3.5" floppy failure mode? Hub not attched to media.

On Feb 27, 2015 12:58 PM, "Chuck Guzis" <cclist at sydex.com> wrote: with head initially on-track. That is, I'm not at all sure that it compensated for any eccentricity. They wouldn't have needed fullblown embedded servo for that. They went to a lot of expense to have embedded servo tracking because it was actually necessary. 96 and 100 tpi was pushing the limits of open-loop positioning with a crappy clamping system. Had that not been a problem, higher t r ack pitch drives would have appeared much sooner. The ability to make heads for narrow tracks was not the limiting factor. Even ignoring the hub clamping tolerance, while mylar has very good dimensional stability (and was chosen largely for that reason), it isn't good enough for 200 tpi open-loop. Even the tempco is a problem, and isn't sufficiently uniform in all directions. Since it's not good for much more than 135 tpi, it's fairly critical just at 135 tpi, and it takes only a minute centering error for tracks to be only partially readable. If you modified the positioning electronics for microstepping, you *might* be able to compensate for increased centering error by attempting to read each track at a range of offsets, getting some sectors at one offset and others at another, etc. I've done the math only for the case of normal positioning, and it looks pretty grim; for large sector sizes even the microstepping probably wouldn't help with much centering error beyond that already caused by the tolerance of the hub clamping. Having looked at the necessay geometey, I'll believe that data recovery by reattaching the hub of a 3.5" floppy is practical only when someone has succeeded doing it to multiple disks, and then I'll want to know how many attempts per disk it took, how much time per attempt, amd what (if any) special apparatus was used. The angle of hub attachment is a far less critical problem. If a controller is used that actually cares about the relative postition of the index pulse when reading data, a one-shot (or digital equivalent) can trivially be used to introduce a delay in the index pulse to compensate for the angular position error.