Go to Antique Computer home page
back to Stories
The One Week School That Lasted For Six.
(Event was between Oct 1960 and May 1962)IBM announced a new feature for the 1401 that allowed it to keep on processing while data was transferred to and from Input and output units. Years later this was referred to as "Cycle Stealing" or "Direct Memory Access." The school was a one week school but it straddled a weekend. I left Joyce and two small children on Monday expecting to be back the next week.
I flew into Broome County airport and caught a taxi to the Fredrick Hotel in Endicott, NY. This was my first stay at the Fredrick as my two previous schools I had brought my family and we rented apartments. The Fredrick was an old hotel and the rooms were very small. There was a good side to this as they were too small to put two people in the room. The bad side to it was you had to share a bathroom between the two rooms. When you went into the bathroom you had to lock the door to the other room then unlock it when you left. The man in the other room was not in my class and working a different shift. Many times he would forget to unlock the door to my room when he left and I would have to call the desk and have them send someone to open the door.
The first day of class I was surprised to see Ralph Snow from Huntsville, Alabama was going to be the Instructor. I had made several trips to The Marshall Space Flight Center at Redstone Arsenal in Huntsville to assist Ralph on problems with 727 tape drives on the 704 system. After a couple of trips Joe Sipple ask me to come over there and spend a week teaching them how to PM their tape drives.
Ralph did an outstanding job with the class. He had correctly figured that most people would have forgotten, a lot of what they had been taught, about the I/O circuitry of a 1401 two years before. He started by giving a fast paced lecture of the old circuitry then began a detailed explanation of the process overlap circuitry. It was one of the best schools I ever had.
The day before the last day of school the Manager of Education ask to see George Bozdog and me after class. George was from Cincinnati, Ohio. When we got to the managers office there were several people waiting for us. They were from Customer Engineering Tech Ops, Manufacturing and Product Engineering. They were looking for a couple of volunteers to install the first attempt at a field bill or materials for the process overlap feature for the 1401. We had been recommended for the job and they had already contacted our office for permission to ask us to volunteer. Naturally, Endicott would pick up our pay and expenses while we were there.
They put the old full court press on telling us how important to IBM it was and how we were the only ones for the job. Their estimation was that it would take two men about a week to do the job.
I didn’t really want to be away from home another week but it did sound like a good chance to get some feedback into the instructions for the field bill. So, we "volunteered".
We were given a preliminary copy of the B/M instructions and we all walked across the street to the plant to see the area where we would be working. I was surprised to see how many 1401s and test cells they had but a lot of them didn’t have anyone working on them. We were given ID badges and told to report there first thing the day after the school was over.
George and I read the instructions and found nothing unusual about them. They were just standard wire wrap, card plugging and cable running instructions. There were actually two different sets one labeled for man number 1 and one for man number 2.
Our school ended and we reported to the plant the next day. The first thing we did was to check the machine out to make sure there were no bugs on it. We each took a set of the instructions and started to work. There was trouble almost immediately. We were both trying to work in the same area of the machine and there was not enough room for two people. It seems that the task for man 1 and man 2 had been allocated so that each man had the same total work to do. There had been no thought given to where they were working.
We decided to split the work between the front and back of the machine. One man would take one set of instructions and go through it looking for work to do on the front of the machine. He would do that work and mark it complete assigning a sequential step number to each section of work. The other man did the same thing with other set of instructions on the back of the machine.
When they had each finished they swapped instructions but stayed on the same side of the machine until finished again. This is when it was realized that there was a lot more work done on the front of the machine than the rear. When the man on the rear finished there was a good bit of work to be done on the front. We divided the remaining work between frame 1 and frame 2 and both worked on the front when possible.
After we finished the installation of the change the real fun began when we turned the power on. There was a lot of noise and smoke from the TAU (Tape Adapter Unit) gate. This machine had the new Nand (Negative And logic cards) TAU. These cards operated from different voltages and used different card pin locations to receive the voltages. The people who wrote the change had not thought of this. The wrong voltages on the wrong pins blew all the Overlap cards that had been added to the TAU gate.
