Go to Antique Computer home page
back to Stories

700 series had many different Tape Control Units

Attachment 7030_Stretch_Tape_Control

... I wondered if any of you ever thought about how a processor interfaced with a tape drive in the previous [pre IBM 1401] IBM systems.

The 700 series had many different Tape Control Units designed by different engineering groups and a lot of wheels were invented again. Below are the IBM machine types for the Tape Control Units I know of.

These vacuum tube control units were similar but different. As an example, I once worked on a problem on a 753 on our 704 system at Lockheed. The 704 Write End of File command caused the tape to erase and End of File gap then a tape mark character 1, 2, 4 and 8 bit on tape. The 704 could write and read tapes in either binary mode (odd parity) or BCD mode (even parity). When we first got the 704 a programmer had written a print output tape using the odd parity write End of File command.. It wrote the gap and the tape mark but it added a C bit to the 1, 2,4, and 8 bit to make it odd parity. The 704 could read this tape and see the End of File because all it looked for in odd parity was the long erase gap. When the tape was put on the tape to print system the 757 didn't recognize the tape mark because of the C bit and kept on printing old data that was on the tape. I sent in a Suggestion to block the C bit on a tape mark in odd parity and on my birthday June 14, 1957 I received a $75.00 Suggestion Award.

To erase the End of File gap the tape control triggered the same 54 msec Single Shot circuit that was use to write the Load Point gap at the beginning of tape. The 753 had 26 Single Shot circuits to generate all the necessary delays for tape operation. All these SS were controlled by an R/C network with fixed components. As the tubes aged the timing changed. When reading in binary mode it looked for the long erased gap as and End of File. In even parity it looked for the 1, 2, 4 and 8 bit character as the End of File.

Another problem I was having was when a programmer wrote two consecutive End of Files the Single Shot circuit would not recover fast enough to create a long second gap. The second gap was only about half the length is should have been and the 704 did not recognize it as an End of File. I could not find any bad components in the Single Shot so I went over to the 705 computer room and looked at the diagrams for the Load Point Single shot in their 754 control unit. The 754 used two 27 msec Single Shots in series and had no recovery problems.

I punched up a small deck of cards with a program that would write multiple tape marks then rewind and read the tape counting the number of End of Files read. I sent that and a description of the problem to the Suggestion Department in Poughkeepsie. On September 30, I received a $25.00 Suggestion Award.

All these different control units added to the training requirements for CEs. Sometime between the 700 series and the 7000 series someone got the idea to use a binary counter that could be run in either microsecond or millisecond mode to generate all the delays instead of a bunch of Single Shot circuits. I have never been able to find out who or which group developed the first TAU.

Attached is a pdf paper, dated December 20, 1958, telling of the decision not to use the "Universal Channel Adapter (later known as TAU) to connect 729 tape drives to the "Project 7000 (STRETCH)". The document refers to a "SWIFT Tape System". This was the 7340 Hypertape drive announced October, 23, 1961. Since the STRETCH machine was a current mode circuit machine I assume the TAU they are talking about was current mode. The first two 7090s delivered to Sylvania on November 30, 1959 had current mode TAUs in their 7607 Data Channels. These may have been called TAU-1 if they had not been the only ones at that time. The original engineering data flow model of the 7070 used current switch circuits, but subsequent analysis showed that greater economy was to be gained by the use of a family of circuits known as "complementary transistor diode logic" (CTDL), CTDL circuits were used to great advantage in the 1401. I don't know if the 7070 data flow model even had a TAU but the 1401 came out with a CTDL TAU-2. Later on when IBM came out with the low cost 7330 tape drive The TAU-9 was built with SDTDL/SDTRL and could handle 729 and 7330 tape units.

The TAU was a great improvement over what came before and is probably another legacy passed on from STRETCH to what followed. It lived on the the System 360 in the 2803 Tape Control used to attach 2400 Tape Drives to the channels of all the 360 models. There was even a 2804 Tape Control that had a complete Read/Write TAU and a Read Only TAU in the same box. It could be reading from one tape drive while simultaneously writing on a different drive. It was referred to as a Simultaneous TAU and was used on small systems with one channel and was good on sort programs. We had one of these at Western Electric in Sandy Springs, GA. The account CE had removed the front cover to open the gate and change the air filter. He had leaned the cover against the side of another machine close by. An operator came pushing a cart full of tapes down the isle and hit the cover knocking it over. It fell into the 2804 and the corner of the cover raked down the backpanel bending and shorting many pins. It blew 33 SMS cards and took me all night to get it back up.

The 3803/3420 Tape Subsystem was announced November 5, 1970. The 3803 was microprogram controlled with two processors, one talking to the channel and the other talking to the tape drives. This ended the reign of the TAU.

I hope I haven't bored you too much with all this but I guess that's what the delete button if for. It's like having someone to talk to.

Van Gardner