15

Microcomputers run off low voltage DC, but usually came with a power supply to produce a suitable DC current from the mains AC voltage. It was usual among early home computers to supply an external power supply, either on lead or connected to a plug. Amstrad CPC, PCW and early PC machines had their power supply in the monitor. Most of these machines were passively cooled.

The original IBM PC and subsequent clones mostly had the power supply mounted in the case. Given that microchips and other components can be sensitive to temperature, this seems a poor decision. Such machines often needed a fan for the power supply and even cooling for the CPU later on.

Was it common practice prior to the IBM PC to build power supplies into the case with the board? Was their a good engineering reason for doing so?

15
  • 19
    "It was usual among early home computers to supply an external power supply, either on lead or connected to a plug" ... it was certainly common, but I'm not convinced about usual. The Apple II, Commodore PET and BBC Micro all predated the PC and had internal PSUs (although you could debate whether the PET's was part of the monitor or not, being an all-in-one device).
    – occipita
    Oct 3, 2020 at 22:43
  • 21
    It was usual for all appliances that ran off mains power to have internal power supplies; I can't see any reason why home computers should be different -- "one more box" is not an improvement to user-friendliness. Oct 3, 2020 at 23:21
  • 14
    Until very recently, a PC needs nearly a dozen voltages. Using an external supply would involve too many conductors in the cable. But even if we can supply a PC with a single voltage, it's still difficult to transmit near kilowatt of power over low voltage over more than a meter. Oct 3, 2020 at 23:23
  • 6
    @user3528438 original IBM PC, as well as PC XT and AT, was using +5V, -5V, +12V, and -12V - that's four different voltages. Modern-day ATX uses +3.3V, +5V, +12V, and -12V - that's also four voltages. Earlier versions of ATX were also providing -5V, similar to XT/AT, making that five voltages. Where did you find "nearly a dozen voltages" and how are recent computers using any fewer?
    – moonwalker
    Oct 4, 2020 at 6:43
  • 2
    Don't know about good engineering reasons but external power supplies are a real pain. Very often the cable to the transformer isn't long enough so it pulls the cable off or bends it in an awkward way so that it starts tearing after a while. The double insulation then becomes single insulation.
    – cup
    Oct 4, 2020 at 6:57

7 Answers 7

35

Internal power supplies, while not universal, were not unusual at the time. The major considerations were cost, ease of design, and safety (that is, passing safety regulations that already existed in most first-world nations), with thermal considerations factoring in somewhere further down the list. The latter tended to point towards implementing a switch-mode PSU rather than a linear one; though a linear PSU tended to be cheaper, it was generally far less efficient. Some early computers, such as the Commodore PET, did use linear PSUs, due to the relative ease of designing them using off-the-shelf components.

When the computer and its monitor were sold as a set, it made plenty of sense to tap the computer's power off the transformer that the more power-hungry monitor needed anyway, and integrate only a rectifier and regulator into the computer proper. A variation of this principle also applies to all-in-one computers such as the Amstrad PCW, the original Macintosh and the PET. The computer logic, at the time, only needed a few watts at most, so this was quite a sensible option.

However, that was not an option for computers designed to attach to some random TV that the user already owned. The PC clone industry also demanded that monitors be interchangeable between computers, so a PSU had to be dedicated to the computer rather than shared with the monitor.

Safety certification required that the user should not have physical access to mains-voltage circuits, even when opening the case of the computer to tinker with the logic circuitry (as was then commonly expected). Putting the transformer in an external box, to replace the tap off a monitor's transformer, was one logical solution. The practice survives today with laptops and phone chargers, mostly to get the weight of the mains-voltage circuitry out of the computer itself.

The BBC Micro serves as a good example of how thermal considerations could be solved by using a switch-mode PSU. Early prototypes used a linear PSU, and there were concerns about how warm it got inside the case. Most of the logic didn't mind being a little above room temperature, though one of the custom ULAs proved to be decidedly marginal in the final design. However, design of a switch-mode PSU module, encased in a stout metal shield to deter user meddling, was subcontracted to a third party and proved to be both much more efficient and sufficiently inexpensive to meet the production budget for the machine. The production BBC Micro scarcely gets warm inside.

PSU modules of this type, designed by companies specialising in PSU design and safety certification, became standard practice and were commonly used in PCs and other microcomputers from the 1980s onwards.