To correct this problem we needed to get rid of the voltage busses feeding the wrong pins to the Overlap cards. I took a 1/8" pin punch down to a grinder and made a chisel point on it. Placing the chisel on top of the voltage buss next to a pin and giving it one hit with a hammer cut the soft copper buss. Doing this on each side of each pin removed the buss. Using some #22 wire I wire wrapped a buss to the correct pins. After replacing all the blown cards we were able to power up.
. (The CE tool bag did not contain a chisel so a small chisel was included in future B/M (Bill of Materials).
The bugs were very difficult to trouble shoot. The main problem was when running in overlap mode, and the machine taking break-in memory cycles. Prior to the process overlap feature the 1401 only did one thing at a time. It would do sequences like: read a card, calculate, punch a card, print a report. With the overlap feature it could read, calculate, punch and print simultaneously. These processes all ran at different speeds and an overlap cycle could occur at any time. This meant no matter what circuit you used to trigger your scope an overlap cycle could break in and what you would see was the overlap cycle instead of the one you wanted to see. There was a switch on the auxiliary console that would disable the overlap feature. We found it was easier to turn overlap off and shoot the normal bugs before tackling the overlap bugs. (I had been working on 7094 systems, which also had a switch to disable overlap, and had learned the plant always started debugging with overlap off).
We were stumped on a particularly hard overlap bug when I had a brainstorm. I had been trying to use the Address Sync pulse to trigger my scope to see the cycle I wanted. This worked ok unless the next cycle was an overlap. There was a circuit in the 1401 called "Delta Overlap" which was used to make a request that the next cycle be an overlap cycle. I went down to the parts crib and got an eight position Garvey SMS card socket. I used some of the cards we had removed from the machine during the change to build a circuit that would block the Address Sync pulse if the Delta Overlap circuit was active.
This worked great for most bugs but some affected the machine so bad it could not loop back and do the operation again. Bugs like these really slowed you down. You got one try at execution of your loop. Then, you had to hit the start reset button, turn the mode switch to alter, press the ISTAR button, press the start key, return the mode switch to run and press start again. All this got you one more look at your scope.
I added some more cards to my socket and built the circuits that would activate circuits in the machine just like you had pressed all those buttons in sequence. It worked great. We finished shooting all the bugs and cleaned up the machine. Then we wrote up all our recommendations to improve the change and reported to our project manager. He was surprised and very pleased. He invited us to attend a meeting with him at the Glendale Research Lab first thing in the morning.
Arriving a Glendale we were escorted to a large conference room that had at least forty people there. We were introduced to them and they were told what we had been doing. Then came a bombshell. We were told that the subject of the days meeting was to decide whether or not to cancel the field installation of the process overlap feature or to cancel the feature totally. As the people in the room told of the problems they were having we heard of 125 systems that was on the production floor that had not been debugged. The line techs could fix the bugs until overlap was enabled but they could not fix the overlap bugs. The Engineers had to go to the plant and fix the overlap bugs. They said they could not sent Engineers to the field to fix all the overlap bugs that might occur.
Some of the concerns brought up by different people were:
Could the average CE in the field install the change?
Could he debug the change? (Could a CE who worked on a 1401 once in a while do as well as a line tech who worked on the same type machine all day?)
How long would it take him to install it?
During a conversation, about how long it should take, I heard someone say, "Don’t forget to add all the time to remove previously installed engineering changes and reinstall new versions to make room for the overlap circuitry in the machine. We had some very early serial numbered machines at Lockheed and we had installed all the EC’s as they came out. This added up to 125 man hours to uninstall the EC’s and install the new versions. Then we could start the overlap change.
Endicott was building a repackaged machine called a 25000 series that had incorporated the all the new version EC’s. The overlap feature could be installed on them without any preliminary rework.
