Tom Jennings, 13 February 2005
The tape drive wasn't too hard to bring up, no deep debugging dead-ends, no particularly serious problems, though I did snap the tape a bunch, but I tested with the scroungiest tape I could find -- one marked "MISSING BOT".
This is the post-event email I sent to the classic computer list last night. It does contain pretty much everything I did. I've added photos (not allowed on the list) of things referenced in the text.
The home-made manometer. It's pretty simple, a transparent tube in a bucket of water. Tube is marked off in inches (I needed only 10" and 20"). The tube is really long because it's for another project and I didn't want to cut it up.
This is the new vacuum column seal made of RTC silicone rubber applied with my finger. See the text.
From email@example.com Sun Feb 13 21:55:44 2005 Date: Sun, 13 Feb 2005 21:54:49 -0800 (PST) From: Tom Jennings
To: "General Discussion: On-Topic Posts Only" Subject: NOVA4/X -- running! Must be the alignment of the planets, favorable for old computers... I got the 6023 75ips tape drive working. Pretty straightforward debugging. With Bruce Ray's hint, the first thing I checked was the vacuum switches (there's five, one for "master vaccum supply" and then two each (low side, high side) in each column as limit switches) and three of them were bad. Found that out a few weeks ago. Swapped switches around with the other drive, didn't "run". So I got scientific (sic). I made a manometer out of a long plastic tube, a bucket and some automotive vacuum hose, some masking tape and a tape measure. Attached each vacuum switch to it with a tee, and hose in my mouth as vacuum supply and adjusted each switch as close as possible (10% spec). One was way off, the "master supply" switch has a different setting and I had mixed them up, from the donor. It attempted to load, but acted strangely, sucked tape into the buffers (vacuum columns) then THWACK the tape, then go slack. Twice it snapped the tape! (I'm testing with a known-bad tape). Yesterday and today I spent about two hours tracing through logic to see why the spindle motors wouldn't move (not even for manual unload). Traced right up to the driver amp -- no power supply! Duh. ... turns out the DC circuit breaker popped. Sheesh. While the breakers were popped, during testing I initiated LOAD, and fondled the reels such that it the servo was happy, positioning the tape loops in the buffer. Limit switches etc all worked OK. During this though, I found that the lucite cover over the vacuum columns wasn't sealing; it'd warped, lifting up at the corners, leaking a lot of vacuum. For now, I taped it shut. Resetting the breaker a few times, I determined that a logic problem causes the two spindle motors, in the middle of the load sequence, to rotate at high speed in opposite directions -- drawing HUGE currents, tug of war with the tape, which snaps or wrenches into the spool. The takeup side decided that LOAD succeeded, and was searching for BOT. The supply side decided that LOAD failed, and was rewinding! More servo problems. Turns out it was another vacuum leak! The lucite cover seals one column from the other, and it was leaking. I made a seal by wiping silicone rubber in a film about 0.010" thick with my fingers, tapered to match the lift of the warped corners. Tape now loads! Now I tweaked the final servo pots, it all behaves properly. Now that I have actual DG terminals, I plugged one on instead of minicom on a laptop. Display works, no keyboard. Terminal wants CTS or RTS asserted. Fixed! Reset, self-test, try to boot the tape: !100022L Nothing happens. Long story short: the cables to the disk and tape were plugged in wrong! Ouch. Luckily, identical paddle card wiring and pinouts. I'd removed and marked them, but managed to install them wrong anyways! Off-by-one problem. Power up, (disk still off), self-test, boot: !OK 000000 !I !100022L DTOS (diagnostic tape operating system) boots! I typed ACCEPT, it ran some tests, then many, many tape retried, !FATAL ERROR (tape read error). Clean head and capstan. Reboot, ACCEPT... runs! but I halted it, powered on the 6070 disk, (cringing awaiting head crash.... head load OK!), reboot DTOS, run ACCEPT... it discovers the tape (duh) and disk! ACCEPT (accept list of equipment found, run all applicable diags one pass) ran... halted when I accidentially hit a key during GALLOPING ROWS mem test? Seems odd... Reboot, repeat, hands in pockets... runs many tests... I leave running and go eat dinner. An hour later, I find the tape slack. There's a crease in the tape and a transparent spot (!), but no sign of a jam, I think the transparent spot triggered BOT (likely an error). On the screen is the DTOS exec saying "NMMD x V 3.0" which is where it loads the tape file; it was searching... so without reset I reload the tape, press ONLINE... and it starts searching again! It gets to approximatel the same thickness of tape... JAM! WHACK! power goes out! 