data recovery from rotating memory with an HP 1650B

Hey, I just realized, I have (here at work) an old HP 1650B, with 96 (!) channels, I think it is, too many to count ("one... two... three... ... many!") but certainly >> 48. It claims to be dig-it-all (but we know what that means) and has a variable threshold (and I'd need level changers to deal with -20V logic, or wherever I pick off the signal). I don't have an operating manual, though I've puzzled out basic operation, but does anyone know how to extract sampled data from this thing? Here's a hypothesis: the sniggle I'll get off the heads will be the same as, or less than, the specified (in the hardw. man.) signal. The voltage won't increase off the head! If true, I set the 1650B thresholds such that it mimics the LGP-21's memory. Take N samples of the disk (all at once!). Check that each sample is self-consistent, eg. every sample is the same. If subsequent samples are not all the same, see what's changing. If one or few bits/track are flapping, it could be setup or weak signal. Repeat taking sample sets, with the 1650B thresh lowered 2 - 5%, say. At some point the differences between sets should radically alter. (I would in fact take a few sample sets with threshold > LGP's setting, for completeness.) In an ideal setup, each sample within a set would be self-consistent, and each sample set (with varying threshold) would be the same until the thresh crosses into the ambigious zone between LGP-21 "0" and "1", at which point it would deteriorate into garbage. The LGP-21 hardw. man. comes not only with complete memory alignment data and techniques, but photos of scope screens. Too bad all machines don't come with their own historical recovery documentation! If the 1650B can "print" to a port I can capture then this nicely formatted data could be scripted into something canonical.