6
  • 3
    Thanks for broadening the issues beyond heat dissipation. Oct 4, 2020 at 10:28
  • A common pattern at the time, exemplified by the PC and Apple, among other machines, and still common in today's desktop computers, was to have a "brick" inside the computer's case which served as the power supply. Having all mains-power connections inside a UL-approved assembly greatly simplifies regulatory approval compared with having power supply circuitry on the main board.
    – supercat
    Oct 4, 2020 at 17:08
  • @supercat Isn't that precisely what I said?
    – Chromatix
    Oct 4, 2020 at 18:50
  • 1
    I think the Apple II exemplified that approach earlier than other machines, though I don't know when the BBC micro adopted it.
    – supercat
    Oct 4, 2020 at 18:56
  • 1
    @dlatikay: That would be another good point. While I don't think it would have been a factor with the original PC (whose total supply ratings totaled I think 65W), regulatory exemption requires not only that devices not use high voltage, but that they be intended for operation on current-limited supplies. If memory serves, the allowable current limits varied with voltage, but were limited to something under 150 watts.
    – supercat
    Oct 5, 2020 at 16:28
11

Aside from Cromatix great answer, the thing to consider was difference in power. Typical home microcomputers (like ZX Spectrum or Commodore 64 etc) used power supplies with about 10-30 watts, while IBM PC XT has chunky 130W power supply (which increased to 192W SMPS in IBM AT). That much power requires much more space, and would be quite unwieldy as separate PS (especially when you add it in isolated casing).

Also, when internal power supply has fan blowing outside, it actually helps with cooling of the case with main CPU, motherboards etc. as it draws air over the components, due to air pressure difference.

4
  • But why did an 8086 PC require an order of magnitude more power than a Z80 or 6502? Oct 4, 2020 at 16:42
  • 9
    @MarkWilliams A key difference between the PC-XT (or PC-AT) and most contemporary home computers was the expansion slots. It meant that the power consumption of a fully populated PC was less predictable than that of, say, a ZX Spectrum, a BBC Micro, or a C64 which have more constrained expansion opportunities.
    – Chromatix
    Oct 4, 2020 at 18:53
  • 6
    @MarkWilliams The rated output of the BBC Micro's PSU is about 35W. Most of that is reserved for an auxiliary power port on the underside, which can be used for a single floppy drive without requiring the latter have its own PSU. A double 5.25" floppy drive still needed its own PSU. Now consider what hardware you could load into a PC, all driven by the original PSU…
    – Chromatix
    Oct 4, 2020 at 19:02
  • 4
    @MarkWilliams To add a little more context: The IBM PC had a supply of "just" 65W, and it turned out to be incapable of driving an internal hard drive together with the mainboard, so the upgrade path was a second box of the same size with a 130W supply and ISA bus repeater cards. The hard drive and its controller were built into the second case. The XT tried to do away with the expansion unit by offerering 8 instead of 5 slots and 130W of power inside the main case. A common inofficial upgrade of IBM PCs was to mount a 130W supply and a hard drive (+ controller) in the main case. Oct 5, 2020 at 6:56
6

The IBM PC wasn't a "home computer" -- it was intended to be a business computer, and needed to suit the needs of a business owner. Having the PS in the case eliminated extra connections (for the user) and made everything fit neatly on a desk.

(And it should be pointed out that the PC was not IBM's first "personal computer". The IBM 5100 was also a single-unit system -- even including the keyboard and display in the package.)

2
  • I originally assumed that this was the main reason - it was what business users expected, - rather than their being a real engineering reason for it. This would explain why the PET had internal power, but Atari machines had external power supplies. However, some of the other answers have made me reconsider my view. Oct 5, 2020 at 15:05
  • @MarkWilliams: Atari ST (m68k) and Mega ST / STe all had internal power supplies. I repaired several of those power supplies for local Atari club members by soldering in a replacement for the main switching transistor, so I definitely remember where the PSU was inside the main case. Oct 6, 2020 at 0:09
5

The earliest personal computers of the 1970's, including the Altair 8800, had built in power supplies. The power supplies were often oversized to allow expansion boards to be added to the system.

As the computers became smaller and more limited in expansion options (dare I say toy-like?) it became possible to power them with an external supply. It was probably easier to get electrical certification if there was no high voltage in the chassis.

The IBM PC, meant to be a serious computer for offices with generous expansion capabilities, was more suited to the built-in power supply model.

1
  • I hadn't considered that expandable machines would require more power than non-expandable (home) ones, which would explain a lot. Oct 5, 2020 at 15:06
4

Good points in earlier answers, but one is overlooked. The Ibm PC needs a set of different voltages, like +5V for the main board, +12V for disk drives, -5V for the RS-232 serial interface, -12V in case you have a special board that needs it. It would be quite impractical to have an external power supply connected with a cable to the PC with that many conductors.

In fact, modern PC's need most power at extremely low voltages, like 3.3 V, 1.5 V, perhaps even 0.8 V for the CPU. At a power level of over one hundred watt, this needs a LOT of amps (I should say: current). Completely nonsense to supply this through wires from an external power supply. These voltages do not even come directly from the PSU, but from dc-dc converters on the main board, just 1 or 2 cm remote from the cpu chip. Just look at those 4 to 8 funny cm-size inductors and transistors right next to the cpu. These cpu power voltages are not standardized, but differ for each cpu type. The central PSU unit of the PC does not even know the cpu type, which is another reason why the cpu voltages are converted and regulated locally at the main board.

That said, you can buy pc cases with a PSU that accepts 12 V DC and generates all required ATX voltages. That gives you the flexibility to hook it up to any external 12 V power supply, e.g. from the 110V AC or 230 V AC wall socket, or from a car battery, UPS battery, solar panel etc.

