Providing seventeen matched vacuum tubes for a national broadcaster

The ask

The broadcaster's engineering department had pulled two Neumann V72 microphone preamplifiers out of storage — units originally installed in the mid-1950s and retired in the 1980s when the studio migrated to solid-state consoles. They had decided, as broadcasters occasionally do, that the V72s should sing again for a commemorative live-to-tape session marking the studio's centenary.

Each V72 uses several EF804S pentodes in its gain and driver stages. Between the two amplifiers, spares, and a hot swap set, they wanted seventeen tubes. Not any seventeen. Seventeen that measured within three percent of each other on transconductance, with plate current inside a five percent band, and gm at operating point verified on a calibrated Neuberger RPG 4/3 tester.

The reason was simple: the broadcaster's chief engineer had been through this before, in 1983, and he knew what unmatched tubes did to the noise floor of a stereo pair.

If they're not matched, the two channels drift apart over an hour. The image narrows, the top end goes glassy on one side, and by the third reel you can hear which one is tired. I don't want to hear it. Neither will the listeners.

The search

EF804S tubes were made by Telefunken from the late 1950s into the mid-1970s, primarily at the Ulm plant, and they were rare when new. NOS stock exists but is scattered, expensive, and often mislabelled — European EF804S are frequently confused with the standard EF804, which has different biasing and is unsuitable for the V72's operating point.

We assembled candidates from three inventories: a former broadcast supply house in Stuttgart with a sealed batch from 1971, a private seller in Vienna with a mixed lot pulled from decommissioned Studer consoles, and our own long-standing stock in the workshop. Eighty-four tubes in total, all with Ulm-plant date codes between 1962 and 1974.

Every tube was labelled, logged, and rested for forty-eight hours at operating voltage on our own Neuberger RPG 4/3 before measurement. Six failed immediately — internal shorts, gassy plates, one microphonic enough to ring when the workshop door closed. That left seventy-eight.

Of the seventy-eight, we found a cluster of twenty-two tubes falling inside a tight transconductance window. From those, we selected the seventeen with the closest plate current tracking and the lowest measured noise at 1 kHz.

The find

The matched seventeen were not from any single lot. Six came from the Stuttgart batch, eight from the Vienna Studer pull, and three from our own shelf — one of those three had been sitting in a box marked "check someday" since 2011.

The final matching spread came out at 2.1 percent on transconductance across the set, comfortably inside the client's three percent tolerance, and inside what we suspect the Telefunken factory itself achieved on any given production week.

The delivery

Each tube shipped in its original Telefunken carton where possible, in a period-correct Valvo carton where not. Every tube was individually labelled with a serial, a measurement date, and its gm, Ia, and noise figures at operating point. The client received a bound test report — one page per tube, with the tester calibration certificate at the front.

The two V72s went back into service at the broadcaster's Studio 3 in early May, in time for the centenary programme: a live retransmission of a 1954 radio play, performed by the same theatre company that had staged the original, into the same microphones, through the same amplifiers.

The chief engineer wrote to us the following week.

No drift. Ninety-two minutes on air, and the image held. I could hear the room in 1954. That is what these amplifiers were built to do.

Two tubes from the extended candidate pool remain in our workshop, matched to each other but not to the delivered seventeen. They are catalogued and available.