Programme Type: AEW&C — Airborne Early Warning and Control Indian Preliminary Work: August 1972 (radar architecture, ISMC semiconductors) Joint Programme Start: 28 May 1973 (Israel joined) Prototype Demonstration: 2 March 1974

Base Platform: Ilyushin Il-18 turboprop transport, extensively modified Fuselage reinforced at 14 structural points. Windows replaced. Rotodome mounted on twin struts above centreline. Wings structurally strengthened. Rotodome diameter: 6 metres. Rotation rate: 6 RPM (one full 360° sweep every 10 seconds). Flight crew: 4. Mission operators: 6.

Israeli Partners:ELTA Systems (subsidiary of IAI) — Radar Warning and Geolocation System (RWGS), detection and identification of hostile radar emissions. Detection range: 600km+. Library of 180 radar signatures. Position accuracy 3km at 400km range.

Israel Aerospace Industries (IAI) — Airframe system integration, rotodome rotation mechanism engineering, overall systems integration management.

Tadiran Electronic Industries — Mission computer framework, COMINT (Communications Intelligence) receiver suite, rotodome rotation gyroscope stabilisation control, operator display architecture.

Indian Partners:Shergill Aviation — Airframe modification engineering (Dr. P. Krishnan), structural design, aerodynamic analysis of rotodome mounting.

ISMC Gorakhpur (Shergill Industries semiconductor division) — 3-millimetre semiconductor signal processing devices. Performance: minimum detectable speed at low altitude 22 knots (conventional systems: 70 knots); Doppler frequency resolution 40× finer than conventional silicon processing; simultaneous processing of 250+ tracked objects.

Netra Radar Team (Dr. S. Rawat) — Multi-mode Doppler radar architecture, clutter rejection algorithms, track management software.

Demonstration Results — 2 March 1974:

High altitude commercial traffic: Altitude accuracy: ±180 feet (specification: <500 feet) Speed accuracy: ±7 knots (specification: <15 knots) Heading accuracy: <1° (specification: <2°)

Low altitude simultaneous dual-track: MiG-21 at 300 feet AGL, 480 knots — acquired in 4 seconds Hawker Hunter at 500 feet AGL, 320 knots — acquired in 4 seconds Both tracked simultaneously. All measurements within specification.

RWGS dual-emitter: Two simultaneous radar stations located. Position accuracy within specification at both ranges. Both stations correctly identified from signature library.

COMINT: Function confirmed. Direction-finding within 1.8° of true bearing.

Post-Demonstration Action Items: Priority 1: Terrain proximity alert display (S/L Nair requirement) — software modification Priority 2: Time-to-closest-approach display metric (S/L Nair requirement) — software modification Items 3–27: Engineering refinements (rotation vibration, data link optimisation, display software, ergonomic adjustments)

IAF Evaluation (Air Vice Marshal Keelor): Formal IAF participation in programme recommended. Conditions: Priority 1 and 2 modifications before operational service; full-fidelity crew training simulator within 6 months.

Key Terms Defined:AWACS — Airborne Warning And Control System. Flying radar aircraft that provides complete air picture to commanders. Rotodome — Rotating radar dome mounted above aircraft fuselage. Rotates to scan all directions. AGL — Above Ground Level. Aircraft height above the actual ground directly beneath. EW — Electronic Warfare. Use of the radio frequency spectrum for military purposes. RWGS — Radar Warning and Geolocation System. Passive system detecting and locating hostile radar transmissions. COMINT — Communications Intelligence. Monitoring of radio communications for military intelligence. Doppler Effect — Change in frequency of a wave (including radio waves) caused by motion of the source or receiver. Used by radar to measure target speed. Ground Clutter — Radar returns from the ground, which must be filtered to find aircraft returns. Track — The radar system's maintained record of an object's position, heading, and speed over time. Pulse-Doppler — Radar processing method using Doppler frequency measurement to separate moving targets from stationary clutter. Cross-section (radar) — Measure of how strongly an object reflects radar signals. Not the object's physical size. Data link — Radio connection transmitting digital data between aircraft and ground station. Mode C transponder — Aircraft electronic device broadcasting altitude to ATC radar systems. CRT — Cathode Ray Tube. The screen technology used in displays before modern flat screens.