April 23, 2024

Watchever group

Inspired by Technology

History of the Computer – Cooling, Part 2 of 2

Computer manufacturers specified the heat load required to be handled, the required flow of conditioned air, and the temperature to be measured at the cabinet entry points. Most systems were installed on false floors, with a space of about 1 meter or 3ft below the floor. This allowed the air conditioning to be supplied under the floor, and exhausted through the false ceiling. This space was also used for cabling the systems. An air lock was often provided for entry and exit.

Other components of the computer system, such as disks, also generated heat and had to be considered. Tape drives were particularly large producers of heat. It was common, on large scale systems, for the tapes and disks to be installed in a separate room, or partitioned off from the mainframe, so that air conditioning requirements could be better controlled. (Printers were also partitioned, but mainly for noise, dust, and publishing requirements.

For instance disks are sensitive to rapid changes in humidity, as condensation forming on the disk surface can cause head crashes, due to the small height the head flies at. Tapes produce a lot of heat due to the powerful motors they use, but are more tolerant of temperature changes.

Integrated circuits, or chips, introduced at the end of the 1960s, produced a similar escalation in heat generation and tolerance. As they became more sophisticated, and LSI (large scale integration) was common the escalation continued. This problem is still with us today – have you tried using a laptop on your lap lately? Apart from ever more sophisticated fan design, coupled with heat sinks and air flow funnels or plenums, another approach has been used from time to time. This is the system we mentioned above, water cooling.

The computer user already provides a close tolerance power supply, sometimes at different frequencies, such as 400Hz instead of 50 or 60. He must also supply air conditioning with temperature and humidity gradients specified within tight limits. Now he is required to supply chilled water, also to tight tolerances.

The computer logic circuits were assembled on multilayer PCBs (or PCAs – Printed Circuit Boards or Arrays) which were ‘plugged in’ to the backplane. The Backplane was a large multilayer printed circuit panel, of varying size, depending on the system, but normally around one metre or 3ft square. Approximately 60-100 PCBs would be plugged in to this panel via a multipin connector, with several hundred pins, on the board.

The multilayer PCB or backplane had been necessary since the early 1970s, due to the incorporation of many more circuits on the board. A single two-dimensional plane could not hold all the interconnections required for this multitude of circuits. The multilayer board incorporates circuits on each layer, with links through the layers, to accomplish the required connections.

For water cooling, the PCB consisted of a large heat sink, around which the circuit components were assembled, through which the chilled water flowed. When the PCB was plugged into the backplane, it also connected via special leak proof connectors to the chilled water supply flowing in the backplane. In turn, the backplane was connected with larger leak proof connectors to the customer supply.

Today’s systems are usually physically smaller for the same computing power, and are designed to have a larger tolerance of air conditioning requirements, some in a so-called ‘office environment’. They are designed with a wind tunnel approach, and require all doors to be closed.