On 01/03/2017 10:58 AM, Cory Heisterkamp wrote: Most likely the same issue as the G-15 we had. Dust was allowed to get into the drum area and pack under the heads. Probably if you pull the heads and clean them, it will restore clearance. Of course, the bearings may be bad, or will have to be replaced anyway as the grease may have hardened. If the drum can be set up to run true again (may need attention to bearings) then I think setting the heads up won't be that tough. I suspect it is done with a feeler gauge, this is low-resolution stuff with large gaps in the heads, so the heads probably run with a gap of at least .005" (~ 0.1mm) (I'd think, without actually knowing). Well, it could be. That sounds really close for the vintage involved. So, maybe the drum or oxide has swelled. Anyway, if there is much damage to the oxide, it may not make sense to try to repair it. if the heads that jammed it left divots in the drum, or the surface is uneven (likely if swelling actually occurred) then it may require extreme efforts to repair. Why not just replace the whole works? If you are going to replace the long lines with electronic memory, doing the short lines and the clock track should be trivial. I think a mid-sized FPGA could do it all quite easily. My guess is that if the surface is uneven, it may not read back data reliably. The high spots might be fine, the low spots will have dropouts. This is all assuming swelling was the culprit. It is also possible that machined parts suffered stress relief over the years. Wrought metal has stress imparted to it when rolled, and then machining will partially relieve the stresses, causing warpage. The warping continues over time. To eliminate this, critical parts are machined close to size, heat treated to relieve the stress, and then finish ground to exact dimensions. It is possible some of the stress wasn't relieved during manufacture. And, nobody expected a 195x machine to be running in 2017, especially as anybody in the computer business knew those transistors were right around the corner, and would almost certainly replace tubes. Jon

On Tue, 3 Jan 2017, Cory Heisterkamp wrote: We relocated some of the heads on our second LGP-30. This is not difficult. We put 3 or 4 layers of 3 micro-meter kitchen aluminium between head and drum. If I remember right, the distance is 12 micro- meter. Maybe, that we made a fine adjustment 3 or 4 layers with the oscilloscope, to get the same output voltage at the head, than the other heads. Maybe that 3 layers of 3 micro-meter Al have something more than 9 micro-meters. If I counted right on every head bar are 3 unused positions as spare tracks. > For kicks, I tried to use a piece of cheap (=thin) (0.004?) notebook paper > as a feeler gage to see if I could identify the offending heads prior to > support removal. This was a no-go as clearance was too tight. So, is it > true these ride 0.001? off the surface? The (german) maintenance manual whose scan is on our website speaks about 12 micro.meters. There are 64 data tracks, 4 timing tracks (including the main clock at the rightmost position) and 3 tracks for the registers. Most problematic are the timing tracks. If they are faulty, there is no simple solution to change the heads, because there is no provision to write them in the machine. If there are only few tracks defective, and not exchangeable, I would try to get most of them in the lower adress-room of the machine, and at least the last track. The first 3 tracks are used by the "operating-system", program 10.4, and the last track is used by the initial loader, that loads 10.4 itself. But timing tracks S1, S2, S3 and the registers themselves must be functional. If this is not, a semiconductor drum-emulator ist indeed the only solution. If you lift all the head bars 1 or 2 mm, then the drum can rotate even with this fake. Klemens > Any input is welcome. -C -- klemens krause Stuttgarter KompetenzZentrum fyr Minimal- & Retrocomputing. http://computermuseum.informatik.uni-stuttgart.de