Alpha Pictures

The drawing above attempts to reproduce a cover plate from an Alpha. AMM autospeed boards were mounted three abreast in a U-shaped piece of sheet metal. The shape was deep enough that the rear which held the boards (about four inch square) was aligned with the rear of the Alpha. This would have left an unsightly gap on the front side. The cover plate disguised the gap and provided branding. Two of the U-shapes are visible above the Alpha in the picture below.

This photo is from the May/June 1979 IPSA newsletter. Rohan is the rear figure in a white shirt (he worked in Montreal sales & support in 1979); Dana is the foreground figure in a striped T-shirt (he was a senior communications technician from Toronto).

The photo shows a typical Alpha field installation. A 19inch relay rack about five foot high held the equipment. On top (near Rohan) are low speed (300baud) answer-only dial modems. I suspect they are General DataComm 113A two card modems rented from Bell Canada. They are housed in a light-coloured enclosure. Below that is equipment which is obscured by the long cables. Other elements in the picture suggest that there must have five AMM autospeed mini-boards mounted three abreast. (There are five thick round cables next to Dana.) The thinner cables would be RS232 cables connecting the autospeed boards to the 113A modems. A few cables would have connected to hard-wired terminals in the office.

Below the unseen autospeed boards is an LSI-2 with a main and an expansion chassis. The two upper boards in the main chassis are processor and memory (16k words of 16 bit width). The third board must have an SMC (Synchronous Modem Controller) paired with either a filler or a UMM (Universal Modem Multiplexor). (I suspect it was a filler as I can't see the flat cable associated with a UMM.) The bottom two boards in the main chassis and the three boards in the expansion chassis are AMM (Asynchronous Modem Multiplexor) boards. The expansion chassis AMMs clearly show the blue plastic covers which covered the wiring and revealed ill-seated rear connectors.

The network connection between the Montreal SMC and the second SMC in Ottawa was provided by Bell Canada Dataroute service. Just above Dana's head is a light-coloured box with a dark rear panel. This is a Computer Transmission Corp DCTE (short distance synchronous modem). It connected IPSA Montreal to a nearby (less than five kilometres) central office with a CTC (later Amdahl) synchronous multiplexor.

The above picture of an LSI 2-2 Programmer's Console was taken from page 3-2 of the Computation Automation LSI-2 Series MINICOMPUTER HANDBOOK. The 16-bit display was used by the concentrator to indicate whether High Order line one was communicating with the adjacent node. The hexadecimal keypad on the right was not normally used by the concentrator. The exception was the "direct bisync" version in which a digit from the keypad and the interrupt button indicated what operation was desired: print, tape read, tape write. restart interrupted tape read/write.

3705 front panel is illustrated here. The two 18 bit displays at the top could be set by software while the 3705 was running. (Software display is shown when DISPLAY/FUNCTION SELECT rotary switch in the centre selects a one of the six functions. Definition of the six functions was left to software.) In other positions, 3705 hardware can display storage or sample an input register while the program is running. (This is most unusual for an IBM console.) STATUS is mainly of interest after a hard stop or if single cycling the machine. TAR&OP REGISTER setting might be of interest to a programmer if the machine had hard stopped. TAR is the Trailing Address Register and contains the source of the previous instruction.

The MODE SELECT rotary switch was used in software debugging. It provided instruction step capabilities and two address compare options (stop and non-maskable interrupt). With MODE SELECT in PROCESS the address compare result was still available for use with an oscilloscope.

The DIAGNOSTIC CONTROL rotary on the right was mainly of use to someone repairing the machine.

The five STORAGE ADDRESS/REGISTER DATA rotary switches near the bottom had two uses. One was in support of various hardware facilities such as address compare, storage inspection, output register alteration, etc. The second was that these switches could read by software to provide parameters to the six functions.

The controls and lights at the lower right are similar to the found in the 360 and early 370 models. In the 360, three rotary switches selected an Initial Program Load device from which a program was bootstrapped. The LOAD and INTERRUPT buttons appeared in exactly that position in the 360 models. The LOAD, WAIT and TEST lights were common to the 360. The power control was slightly different as the 3705 was considered a peripheral for purposes of power on/off. With the power control switch in REMOTE, the central processor controlled 3705 power just like any other peripheral. The local power controls were only available with the switch in local. Mainframe power control included a red circular pull switch labelled Emergency Power Off. Pulling it immediately powered down the mainframe and devices whose power was controlled by that mainframe. The 3705 as an I/O unit lacked an EPO.

