On Thu, 30 Jun 2016 16:52:17 -0700, Tom Brennan wrote: >Way later than the 1950's, a tape operator called me to watch a Storage >Tek reel drive do something strange: He mounted a scratch tape and the >job took off spinning the reels so fast that soon the metal capstan got >hot enough to melt the tape, basically destroying both tape and capstan. > Then he pointed to the drive just to the left that had failed the same >way a few minutes earlier. > >Since the tape was moving so fast with no apparent i/o breaks, I thought >I could reproduce the error by running a big IEBGENER with a large >blocksize and BUFNO=255. So I ran that, and sure enough ruined yet >another capstan! I think they were about $900 each, but at least we >found the cause. > Ah! Black Team at work!
Did StorageTek fix it, perhaps by capping the duty cycle? It's hard to imagine that this was purely a matter of saturation. I'd expect more typical jobs to reach perhaps 75% and just take longer to melt the tape. Perhaps the stop-start sequence exercised the pneumatics and supplied a cooling airflow. CDC 6400 FORTRAN provided low-level I/O facilities. With the analogue of BUFNO=2 in my FORTRAN program I could keep a tape moving nonstop. The usual consequence was that the operator perceived this as "runaway tape" (all record gap) and cancelled the job unless I had provided special instructions. Never overheated. IIRC, those same drives did high-speed rewind leaving the tape in the vacuum columns, the capstan driving the tape and the hub servos controlling the reels. The heads were elevated from the tape. A photosensor detected when the tape got near the hub and slowed down the process to finish at normal R/W speed. (I may be confusing manufacturers.) -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
