AT&T Long Distance via Shortwave

How America phoned the world, 1927–1975 — and why every overseas call was a radio call.

The first transatlantic telegraph cable was laid in 1866. The first transatlantic telephone cable was laid in 1956. In the ninety years between, you could send a telegram under the ocean but you could not send a voice, because the cables did not have the bandwidth to carry one.

So when AT&T opened commercial telephone service to London in 1927, it had exactly one medium available: radio. And for the next three decades, every overseas telephone call placed from the United States went over the air. Not as a novelty or a backup — as the only way it could be done. The plant that carried those calls was one of the largest radio installations ever built by a private company, and almost all of it is gone.

The problem: cable could not carry a voice ↑ contents

Transatlantic telegraph cables had been in service since 1866. They worked, they were profitable, and they were useless for telephony — a telegraph cable is a very long, very lossy capacitor, and it will pass dots and dashes but not the several kilohertz a human voice needs. There was no way to put an amplifier on the sea floor, two miles down, and have it survive.

That problem was not solved until repeaters could be built rugged enough to be spliced into a cable and dropped into the Atlantic. Until then, if you wanted to speak to Europe, you spoke through the ionosphere.

January 7, 1927 — and it was not shortwave ↑ contents

Commercial transatlantic telephone service opened on January 7, 1927, with a call between Walter S. Gifford, president of AT&T, in New York, and Sir Evelyn Murray, Secretary of the British General Post Office, in London. A test call had been made the day before. The recording was added to the Library of Congress National Recording Registry in 2005.

The service opened on longwave, not shortwave — frequencies in the range of 58.5 to 61.5 kHz. And it was not one radio link but two, because each direction had its own transmitter, its own receiver, and its own continent:

Houlton, Maine is unambiguously AT&T's: the AT&T Transoceanic Receiver Station, with a Beverage antenna over three miles long and two miles wide, on ground that is now straddled by Interstate 95.

The move to shortwave ↑ contents

Longwave worked, but it was enormously expensive in power and real estate for a single circuit. Shortwave, refracted off the ionosphere, could do the same job with a fraction of the transmitter and a directional antenna — and it could do it many times over, on many frequencies at once.

Rugby put a shortwave telephony transmitter on the air on July 6, 1928. AT&T's shortwave circuits, in the 6 to 25 MHz range, came in from 1929. From that point the longwave plant was legacy, and the shortwave plant was the business.

The Pole Farm: Lawrenceville, New Jersey ↑ contents

In 1928 AT&T bought roughly 820 acres of farmland across Lawrence and Hopewell townships in New Jersey, chosen for being flat and for being far from the electrical noise of New York City. What went up there became, by contemporary accounts, the largest radiotelephone shortwave transmitting station in the world. Locally it was known as the Pole Farm.

From 1929, every telephone call from the United States to Europe, Africa, or the Middle East left the country through Lawrenceville. Buenos Aires came online in 1930.

The curtains ↑ contents

The signature of the site was twenty-six steel towers, 180 feet tall, arranged in a V — two lines running about a mile each, one aimed at London and one at Buenos Aires, with the towers spaced roughly 250 feet apart. Strung between them were wire-mesh curtain antennas.

And here is the part that stops a radio person cold. The curtains were raised and lowered on Roebling cable, like window shades, on a schedule. Machinery hoisted the appropriate curtain at the appointed hour, because the frequency that works to London at 3 a.m. is not the frequency that works at noon, and the antenna has to change with it. The ionosphere moved, and the antenna farm physically reconfigured itself to follow.

The transmitting tubes were water-cooled. At its height the station employed about eighty radio men, whose principal occupation was replacing tubes as they burned out, plus riggers to keep the outdoor plant standing.

Lawrenceville transmitted. It did not receive — a transmitter that size would have deafened any receiver within miles. Receiving happened fifty miles away at Netcong, and later at Manahawkin.

The enemy: selective fading ↑ contents

Every serious engineering decision AT&T made in this system was an answer to one problem, and it was not weak signals.

A shortwave signal does not arrive at the receiver by one path. It arrives by several — one hop, two hops, different layers — each with a different path length and therefore a different arrival time, and each striking the receiving antenna at a different vertical angle. Those copies add and subtract at the antenna. Some frequencies within the voice band reinforce; others cancel. The result is not quiet audio. It is mangled audio, and it changes second by second.

You cannot fix that with more power. More power gets you more of all the paths at once. The only fix is to accept one arrival angle and reject the others — which means directivity, not in the horizontal plane where a rhombic already gives you plenty, but in the vertical plane, where nobody had it.

The MUSA — a steerable antenna, in 1937 ↑ contents

The answer came from Harald Friis and C. B. Feldman at Bell Labs, and it is one of the most remarkable pieces of radio engineering of the era: the Multiple Unit Steerable Antenna, the MUSA.

