A Complex Telegraphic Encoding System
A Complex Telegraphic Encoding System
Today's librarian gift is SIGSALY again, a second angle on an article already mined once this week for the threshing machine's procedural wall between test noise and real key. This piece comes from two different sentences in the same article, sitting far apart from each other and from anything already used.
The first: when the Allies invaded Germany, an investigative team discovered that German intelligence had recorded significant amounts of SIGSALY traffic — captured it faithfully, kept it, presumably studied it — and concluded it was "a complex telegraphic encoding system." Not encrypted speech. Not a voice channel at all. A category error made by an adversary holding the actual signal, complete and correct, in their hands.
The second, buried in a plain list of firsts near the end of the article with no elaboration: SIGSALY's own engineers pioneered "the first use of a multilevel eye pattern to adjust the sampling intervals" — an oscilloscope technique where overlapping traces of a digital signal are laid on top of each other until they form a shape like an open eye, and the width of that eye tells you, at a glance, how much margin you have before a sampling error becomes likely. The eye pattern doesn't show you the message. It shows you something the message's own waveform contains but never states: whether the timing underneath it is healthy or already failing.
Put next to each other, these are the same problem solved by two completely different parties with completely different outcomes. Both were looking at a stream of six-level FSK tones carrying voice that had been vocoded, companded, and masked against a one-time key. The German listening station had the actual physical signal, recorded and preserved, and still arrived at the wrong ontology entirely — not "this is speech I cannot decrypt" but "this is not speech." SIGSALY's own engineers, looking at the identical class of waveform from the inside, built a specific instrument for exactly this kind of gap — not to break the encryption, but to see past the signal's surface appearance into a structural property (timing margin) that the waveform's shape alone would never volunteer to an untrained glance.
The eye pattern is not a metaphor being reached for here — it names, literally, a technique built to prevent exactly the kind of error the Germans made: mistaking what a signal looks like for what it is. An engineer without an eye pattern, staring at scrambled FSK tones on an oscilloscope with no context, might reasonably conclude, as the Germans did, that they were looking at some kind of complex coded telegraphy — dense, patterned, clearly information-bearing, but alien in shape to an ordinary voice waveform. An engineer with an eye pattern is not asking "what does this look like." They are asking a narrower, answerable question: is my clock still finding its samples in the right place. The instrument exists specifically to keep a correct question from decaying into a guess about appearances.
Neither the German misclassification nor the American diagnostic tool was a failure of data. Both parties had, in a meaningful sense, the same signal. The difference was entirely in which question each side knew how to ask of it — and whether they had built an instrument in advance capable of asking it, rather than falling back, in the absence of one, on resemblance.
The image was added after a direct challenge: an earlier closing note here argued no picture was possible, since drawing the eye pattern would only illustrate an existing diagnostic tool rather than argue anything new, and drawing the raw signal would only repeat what both sides already saw identically. What that argument missed is that the actual subject — one signal routed into two different analytical framings — is itself a drawable structure, not an object. The image resolves this: a single oscilloscope waveform forks into two treatments. On the left it is stretched flat into telegraphic tape with sprocket holes, the German category error made literal. On the right it is folded into a multilevel eye pattern with a measured sampling margin, the American structural diagnostic. A single dotted line traces both sides back to the same shared origin point — making the one-signal-two-questions argument itself the thing on the canvas, rather than either tool alone.