In all, it was not a poor decision to build in the power supply. As if IBM did not have smart engineers. The build-in power supply generated the exact voltages as required by the main board and peripherals, at the required power levels, to guarantee reliable operation.

As for the fan, most heat is generated by the CPU. The efficiency of the pwer supply is over 80 percent, whether linear or switched mode. That means that 80 percent of heat is generated by the main board and peripherals. Using an external power supply would not make a fan superfluous. Most laptops have external power supplies, and many have also an internal fan.

3
  • 2
    Decent answer; thanks for contributing. It is possible to generate multiple voltages from one DC supply. The ZX Spectrum converted 9 V DC to 5 V, 12 V and -5 V DC internally, admittedly all at quite low currents. RS-232 voltages are typically generated by charge pump ICs such as the MAX232 from a 5 V supply, too.
    – scruss
    Oct 7, 2020 at 12:46
  • @scruss MAX232 is an answer to the pain of needing two voltages, sure, but back in 1980 we needed negative voltage from the PSU. Later, we used versions of RS232 with one polarity, with lower voltage, and now it is replaced completely by USB. The main point is that the PC, even today, needs multiple voltages from the PSU. Just look at the ATX connector. And these voltages need to switch on in a coordinated way, you cannot just hook up each voltage separately with e.g. banana plugs, the PC would not boot correctly. Hence internal PSU
    – Roland
    Oct 7, 2020 at 20:32
  • Yes, I know about the multiple voltages. That's why I gave an example of a computer that generated negative voltages from one power plug. It wasn't anything like at the current that PCs need, but it can be done with one cable. Old CPUs (particularly the 8080) needed negative bias voltages for the substrate, and that had to be carefully timed to avoid frying the CPU
    – scruss
    Oct 8, 2020 at 0:13
4

Another issue, outside the heat and multiple voltages issue, is the safety issue of low-voltage DC cabling. Home computers didn't draw much power, so a 5 – 12 V wall wart on a fairly short, thin two-conductor cable was considered safe enough. For example, the ZX Spectrum used a 9 V 0.7 – 1.4 A supply feeding a simple 7805 regulator, consuming ~650 mA at 5 V (so 3.25 W or so), dumping the rest as heat.

Once you draw more power (say the 65 W of the original IBM PC, which is not that far away from a modern laptop), DC cabling starts to cause problems:

  • you either have to use a much heavier and more expensive cable, or you accept the losses and allow a cheaper cable to get warm in regular use. I²R losses get very expensive when you're specifying PSUs to be sold by the millions.

  • Unlike AC, DC arcs don't self-quench. While the arc from a broken connector for ≤ 12 V is tiny, it's still hot enough to create a fire risk.

Laptops now tend to use a higher DC voltage to reduce the copper losses in the cable, and the cable is thicker and better strain-relieved than the old home computer wall-warts. But the strain relief only came to some manufacturers after reports of laptop adapters on fire.

1

They still have internal power supplies. When you see an external supply, it's usually an AC-DC converter taking mains AC and converting it to an intermediate voltage, 5-20V. The computer contains DC-DC converters to make the voltages it needs from that.

8
  • 3
    I'm not sure this answers the question, nor which computers this applies to. Oct 5, 2020 at 19:54
  • @MarkWilliams I am surrounded by five computers. The Mac uses an AC -DC converter to get my 120V AC mains power down to 20V DC. I don't know Mac internals, but there can't be much in a Mac that runs on 20V. Then there are two Raspberry Pi's running on 5V power bricks, but they each have a four output DC-DC converter (MxL7704 chip) to make their internal voltages. Another Raspberry Pi is in a case with an internal AC-DC converter. And my under-desk tower has an internal AC-DC converter, too. Can you name a computer that gets its internal low voltage DC from an external supply?
    – John Doty
    Oct 5, 2020 at 20:09
  • I believe the Amstrad CPC machines run on 5v and ground, with 12v for the disc drive. Both are supplied from a transformer in the monitor. Oct 5, 2020 at 21:51
  • 1
    @MarkWilliams Perhaps the question is what is a power supply? Is the AC/DC converter THE power supply? Is the DC/DC converter next to the cpu socket not a power supply? Is the power adaptor for a laptop, 230 VAC to 19VDC, THE power supply? And is the module inside the laptop that creates 12V, 5V etc not also a power supply? Does a TV set have an internal power supply? How about a 230V AC led lamp? How about that small converter (joule thief) in my tiny GP led flashlight on one AAA battery? Are you only interested in old PC's? Do you worry about bad engineering at IBM?
    – Roland
    Oct 8, 2020 at 13:04
  • 1
    @Roland, that is indeed a good question. I am using the term to refer to the AC-DC conversion, but I am mindful that voltage rectifiers etc are also part of the power supply chain and are often thermally inefficient too. I ask the question in the context of old PCs, and the contrast between home computers and business computers. Was the AC/DC conversion in the case because that was obviously best (at the time), or because that was normal and convenient for a business machine (e.g. a mini computer or a terminal) and would be expected by a customer? Oct 8, 2020 at 15:56

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .