The original IBM PC supplied several voltages: +5 V, −5 V, +12 V, and −12 V. Which computer supplied the most DC voltages?

To facilitate the comparison, let's agree on some uniform criteria for what counts:

  • Only DC supply voltages count. AC line voltages (typically 120 or 240 V) do not count (they are not DC). Analog signals do not count (they are not a supply voltage).
  • Do not count ground / zero volts. Thus, a "single-supply" computer counts as one supply voltage.
  • Negative voltages count separately from positive voltages. Thus, −5 V counts separately from +5 V.
  • Two or more supplies with the same nominal voltage only count once, even if they come from different circuits. So a regulated +12 V and and unregulated +12 V count only once.
  • Voltages created on-chip (e.g. by a charge pump) count, if you can specify their amount.

By these rules, the original IBM PC had four supply voltages.

I suspect that a vacuum-tube computer might be the winner.

  • "So a regulated +12 V and and unregulated +12 V count only once."- so when you say 'greatest number' you mean greatest number of different voltage designations, not their sources? Mar 7 at 6:13
  • 1
    "Voltages created on-chip (e.g. by a charge pump) count, if you can specify their amount." - what about voltages created on board but not inside a chip? What do you mean by 'created'? Mar 7 at 6:18
  • One more question - do all the voltages have to be supplied continuously at the same time? Mar 7 at 6:20
  • 1
    In the PC world, the ATX standard added the 3.3V to the power supply so that's five. But the motherboard could interally use virtually any amount of different voltages for components and buses. At least there would be converters for supplying the CPU and memory with correct voltages.
    – Justme
    Mar 7 at 9:11
  • 1
    " As do the higher voltages needed by CRTs and vacuum tubes"- Now it's getting interesting. But as this is Retrocomputing, you need to set a cut off date. Mar 7 at 17:37

The IBM 704 and 705 -- both vacuum-tube computers -- used twelve distinct power supply voltages:

IBM 704/705 power supply voltages

The 704 was a scientific computer, and used the unique voltages +220, +150, +15, +10, -15, -30, -100, -130, -150, -160, -200, and -250 V. The 705 was a business machine which used +270, +140, +72, +48, +40, -12, -60, -130, -140, -150, -180, and -270 V. Of course, IBM had a color code for each of the voltages (described in the source). These are not signals, which are described in the following section of the source. In addition, the filaments ran on 6.3 or 12.6 V, but this is AC, which does not count.

Source: IBM 700 Series Component Circuits, PDF page 379, logical page E27.

  • How do you know earlier computers (like the ENIAC-class ones) didn't use even more voltages?
    – dirkt
    Mar 9 at 6:01
  • @dirkt: If you can find a better example, please write an answer.
    – DrSheldon
    Mar 9 at 6:18
  • The point was that the question "which computer used the greatest number" is not answerable. We can give examples of computers with many voltages, and more examples, but we'll never know if we found "the" one.
    – dirkt
    Mar 9 at 6:47
  • 3
    @dirkt So what? Most questions are not answerable in the absolute, but a relative answer can nonetheless be an interesting endeavour. The world is not black or white. As the wise man said, sometime it's the path that is goal. Mar 9 at 12:09

I'm working on a relay machine (not otherwise written about in much detail yet - it's still a WIP), but it doesn't even use DC voltages - everything is AC, and there are multiple voltages. Everything is in VAC RMS:

  1. 3.0V
  2. 5.0V
  3. 6.0V
  4. 10V
  5. 12V
  6. 12V inverse polarity (akin to "-12V")
  7. 18V
  8. 24V
  9. 48V
  10. 62V
  11. 96V
  12. 140V
  13. 170V

So, what sort of magic is all that? Not very magical, I'm afraid. There are simple reasons for this multitude:

  1. Since the design uses thousands of relays, and they all had to be purchased secondhand, I didn't have much control over the coil voltages. I had to buy what was available at the best price and fulfilled the speed and load/voltage capacity requirements. That meant <$0.25 per relay for the vast majority of them. I got 2.0V and 2.5V coil relays (really), and also 3V, 5V, 8V, 12V, 18V and 24V.

    I'm not even sure I got 24V ones since they had an obscure custom type number with no data available and no indication of coil voltage, but they work well at 24V, don't get too warm, and are quite fast - actually the fastest reed relays I got, with turn-on time when new of about 70us, and the "worn in" turn-on under 120us. They of course take about 1000us to stop bouncing on turn-off, so required some care in applying them.

  2. All of the relays I use, save maybe a dozen for high-current I/O switching, are physically small. Some of them are 5x5x9mm in size. The tiny contacts thus have relatively low current ratings. There are many circuits that are limited to 40mA max currents - much below the contact maximums, but larger loads take excessive cut out of the contact life. If there's a control signal that needs to drive 8 loads (coils) at 2.5V each, and the loads are all on the same card, it's preferred to have a 12V signal driving these loads in series. This will reduce the load current by 8-fold, as well as reduce the voltage drops and thus heat dissipation in the connectors and traces on the plug-in boards and in the backplane. Driving eight 12V coils in series takes a 96V signal, of course driven by a "bus transmitter" with a contact with sufficient voltage rating (including some derating).

  3. The gas discharge tube (GDT) switches I found for "pennies on the dollar" need 140V to reliably ignite when cold, dark and with their aging/wear taken into account. The design is semiconductor-free, so the only alternative to diode matrices is GDTs. GDTs also can be coaxed to work in sequential circuits like shift registers that wear out much slower than the very fast reed relays needed otherwise would (e.g. in a UART application). I initially used neon bulbs instead of GDTs, but they had to be paralleled to support the currents I needed (10-40mA) while surviving longer than minutes, and would end up costing more than GDTs and of course taking way more space. That was specific to the particular "bargain basement" components I had access to, and isn't any sort of generic advice.

  4. Neon bulb displays can be simplified when voltages are available to drive multiple neons in series. 170V is only for this use, but lower voltages are used as well where they fit.

For development work I'm using a multi-channel D-class power amplifier to generate the AC voltages, since it has good current limiting - essential when desperately trying to avoid smoking a few dozen relays every time I make a mistake :) I started with transformers but fusing them for development work was too unwieldy. Eventually the power will come from taps on a couple of custom transformers - the ultimate partitioning will depend on the relative load power of various supply voltages.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.