Take a line of rhombic antennas, spaced along the great-circle bearing to the distant transmitter, all pointed the same way. Run each one back to the receiving building on its own coaxial line. Now, at the building, before you combine them, put each signal through an adjustable phase shifter working at intermediate frequency — and gear the shifters together, so that turning one assembly changes them all in concert.

When the phases are set so that the outputs of all the rhombics add up for a signal arriving at one particular vertical angle, that angle is reinforced and every other angle is suppressed. Turn the assembly, and the antenna's beam sweeps up and down in elevation. You have steered the antenna without moving it.

And then the detail that turns a clever antenna into a system: they built several sets of phase shifters in parallel, each a separately steerable branch. So the station could lock onto two or three different arrival angles simultaneously and combine what each one heard — a diversity receiver that follows the ionosphere in real time.

The paper is "A Multiple Unit Steerable Antenna for Short-Wave Reception," Bell System Technical Journal, Vol. 16 No. 3, July 1937, pp. 337–419, and it is freely readable on the Internet Archive. It is worth an evening.

Only three MUSA arrays are believed to have been built: the experimental one at Holmdel, New Jersey, a full array at Manahawkin, New Jersey, and one at Cooling, on the Thames marshes in Kent — the British end of the circuit from Lawrenceville, sited on brackish tidal marsh precisely because salt-saturated ground makes a superb RF ground. Cooling's array was two miles long, sixteen rhombics on sixty-foot poles, each side 315 feet, and it came into service on 1 July 1942.

The National Radio Astronomy Observatory has called the MUSA the first radio astronomy interferometer. It was built to make a phone call sound better.

The stations: AT&T's shortwave plant in the United States ↑ contents

Transmitting and receiving were always separated by tens of miles. The pairing below is the essential architecture of the whole system.

The Pacific circuit is a Bay Area story. On December 31, 1931, radiotelephone service was established between the AT&T station on the California coast and Hawaii. Direct service to Sydney followed. Dixon transmitted; Point Reyes received. The Point Reyes receiving building still stands, and it sits next door to RCA's KPH — which is still on the air, kept alive by the Maritime Radio Historical Society.

Double duty: the same stations talked to ships ↑ contents

Look at the table again and something jumps out. Ocean Gate, Dixon, and Pennsuco are listed twice over — overseas and high seas. That is not sloppy record-keeping. AT&T's own station index describes Manahawkin as a "former overseas and high-seas shortwave receiving station" and Ocean Gate as a "former overseas and high-seas shortwave transmitting station."

The reason is pure physics, and once you see it you cannot unsee it. A rhombic aimed at Europe crosses the North Atlantic shipping lanes on the way. The directivity you build to reach London is the directivity that reaches a freighter eight hundred miles out. Building a separate marine network would have meant building the same plant twice, pointed the same way.

So AT&T did not build an overseas network and a marine network. It built transmit sites and receive sites — and the ones whose beam headings suited both jobs did both. Calling London and calling a ship used the same tubes, the same antennas, the same riggers, the same buildings. The only difference was which circuit the operator patched you into.

It went further than ships. The June 1958 ITU List of Coast Stations records WOO as "open to correspondence with aircraft" — one of only nineteen US coast stations so authorized. These sites were the general-purpose radio edge of the telephone network: whatever was out there and could not be reached by wire, they reached.

Lawrenceville appears to be the exception — the big dedicated overseas plant, listed as an overseas transmitting station and nothing else.

The marine side of these stations — the ship-to-shore service, the operators, and how you actually placed a call to a vessel at sea — is covered on the companion page: Coast Stations.

The scrambler the Germans broke ↑ contents

A shortwave telephone call is a broadcast. Anyone with an HF receiver and the right frequency hears it. AT&T's commercial answer, developed before the war for businessmen who did not want competitors listening, was the A-3 scrambler — voice inversion and band-shuffling, with the key changing at intervals. Against a casual eavesdropper it was fine.

Against the German post office it was nothing. The Deutsche Reichspost built an interception station at Noordwijk on the Dutch coast, and from 1940 they were reading transatlantic radiotelephone traffic, including Roosevelt and Churchill. The method was almost insulting in its simplicity: record the call on a BASF magnetic tape recorder and replay it through the possible scrambling combinations until it came out intelligible. In time they had it running near real time, losing only a few syllables after each key change until the machine found the new key.

On 29 July 1943, Churchill and Roosevelt discussed a possible Italian armistice. The Germans decoded it. Hitler's uncertainty ended, and he ordered the occupation of Italy — putting twenty divisions in the Allies' path instead of eight.

AT&T had known the A-3 was weak since at least 1936. The replacement, begun in October 1940, was SIGSALY — also called Project X, the X-System, and, for the sound it made, the Green Hornet. A vocoder-based digital system with a one-time key on phonograph records. Alan Turing was the only British officer cleared to evaluate it at Bell Labs. Twelve terminals carried roughly three thousand conversations.

