A Technician's View

I mentioned that I had a disagreement with an news article that reported that people associated Apple with innovation and IBM for something like reliability.

Having delivered opinions on Apple, I start in on IBM.

I do realize that the GMR head on hard drives has been an absolutely essential development in hard drive technology. But IBM has a poor record of technical developments - at least from what I have seen.

If you read "When Computers Went to Sea", one of the devices mentioned is the IBM disk file that went aboard the USS Wasp. Storage technology of the day was by dancer arm tape deck.

Searching for records on a reel of tape was very slow. An aircraft carrier may have over 100 contacts - aircraft and surface units - to keep track of. The computers used were limited to 32K of memory. It took two to three of them to run the NTDS (naval tactical data system) software and have a bit of storage for each track on the system. The track buffer was pretty limited, so something more than a tape deck was in order.

Enter IBM and the 1311 (militarized) disk file. It held two removable packs with 10 platters each. The read/write heads operated in pairs to cover both disk surfaces. Latency was not too bad, and the beast used a high speed interface - 250K words/sec. The computers clocked at 1 MHz, but the clock got divided into 4 phases, so execution time was from 4 uS to 48uS depending on the instruction.

While stationed at Dam Neck, Several significant thing s happened to me. First, I really irritated our E-9 chief. This was not smart. But we had received a sympathy card from his last division before he arrived, and he was just bad news.

Secondly, we got the IBM1311 off the Wasp. In a somewhat discarded (we used another word) state. One disk pack was missing and the other had been dropped. It still worked, but the four lower heads were a blur as they ran up and down the bend in the lower platters.

So we had this marvel of technology and the chief gave it to me, with dire warnings about what would happen if I didn't get it running.

This was model XN-1, serial number 1, by the way. Real bleeding edge stuff for 1968 when they built it. It used IC's, but not standard ones. They were not DIP shaped, had crimped on metal covers, and were made by IBM - don't bother looking for substitutes.

The logic planes were hinged to swing out for access. The IC's were in sockets which made contact with copper foil in multiple layers under the top surface. The contact pins also led out the rear of the board and could be used for wire wrap.

This was handy, as the original foil was mostly unused by the time we got the 1311. Included was a handy cutter that fit around a contact pin. When run in to where the cutter bottomed out on the pin, it left the contact mechanically held in place, but electrically isolated. If you worked on one, you got lots of OJT in wire wrap.

Even more clever was how the 1311 moved the heads. They were all in a frame so they went in and out together. The frame was pushed by hydraulic oil - at 90 PSI. There was also a rack and pinion gear on the frame. The rack moved with the frame, and the pinion gear stopped it.

Actually, it was an arm with a steel tooth that stopped it. The arm, also hydraulically controlled, dropped into the pinion gear to stop it turning, and somewhat hopefully leaving the heads on the right track.

On a big move, the arm could skip over the tops of the gear teeth and let the frame go all the way forward. This made a horrible mess. a the hydraulic piston seal could blow out, dumping oil all over.

The crowning glory was the power supply. It was the first switcher. IBM called it time regulated, but it was the prototype SMPS.

Sigh, if only FET's had been available. Instead, they used SCR's. The transformer had a center tap to ground on the primary side. Two SCR's alternated dumping charge from a pair of capacitors through each half of the primary. The charge on the capacitors was controlled by another pair of SCR's.

The whole mess actually used more SCR's to deliver charge from the secondary windings to the LC filters. The time each was gated on controlled the output voltage, which was where IBM came up with the time regulated descriptor.

The main frequency was 20 KHz. The control circuitry was located on several PCB's. These used conventional discreet parts, but IBM potted a tough black rubber on top to make the components invisible and unreachable. The control circuit was a black box on the schematics.

The power supply was not reliable. Every now and again, an SCR died. It always went short to the gate, and always took out one of two control boards. We referred to them as either the $1100 board, or the $1600 board.

We never did find the source of these failures, although I found out a lot about intermittent problems and overbearing chief petty officers. It took a whole year, but I got rid of the 1311 and got the chief reassigned to software only. It only cost me one article 15 charge and some extra duty.

Beyond the 1311 disk file, I should only have to mention the 5150 to have people recall what a botch IBM can make of something.

Blog entry information

Last update

More entries in General

More entries from beenthere

Share this entry