On 02/27/2015 11:40 AM, Eric Smith wrote: I'm not sure that the "fine" stepper was used other than to bring the head initially on-track. That is, I'm not at all sure that it compensated for any eccentricity. --Chuck

On Sat, 28 Feb 2015, Eric Smith wrote: Good point. For 512 byte sectors, the runout error would be minor (a small fraction of the min/max error o0f the track) within each sector. Still a small fraction of the total error even with 1024 byte sectors. But, not as easy with something such as Amiga, where the entire track is read, and then sectors parsed in software. But, that technique could be considered with a track read (ala WD179x or flux transition hardware), and then identify "good" and "bad" sections of the track with multiple reads, and splice together "good" sections to recreate the track. For READ (who would want to WRITE to a damaged disk??), even the NEC765 can READ with the index puls blocked. 'course then you have a few trivial hassles, such as errors being misreported as 128/"Drive Not Ready" Would disclosing fluid be adequate for corrcting the rotational error? It would be nice to have the index pulse. Using the premise behind the radial alignment tests of the Dysan Digital Diagnostic disk, it would be possible to calculate the direction of the radial/runout misalignment, and be able to keep track of which sectors are out of position. IFF the sector sequence of the disk is known, and consistent from track to track, (not Apple][ and Commodore64) then micropositioning based on one track could be applied to the entire disk. If the micropositioning could be done in real time, then correction could be applied for a "clean" read, or, if not, it would still make it possible to know which sectors will be readable at each microposition and read those sectors from every track before again modifying the position offset. I did, on a very few occasions, deliberately take a 5.25" drive out of radial alignment to read disks created on a misaligned drive. If I were still actively doing such, I would rig up a way to calibrate (maybe with a micrometer head?. On one group of 8" disks, light finger pressure pushing the head further towards the spindle made the difference for successful reads! Of couse, 48TPI is quite a bit easier than 135, and access to the positioner is easier on big drives. So, if radial runout is the problem, read all of the sectors that can be read on the disk, throw the radial alignment out slightly, read all track, throw it out the other direction, . . . And spend a ridiculous amount of time on a single disk. -- Grumpy Ol' Fred cisin at xenosoft.com

On Sat, Feb 28, 2015 at 10:57 AM, Chuck Guzis <cclist at sydex.com> wrote: Using algebra and geometry it is readily apparent that as little as 0.1 mm of centering error will cause problems. I don't have to do any experiment to determine that. Whether the hub reattachment is self-centering to within 0.1 mm is another question which I don't have an answer to. I rather doubt it, as there was no reason to engineer it to self-center to any significantly better tolerance than needed to ensure that the medium doesn't hit the edges of the casing. However, not having done any experiment or analysis of that aspect I'll readily concede that there's some chance that it's good enough.

On 02/28/2015 10:14 AM, Mike Stein wrote: I'm saying that it's very unlikely that it did. The OEM manual makes no reference to such behavior. So, getting on-track was likely a matter of reading an embedded servo mark or two, tweaking the "fine" stepper and signaling end-of-seek. --Chuck

On 28/02/15 19:17, Mike Stein wrote: I'm very curious what the magnetic trace data looks like, whether you're getting anything valid at all. Do you have access to a DiscFerret, Catweasel or Kryoflux type disc reader? I'd be inclined to do a non-indexed read, then bash together some code to realign things about the "real" index. If you know what the offset is, you can just shift the flux transitions (think "ring buffer" or "barrel shifter"). Cheers, -- Phil. classiccmp at philpem.me.uk http://www.philpem.me.uk/

On 02/28/2015 02:57 PM, Eric Smith wrote: Well, again, I have my doubts. Take early 100 tpi drives, for example, say the Micropolis 1015 SS drives. Initially, they did not "spin" the disk as the door was being closed, which could cause the disk hub clamp to severely mangle the early DSDD disks--and this was before the reinforcing rings appeared. We ran into the problem quite often, with the wrinkled hub area. While a severely mis-registered clamp would cause read errors, I don't think we ever lost data by simply carefully re-clamping the disk in the drive. Note that the early Micropolis units did not use the Shugart "epanding cone" clamp, but rather one of their own design that consisted of a projecting hub piece that fit a recessed piece on the other side of the drive, sandwiching the media between the two. What do you suppose that the mis-registration would have been in that case? I've got several Drivetek drives, including the later Kodak versions, and I can't ever recall seeing the "flag" on the fine positioner move once the drive was on-track. So, while it may have been possible to compensate for eccentricity, I'm not aware that it was done. Since neither you nor I are willing to spend the time to investigate real-world samples, I think the discussion is academic at best. --Chuck

On Sat, Feb 28, 2015 at 11:14 AM, Mike Stein <mhs.stein at gmail.com> wrote: Yes, it did fully closed-loop servo positioning, with embedded servo bursts in all inter-sector gaps. That's why they have not one, but TWO stepper motors for positioning, with a mechanical adder mechanism. The coarse positioning stepper doesn't have the resolution needed for track following. This can also be seen in their US patent 4,630,145: "As the track densities are increased, the misalignment error may approach or become larger than the track spacing. The noise-to-signal ratio then becomes very large and the desired signal cannot be read." "The second stepper motor 46 is stepped by track following control circuits operating in a closed loop." "It is yet another object of this invention to step the head towards the track centerline by a predetermined discrete fine increment equal to the maximum amount of track eccentricity per sector." They didn't do this stuff because it was cool and they wanted to make the drive more expensive. They did it because it was necessary to make it work reliably at all.

On 02/28/2015 04:00 PM, Mike Stein wrote: They were 5.25" drives. Drivetek went bankrupt, but managed to get drives into the Kaypro Robie machines. Kodak picked up the pieces and offered some of their own incarnations--there was a 6MB drive. The killer was that media was fairly expensive (about $15 per disk) and was pre-formatted. i.e., there was no low-level FORMAT utility. The positioner mechanism was a little Rube-Goldberg-ish. Think of a coarse leadscrew on a sprung slip fitting over a stepper shaft. Moving the slip fitting was a fulcrum levering off a fixed pin at the end nearest the coarse stepper. On the other end of the fulcrum was the fine stepper driving the fulcrum back and forth with a fine leadscrew. Somewhat like driving a tapered shim into the works to provide a fine fit. Both steppers have "flags" on the ends, so you can see what's going on. FWIW, I did determine that the Robie disks could be read quite easily in two passes by using two 96 tpi drives, one being offset slightly from the other. --Chuck