1) The tape caught somehow, likely that crease, cross-wound on the capstan. These reel motors are POWERFUL, likely break your finger. That's what it tapes to swing a 5lb reel of tape at 75 ips. 2) The jam causes motor current draw to skyrocket. I've got the terminal, CPU, tape drive, and disk, all plugged into a quadbox extention, plugged into the (heavy duty) outlet strip on my workbench... the little breaker on that tripped. Well that was stupid. The disk heads were loaded at the time, but wasn't operating (it needs formatting anyways) and I double-checked that the head-unload mechanism (monster capacitor + N.C. relay) worked, so it's fine. I checked. But I likely ruined the already-creased DTOS tape (will find out Tuesday night) and the other DTOS tape Bruce lent me won't boot; read error. But it was a lot of progress anyways! I took some pictures, like the goofy manometer, I'll put 'em up this week. From firstname.lastname@example.org Mon Feb 14 09:09:23 2005 Date: Mon, 14 Feb 2005 09:08:21 -0800 (PST) From: Tom Jennings To: "General Discussion: On-Topic and Off-Topic Posts" Subject: Re: NOVA4/X -- running! On Mon, 14 Feb 2005, Tore S Bekkedal wrote: >> It attempted to load, but acted strangely, sucked tape into the >> buffers (vacuum columns) then THWACK the tape, then go slack. > > That's the same thing happening with the NORD-10's drive. Although it > fortunately cuts out in time, before harming the tape :) Well I learned a lot when the servo motors were turned off: That let me verify that the limit stops worked precisely right. (For those not familiar with vacuum column tape buffers, there's a deep rectangular box, open at one end. At the bottom of the box is vacuum generated by a pump (a noisy pump). There's two boxes, one upside down, for each reel. It sucks a loop of tape down into the box. Picture a loop of tape sucked into the box, but held so that it goes half way; it wants to suck all the way to the bottom of course. Then...) For each column, there's two limit (vacuum) switches, one near the open end. The switch near the bottom should have vacuum on it (tape blocks off the bottom of the box), the one at the top should not (tape is sucked past the switch). That's the limit stuff. The servo loop is actually straightforward, sort-of. In addition to the limit switches, the column has a row of small holes into the column box, which on the backside go into a manifold; as the tape moves up and down the box more or less holes are passed by the tape. The holes towards the bottom of the box apply vacuum, the ones above the tape loop are at atmospheric pressure. Therefore, given a whole bunch of givens (pump capacity >> flow of the little holes, etc mostly static physics) the amount of vacuum (OK pressure you pedants) in the manifold is proportional to how far down the column the tape is! Then there's this hokey-looking (but reliable) piston/lever/spring thing that bends a strain guage, sniffing the vacuum manifold. This produces a resistance proportional to the amount of tape in the box. This is the feedback for the 10-ampere at 12-volts opamp that drives the tape motor. Linear! Heat! Tape-snapping torque! Dumb as a bag of hammers. It really needs some fast and sensitive current-limit (rate ramp-up) to avoid tape-snapping. Its the sort of thing you make a microprocessor do. Ahem. Anyways, the point of this: I did not do this step, but you could: manually arrange for the right amount of slack tape to fill the buffers half way, and tape down one of the reels. UNPLUG THE MOTORS from their drivers. Go through the "LOAD" sequence, with one hand on the free reel, and position the tape into the columns correctly. This is fairly easy. Now you've isolated the free-reel's servo system for open-loop testing: you can put a DVM on the feedback resistor strain guage thingie and rock the free reel back and forth and see it produce an error voltage. Or better yet, put a "big" resistor in series with the motor to limit it's torque to something absurdly low (so it won't tear all the skin off you hand and snap the tape). The D.G. drive wouldn't mind this, other designs might get upset. DG's tape servo is simply a gargantuan discrete opamp, all linear and ceramic .1-ohm, 50-watt resistors. No PWM, no nothing! Many watts! Tres simple. I wouldn't have known to do this at the time, but with the system LOADed, and the reels taped down into place, motors off, physically checking for vacuum leaks (pressing covers, wiggling hoses, etc) would have produced a BIG feedback signal change. This would have been Wrong and a big hint on what to fix. Oh yeah, I forgot to say, the hoses that pressed onto the plastic vacuum switches were stretched out and likely leaking; they were loose, on and off. I cut off .25" and fixed that. It was an accumulation of problems. The other end of the hose, that pressed onto a brass nipple, was NOT stretched out, so it was a plastic::plastic problem, probably one of them outgassing etc.