The two counters in the upper left corner were common to both mainframes and peripherals for at least a decade. They were used to measure the number of hours of operation of a machine. The upper counter with value 9446.67 measured customer time. When the IBM Customer Engineer operated the keyswitch between the counters, the lower counter (reading 0.42) measured time. (On some machines such as the 370/145 operating the keyswitch also changed the behaviour of the machine.) These counters were used by IBM to bill rental customers for excess usage beyond 176 hours per month. When leasing replaced rental, these counters were dropped from the IBM design.

Rear view of 3705 control panel is shown here. The picture does not show enough detail of upper display wiring to accurately reveal the clever wiring scheme. The yellow wires are signal wires for both upper and lower rows of lights. A black diode connects a yellow wire to a bulb socket. The upper row has a ground circuit which is different from the lower row's ground. The black wires with numbers starting at label 218 connect adjacent bulb sockets in the lower row. This chain leads to lower row ground.

click for larger image (260KB)

These photos exhibit a single bulb assembly from a 3705 control panel. This bulb technology was used in the 360/40 and larger models. It continued into the 370/145 and 370/155. It was discontinued with 370/158 which used a CRT console. It also appeared in other devices including the 2860 and 2870 channels and many peripherals. The driving electronics in the 3705 were far better than in earlier machines. Voltage was only applied for about 20% of the time using a 60hz chopping circuit. This gave much longer bulb lifetimes than in System 360. (A 360/50 with a population of about 150 bulbs suffered about one failure a week.)

The grey ring at the top protruded from the panel and gave a purchase for removal without tools. A transparent plastic cylinder joined this ring with the green plastic base. A tiny bulb with a black base can be seen below the grey ring.

This picture shows two connectors from the 3705 control panel. The dark green plastic moulded shell was used in many 360 and 370 devices (both SLT and MST). Printed circuit cards connected to the mother board with 24 pin connectors like these. One, two or four connectors were used depending on the height of the card. These connectors plugged into the top of a 3705 card. B3U4 label is the plug co-ordinate. B3 selects a board (CCU extras); U4 selects card U2 third plug from the top.

This is the power control connector for the 3705 console. IBM machines used relays for power control and so higher currents were required than in signal connectors. (These relays made IBM machines quite sensitive to single cycle power outages.)

This shows a 3704 (with protruding control panel and table top) and a two frame 3705.

This picture from the University of Hawaii was taken before September 1975. The missing floor tiles and open machine (370/158?) behind the 3705 suggest the picture was taken during installation. The grey cables appear to be channel cables connecting I/O devices such as the 3705 to the mainframe.

Inside the 3705, yellow wirewrap above the power supply enclosure is the B3 board. The 3705 power supply was more complex than in some machines as +12V and -12V were required for the RS232 interfaces in addition to the -3V for the MST circuits. (Note the vertical capacitor array near the centre of the machine.) Two small swingout frames are visible. One of these contained the storage (core in this model).

This photo was provided by the IBM-Collectables Gallery. It shows several MST quad cards. The 3705 was constructed almost entirely from quad cards (a Type 1D line adapter had one quad and two double cards). The dark plastic structure at the top of each card is the extractor.

This photo which has been kindly provided by the Gallery of Old Iron shows an IBM inter-board connector. In the 3705 they were used to connect adjacent boards. They could be stacked and were thus more difficult to replace than MST cards. Unlike the control panel cables shown above, the conductors were small solid wires. They are visible in the image. Alternate conductors were grounds.

I had the misfortune of having an intermittent fault in such a cable on a 3705 which was an important node in IPSANET. The faulty wire was the lowest speed business machine clock. As the attached lines were synchronous and full-duplex, the internal 3705 clock was not needed for normal operation. Error recovery from situations such as carrier loss or link reset required this clock signal. I procrastinated complaining to IBM for almost a year as I knew the node would have to be out of service for an hour to replace the cable. (With the 1980 routing scheme and extra high-order ports taking a node out of service was less disruptive.)