Power before all else

THE POWER SYSTEMS IN IBM MAINFRAMES ARE COMPLEX

IBM systems make use of a diverse range of voltages, in part to leverage existing subsystems and designs. Thus, the IBM 1130 and 360 systems mixed SLT circuitry with sections built with the predecessor SMS technology and components built originally for vacuum tube based products. That introduces a laundry list of voltage requirements and the result is a surprising number of different power supplies.

Main frame of 1131

Memory expansion frame (blister)

Our first attention was strictly on the power box that manages the high voltage AC. Already you see that IBM might have a 230V machine but power some components such as fans and the convenience outlets at 115V. In addition to that, the box produces 24VAC which is used to drive power and emergency stop buttons/switches.

The box itself is the 1131, by the way, because an 1130 system is the 1131 plus peripherals - including 1132, 1442, 1055, 1134, 1627, 1133, 2310, 2501, 1231, 2250, 1403, 2420 and others.  The mix of product numbering schemes reflects the leveraging of boxes from other systems. The 1627 plotter was originally developed for the 1620 system. The 1442 reader was originally developed for the older 1440 system. The 2xxx boxes were developed for 360. The 1055 was designed originally for an older communications product family that included the 1052 and 1053 typewriters and even a card reader mechanism that morphed into the 1442. 

One other major AC voltage used in the system is 7.25VAC for many of the light bulbs in the system. However, even then, some bulbs were 12VDC or even 48VDC depending upon the circuit driving them. Not only are we distributing 115, 230, 24, 7.25, 12 and 48 inside the 1131, we may be delivering some or all of those voltages over cables to peripheral boxes. 

The we need to provide power for the logic circuits. SLT logic (30ns family) requires three DC voltages - +3, -3, and +6. SMS logic can require +6, -6, and +12V and others. Solenoids are typically 48V. Pushbutton switches tend to use 12V which is the voltage level of the slowest SLT family, as that gives more margin to handle a bit of contact corrosion and increased resistance. 

Core memories have their own voltage levels and the two different core memory designs used in the 1130, SJ-4 for 3.6 us and SJ-2 for 2.2 us operation, had different requirements. The blister extension frame that holds larger core configurations or any SJ-2 memory has its own sequencing box and can have its own power supply. There can be a -15, a +6 and a +3 depending on memory type, the main power supply type (MPS versus midpack) and the memory capacity.

The Synchronous Communications Adapter uses RS-232B signaling levels thus an additional -12V supply exists in 1130 systems with the SCA feature.

The usage meters have their own 40VAC supply in a tamper resistant box. 

MAIN DC POWER SUPPLY - UNREGULATED

The midpack power supply, distinguished from the earlier Midrange Power Suppy (MPS) set of power supplies in older 1130 systems, produces unregulated DC. It also outputs 24VAC that the SCA gate will convert to +12 and -12VDC. There is a 7.8VAC output but that is fed to two of the DC regulators as a 'bias level'. 

The midpack supply is not completely unregulated, as the transformer that generates the +12 and +48V levels is ferroresonant, thus line voltage swings have little effect on the output voltage. However, the other transformer in the supply that will ultimately produce +3, -3 and +6 is unregulated. 

VOLTAGE REGULATORS FOR LOGIC VOLTAGES

Three massive voltage regulators are used to convert the raw unregulated voltages from the midpack supply to the precisely managed +3, -3 and +6 for the SLT logic. These have an adjustable regulator card and a crowbar card, based on SMS technology as these are derived from the SMS era supplies of machines like the 1401. The crowbar will short the outputs causing the circuit breaker to open if the output voltage of the supply gets above a target level, in order to protect the SLT cards from overvoltage damage. 

The voltage regulators use Germanium power transistors in parallel banks to support the current capacity of the regulators. The +6V supply is rated at 24A while the other two are 20A ratings. One common failure mode of the IBM 108 power transistor is to become an open circuit, thus the other transistors carry more of the load. The supply at low load appears to be working properly, but begins to sag as it approaches its full load rating and can cause cascading transistor failures. 