After they all finished talking to each other they turned to us with a load of questions. We had been the first field personnel to install the feature and the first to trouble shoot the bugs. They were very interested in the circuit I had designed to sync my scope and restart the machine. I told them I was going to submit it to the suggestion department for consideration. They told me to be sure and mention in the suggestion that I had divulged the idea to them. (October 31, 1962 I won a $50.00 Suggestion Award number D03556 for my Process Overlap Address Sync service aid.)
They asked what we thought about CE’s installing the Change. I told them my big concern was all the old machines in the field that had to have all the EC rework done first. The CE’s had too big a workload to be pulled away from the rest of their territory for 125 consecutive hours of EC rework. My suggestion was that for the old machines a plant team should be sent to remove the old EC’s then install all the new ones using a composite wiring list to do them all at once. That was the way the plant upgraded a machine when it was returned to them.
When the subject of CE experience came up I told them that in Atlanta the only accounts that had more than one 1401 were Lockheed and Southern Railway. All the rest were scattered through individual territories. If a CE’s 1401 never broke he didn’t get any experience until it did. I had tried to get a group of 1401s made into a territory but local management would not buy it. There were two large offices in the country making 1401 territories with great success.
They wanted us to stay and do another machine to check out the corrected and updated instructions. After hearing the seriousness of the problem how could we refuse?
The instructions were rewritten and the B/M assembled and we started on our second machine. The first thing we did was to cut the buss bars out of the TAU gate. It caused so much damage last time we didn’t want to go through that again. We made a note to have that moved to the first item on the change. We figured you usually didn’t accidentally skip over the first item.
The change went a lot faster this time due to the new split in work assignment and the fact we had done it all before. When time came to power up the machine came up but with a few bugs. Some of these were the normal kind of thing you see after a large change like, loose cable paddle connectors, bad cards, bent pins, etc. Some were due to some of the corrections we had written on the old instructions had not been incorporated into the new instructions. We could not understand that.
The next week I spent some time with the Engineer that was going to do the official design of the circuitry I had built to make the machine recycle when shooting overlap bugs. He had decided to keep the cost of the 1401 down and not use up any spare card sockets he would make an external tool. We called it a recycle box. If an installation had multiple overlap 1401s they would only need one tool. It was about the size of a shoebox but much heavier. It had a cable that plugged into the 1401 for power and signal connections. It was really a good idea from a money saving position.
George and I were ready to go home. Every weekend when I called home my wife would say, "When are you coming home?". I had been saying one more week but enough was enough. The project manager came in to ask us to do another machine. While he was talking to George I took a quick look the latest instructions and there were still some of the original mistakes we had written corrections for. I had had enough and told him I had to go home and George told him the same thing. He said he understood and thanked us both for the work we had done. We left the next day and when I got home I had been gone almost six weeks.
Lockheed had ordered the Overlap feature for our largest 1401 system. It was also one of our oldest machines. There was 125 man hours of EC removal and reinstallation before the overlap change could begin. The plant sent two wire wrap men and a trouble shooter to help with the change. We furnished two other 1401 trained men besides myself to work with them. We made sure a local man was always there for logistics and to deal with the customer. It took two weeks around the clock to install and debug everything.
Johnny Malone was a 705 Customer Engineer that had become a Systems Engineer. He was assigned the job of writing a peripheral program that would do card to tape, tape to punch, and tape to printer on two printers. The machine had an RPQ that installed a second print buffer gate in a frame like the 1406 memory used. The program requirement was that it could drive the reader, punch and two printers at full speed.
The card reader could read at 800 cards per minute. The punch could punch 250 cards per minute and the printers could print 600 lines per minute. Johnny explained his idea of how his program would handle this. He envisioned a software Ferris wheel where each seat was a time slot to service one of the I/O devices. This also had to include the four tape drives.
I worked closely with Johnny when he began to debug his program. He didn’t really need any help until he got all the devices running and he ran into some speed problems. Sometimes a printer or the reader punch would slow down then speed back up. We would use the address sync pulse and a scope to take time measurements between different routines to find where the trouble was. It didn’t take too many days for Johnny to have the machine running full speed. It was a very successful program that was used until it was replaced by a 360 Model 30, that eventually ran five printers and a reader punch.