And Churchill mostly did not use it. SIGSALY meant walking to a specific room; the broken A-3 was closer to hand. He did not use SIGSALY until April 1944.

The only phone company on shortwave ↑ contents

AT&T was not the only American company with big shortwave stations. It was the only telephone company with them — because the international radio business was divided by service, not by geography:

You can read the division straight out of the litigation. FCC v. RCA Communications (1953) is a fight between RCAC, Mackay, and Western Union, and every circuit at issue is a radiotelegraph circuit. When the court lists what those carriers competed against, it names cable telegraph, air mail, and "direct radiotelephone service" — radiotelephone appears as a separate, adjacent service, not as something RCA or Mackay offered. They moved telegrams. AT&T moved voices.

Western Union, incidentally, did eventually go to radio — but it was microwave, not shortwave, and it was postwar: an experimental relay authorized in March 1945, commercial traffic by 1948, a transcontinental beam in 1964. A different technology in a different decade.

How much traffic, actually ↑ contents

Less than you would think, and that is the point.

The FCC's 20th Annual Report, for fiscal year 1954, records that international radiotelephony — born in the 1920s — exceeded one million calls for the first time, an increase of 7.1 percent over the previous year, carried on 63 direct circuits reaching 111 foreign countries and overseas points.

Narrow that to the transatlantic route and the numbers get smaller still: by 1955, after nearly thirty years of service, only about 100,000 calls a year were being made between the United States and Britain. That is a couple of hundred calls a day, on the busiest and most famous international telephone route in the world.

This was never a mass-market service. It was radio-frequency spectrum, an enormous capital plant, and a small number of very expensive conversations.

The cable arrives, and it all ends ↑ contents

TAT-1 opened on September 25, 1956 — the first transatlantic telephone cable, two coaxial cables with repeaters, Clarenville to Oban. In its first year it carried about 220,000 calls between Britain and the United States: double what the radio circuits had carried in the previous twelve months.

HAW-1 followed in 1957, from Point Arena, California to Hanauma Bay, Oahu — and did to the Pacific circuit what TAT-1 had done to the Atlantic.

Cable did not beat radio on capacity alone. It beat it on quality. No fading, no ionosphere, no waiting for the band to open. Everything the MUSA had been built to fight simply stopped being a problem, because the signal now travelled through a copper tube on the sea floor instead of a plasma sixty miles up.

The shortwave plant did not vanish overnight — the Pole Farm ran until 1975, and high-seas service to ships continued until the FCC authorized AT&T to shut it down in 1999. But after 1956 it was a legacy system, and everyone knew it.

What is left ↑ contents

Lawrenceville is a county park. The Pole Farm is now part of Mercer Meadows; the towers are gone, and people walk dogs where the curtains hung.

Rocky Point was decommissioned in 1978 and demolished in the 1980s. It is now New York State land in the Long Island Pine Barrens — concrete ruins and fallen poles.

Ocean Gate is the saddest one. The FCC ended high-seas service there in 1999; the marsh became a wildlife refuge; and in 2025 the US Fish and Wildlife Service began removing more than 500 antennas and poles across 222 acres of salt marsh at Good Luck Point. The transmitter building stands empty with its windows out.

Point Reyes is the survivor. The AT&T receiving building is still there inside the National Seashore, with an interpretive marker, and RCA's KPH next door is still transmitting — operated by volunteers of the Maritime Radio Historical Society. If you want to stand in this history rather than read about it, that is where you go.

Two numbers I cannot reconcile ↑ contents

In the spirit of not pretending to more certainty than I have, two figures that appear widely and disagree with each other:

First-year traffic. Britannica states the 1927 service carried 11,000 calls in its first year. A British Post Office history of TAT-1 states it was a single circuit averaging about 2,000 calls a year. That is a fivefold gap on the most basic number in the story.

The price of a call. The 1927 rate is given as $75 (£15) for three minutes by some sources and as roughly $6 a minute by others. The British Post Office history says the New York rate was reduced to £9 for three minutes in 1928. These do not reconcile cleanly, and I have not found a primary source that settles it.

Both would be answerable from FCC annual reports or AT&T Long Lines statistics. Until someone works through those, treat any specific figure you read — here or anywhere — with suspicion.

Sources: Bell System Technical Journal, July 1937 (Friis & Feldman, "A Multiple Unit Steerable Antenna for Short-Wave Reception"), AT&T Long Lines (long-lines.net), the Library of Congress National Recording Preservation Board, Britannica, the FCC 20th Annual Report (FY1954), the Point Reyes National Seashore interpretive markers, the Maritime Radio Historical Society, The SWLing Post, the International Churchill Society, and the Morven Museum. Compiled by N6JET as a noncommercial hobby reference. Questions or corrections? Email chris@n6jet.com.

Return to n6jet.com