The U-2 shootdown over Sverdlovsk in 1960 settled an argument. Altitude was no longer sanctuary. If you wanted to reach a defended target, you would have to go under the radar rather than over it, at night, in weather, at a few hundred feet.

Which raises an immediate problem. A pilot cannot see a ridge in cloud at 500 knots, and by the time he could, the aircraft would already be inside its own turning circle.

The answer was to let a radar look ahead and fly the aeroplane over the landscape automatically. It worked, it has killed people, and it is quietly being replaced by something that does not transmit at all.

The electronic ski-toe

The principle is more elegant than the hardware suggests, and the clearest description of it comes from the engineer who led the Ferranti programme for the TSR.2.

Imagine a ski held ahead of the aircraft, its flat section sitting at the height you want to maintain, its upturned toe reaching out to the horizon. Slide it along the ground profile the radar is drawing. Wherever the terrain pushes the ski up, the aircraft must climb; wherever it falls away, the aircraft may descend.

The template was a physical thing in the cockpit sense. The F-111E preflight checklist has the crew verify that the zero command line moves properly and produces a smooth curve across all three ride settings and every clearance setting, because both knobs reshape the ski.

The original General Dynamics training film for the F-111 system is above, and it is the best surviving explanation of how the equipment was actually operated.

Hard, medium and soft

The ride switch is the part crews remember. On the F-111 it selects how much negative g the system is allowed to command as it pushes over a crest: zero on hard, half a g on medium, three quarters on soft. Hard keeps the aircraft pinned closest to the terrain and is thoroughly unpleasant. Soft computes an earlier, gentler anticipatory command and costs you exposure.

Crucially, the flight manual is explicit that the fail-safe fly-up signal is not affected by the ride switch. You can choose how roughly you are flown. You cannot choose whether the system saves you.

Nor is the negative-g limit arbitrary. It exists because the radar cannot see behind a crest. Push over too aggressively and the aircraft arrives in a valley the system has not yet drawn, with a second hill waiting in the shadow of the first. Blain put the TSR.2 limit at half a g for exactly that reason, and the F-111 numbers bracket the same value.

When it stops working

Two failure modes are built into the physics rather than the engineering.

The first is low-reflectivity ground. Over calm water, dry lake beds, smooth sand or snow, very little energy comes back. The F-111 manual states plainly that terrain following will then be commanded by the radar altimeter, which looks only downward and has no forward-looking capability. On rising ground that is a serious deficiency, and the manual quantifies it: the aircraft will fly below set clearance by 87 feet for every degree of slope.

The second is that the system knows when it is blind, and its response is violent. Any internal fault, any loss of roll or air-data or radar-altimeter validity, or a radar altitude below 68 per cent of the selected clearance, triggers an automatic fly-up. The initial response may reach 3.8 g, hitting limit pitch attitude in about two seconds, with pitch trim disabled and the stick centred.

Edwards Air Force Base footage of B-1B low-altitude terrain following is above.

The accident that shows the trust problem

On 30 November 1992 a B-1B from Dyess Air Force Base flew into a ridgeline 36 miles south-south-west of Van Horn, Texas, on a night low-level sortie. All four crew were killed. The aircraft struck roughly 300 feet below the crest of a 6,500-foot ridge.

The automatic terrain-following system had generated a fly-up thirteen seconds before impact. Just before the crash, the crew manually interrupted it.

That is the central tension of the whole technology. A system aggressive enough to save you will also cry wolf, and a crew that has been thrown around by unnecessary fly-ups will eventually reach for the override.

The Albanian accident, which killed nine, followed a similar shape: the crew missed the start-climb point for a 5,500-foot saddle, switched the terrain-following system on and immediately received an obstacle warning, and stalled about 200 feet above the terrain. The board also noted the crew’s focus on terrain masking rather than terrain avoidance.

Who flew it

By February 1964 the USAF already had terrain-following radar in the B-52H and the F-105, and Texas Instruments held subcontracts for the RF-4C and the F-111. The contemporary trade press was unusually candid about why: proof of Soviet ability to knock down high-flying aircraft had forced the United States to build capability for low-level penetration.

IL Tornado di Panavia carried two separate nose radars, the smaller one dedicated to terrain following, coupled to the autopilot for automatic flight at around 200 feet. Both radars were built by Texas Instruments, making them the only major items in the baseline aircraft not manufactured in Europe.

IL F-15E took a different route, carrying the function in the LANTIRN navigation pod rather than building it into the airframe. The Soviet Su-24 used a dedicated set named Relyef alongside its main scanners. The MC-130H Combat Talon, officially cleared as low as 250 feet in adverse weather, used a redundant dual-band forward-looking radar.

Why it is going away

A terrain-following radar has one unavoidable characteristic: it transmits. It points energy forward, at low altitude, in the direction it is flying, which is precisely where anyone who wishes to intercept it is standing.

The clearest evidence of how seriously that was taken is the engineering response. The B-1B’s AN/APQ-164 uses a fixed electronically scanned antenna canted downward specifically for reduced radar observability, and its manufacturer advertises low probability of intercept terrain-following modes. Low-probability-of-intercept modes only exist because ordinary ones are interceptable.

The modern answer removes the transmission entirely. TERPROM, conceived in 1977 as a private venture at British Aerospace, compares radar-altimeter readings against a stored digital elevation model to work out where the aircraft is and what is ahead, without emitting anything forward. It has been fitted to the F-16, Harrier, Jaguar, Tornado, A-10, Typhoon and C-130 among others, and it delivers a limited but entirely passive terrain-following capability.

The F-35 has no terrain-following radar at all. Its Auto-GCAS uses onboard digital terrain data to determine whether a collision is imminent and, at the last instant, commands a roll to wings level and a 5 g pull. The system was fielded from June 2019 and won the Collier Trophy.

Which makes a neat closing comparison. In 1967 an analogue computer sliding an electronic ski over a radar picture would save your life with 3.8 g. In 2019 a digital elevation map does the same job with 5 g, tells nobody you are coming, and never needs to see the hill at all.

Sources: T.O. 1F-111E-1 General Dynamics F-111E flight manual; Bill Blain, TSR2 Terrain Following Radar Development, recollections; Electronics magazine, 21 February 1964; USAF Aircraft Accident Investigation Board report, MC-130H 87-00127; Aviation Safety Network wikibase 48343; USAF and AFSOC fact sheets; Northrop Grumman AN/APQ-164 material; Lockheed Martin Auto-GCAS releases.