Therefore it is essential to build a test for the regulators that will draw full load, that is the only way to be assured that the regulator is working properly. It would have been nice to own an electronic load that could be dialed to 20 or 25A of load, but I don't have that luxury. 

I use banks of power resistors in series and parallel to achieve the target resistance that will consume the rated current. For the 3V regulators, which can supply up to 20A, we need a net resistance of 0.15 ohms that can dissipate 60W in total. The 6V regulator requires 0.24 ohms net resistance and must handle 150W of power for short periods of time. 

RELAYS INVOLVED IN SWITCHING SOME OF THE DC VOLTAGES

The power sequence logic verifies that we have good levels on +3, -3 and +6V before it allows the rest of the circuitry in the machine to receive the other voltage levels. Thus, a relay controls +48 and +12VDC to the machine. One symptom of a dropped logic voltage level is that the indicator panel lamps to the left of the keyboard do not  light, even ones like Disk Unlock or Forms Check that one expects. 

Relays also switch the voltages to core memory, so that we don't effect any data with random signals if the SLT logic is not properly powered. These tend to be documented on whatever page had free room when the engineer was tasked with drawing them, often nowhere near the rest of the circuitry that either drives the relay or consumes the switched power it controls. Nothing new here, this is a common deficiency in IBM documentation, that it is technically correct but may be unhelpfully obscure. 

POWER SEQUENCING LOGIC

An SMS card that sits just outside the main power box implements the power sequencing logic for the system. There are a series of three reed relays, each energized by one of the main SLT power levels. The contacts are wired in series to power a relay (R1) that energizes when all SLT voltages are present.

Only when R1 has energized is relay R2 activated, to provide the +12 and +48VDC to the other circuits of the 1131. Relay R1 also has contacts that are part of a safety circuit to lock out the machine from powering up if it had dropped an SLT voltage during operation.

The lockout circuit involves a time delay relay TD1 which will cause a lockout reed relay RR1 on the SMS card to energize if R1 is not on at any time. If one of the voltage levels is off before you first power up the machine, the only consequence is that R1 and R2 don't engage and we have a machine that won't do anything. 

However, if one or more SLT voltages drop after TD1 is energized, it fires RR1. RR1 has contacts that will keep it energized even when the system power switch is turned off. As long as the main circuit breaker of the 1131 is turned on and the plug is inserted, RR1 locks out the power switch from operation. The symptom of this is a machine that won't turn on. A CE Reset switch can turn this off, or flipping the main CB, or unplugging the machine. 

I believe the logic behind this is that one of the voltages dropping during operation might have been due to the overvoltage protection, when the crowbar card forced the regulator CB to trip off. If the regulator is producing dangerously high voltages, we don't want customers switching the machine on repeatedly to try to get it to come up. 

That is why this is effective if R1 had picked due to good levels and later it drops, but they are less concerned if the regulator doesn't come up initially since that is not a symptom of an overvoltage fault. The power sequencing card also ensures that when the customer flips the system power switch off, it first drops +12, +48 and core memory voltages while the regulators are still delivering good SLT voltages. Thus we are protected from random signals from SLT causing issues with core or solenoids in peripherals. 

For machines with the blister frame installed, there is a second SMS sequencing card that is used to drive a relay B in the blister that won't deliver +12V to the core unless SLT power is good. 

EPO SWITCHES AND THE 1133 COMPLICATE THINGS

For safety reasons, IBM systems have an Emergency Power Off pull switch on the console. That will drop power immediately across the system. Unlike the system power switch, which leaves the 115VAC convenience outlets powered up when the system is off, the EPO also drops power to the outlets. 

The 1133 Multiplexer box is an expansion for the 1130 system containing circuitry to attach peripherals such as the 1403 printer, 2310 disk drives and others. Since the 1133 handles peripherals that need 3 phase power, such as the 1403 printer, it is the primary power connection for a system that has an 1133, and the 1133 will feed AC power to the 1131. The building power connection is to an 1133 when it is part of the system otherwise to the 1131. 

Because of its role supplying overall power to the 1130 system, the 1133 has its own EPO pull switch in addition to the one on the 1130 console. These are wired in series. The 1133 also supplies the 24VAC to the power and EPO switches, not using the 24V output of the midpack power supply.