8

In answering another question here on the S-100 bus (from my admittedly less-than-perfect recall), I wonder why the bus supplied 8 and 16 volt signals, relying on the boards to regulate this down as appropriate.

Surely it couldn't have been too hard to provide one regulator that would supply (the very common) 5V on another pin so that cards could be built much cheaper.

Even though the standard uses all pins, there are multiple GND pins that could have been re-tasked for this (obviously not on a card that expected it to be GND but could they have changed this during the IEEE process, yes?).

3
  • 1
    Changing a ground pin to a power pin in the IEEE process would have risked making existing hardware incompatible. That wasn't really the idea. Mar 23 at 12:39
  • Isn't "powers of 2" the answer here too? ;-) Mar 23 at 21:56
  • @John, that was the reason for my parenthesised comment. You would have had to confirm that the switch from GND to 5V affected no boards out in the wild (at least not any major ones). Either that or make the standard S-100B :-)
    – paxdiablo
    Mar 23 at 22:44

2 Answers 2

15

Remember, we're talking about mid 1970s, that's before switching power supplies became common. The Apple II was an exception at that time. And it was almost half a century before tiny sized, dirt cheap, multi amp step up or down regulators.

I wonder why the bus supplied 8 and 16 volt signals, relying on the boards to regulate this down as appropriate.

Main reasons for redirecting regulation to cards

  1. Without the base system can be lower priced

  2. (relative) large transformer, rectifier and large capacitors are among the cheapest parts here

  3. Components for reliable fine regulations are expensive

  4. Today's cheap components were not available.

  5. Especially for high current the effort was huge - at that time already 3A was considered high current

In a system that offers a high variability of power need between small and large (4 to 22 slots) all these components multiply. A regulated PS for a 22 slot system would be prohibitive expensive.

'Outsourcing' that to each card allowed to use small regulators, just fit for the card they are on, like the new highly integrated 78/79xx series.

Surely it couldn't have been too hard to provide one regulator that would supply (the very common) 5V on another pin so that cards could be built much cheaper.

Yes, it would. Power electronics at that level were simply not available. A 1-1.5A 78/79xx was about the pinnacle of high density regulation in 1975.

Keep in mind S100 was designed to be as simple and cheap as possible. It was never meant to be a professional system. As so often, history decided different (*1).


*1 - A Rant:

I firmly believe that there is a fundamental law of the universe - or at least of all human creation - that always the worst solution that barely does the job will be the one with the greatest success.

Then again, maybe it's just because of our short lifetime and the short lifetime of these solution that they never really get a chance to evolve past that first bad implementation.

Or it's just in our twisted memory. After all, not long after S100 Kontron introduced ECB. Basically still an x80 bus, but using a proper connector, a well defined enclosing shape and off card regulation for multiple voltages. It was quite successful, way past S100, but in the end, all we remember is the cludge MITS did build.

11
  • Someone told me that some users were afraid to extend their RAM to 48K because of potential power problems, and so they kept on using 16K. Mar 23 at 7:02
  • 4
    Efficient switching supplies were expensive exotica in those days, so linear regulation was the method of choice. That makes a lot of heat. If you centralize regulation, you concentrate that heat generation. That makes the regulator very hot and reduces its reliability. It was better to spread the heat generation around the system.
    – John Doty
    Mar 23 at 13:08
  • 4
    It's not as bad as that. It's not the worst solution that usually wins. It's just the first solution that barely does the job... Mar 23 at 14:40
  • 2
    Even 10 years later in de mid-80's the 7805 was still considered a small miracle by many old-school electronics people.
    – Tonny
    Mar 23 at 19:31
  • 1
    @supercat: It's pretty cheap and easy to establish a reference voltage, the expensive bit was the power transistor to drop the main supply. Adding an extra line with 5.7V would just not be worth it when a 7805, a diode and a resistor would do the job on-board.
    – user24174
    Mar 23 at 21:56
4

One issue was voltage drop. Another is fault isolation.

Voltage Drop:

For 7400-series TTL logic (and similar), the Vcc range is 5.00V +/- 5%, or 4.75V..5.25V.

Consider that precision, fast-response power supplies were quite expensive, and these were hobbyist machines. Using a central power supply would mean e.g. setting the central supply to 5.1V +/- some tolerance, then the current draw along both ground and Vcc traces, through connectors, and through the board's power network, plus any voltage spikes from local transients, must not exceed a (transient) voltage drop of 0.35V minus the central supply tolerance. Including load transients occurring on other boards.

Local regulation simplifies the problem to keeping the ground voltage differences down, and the ground plane on a motherboard was large, with multiple ground pins. 'Bad neighbors' with high load transients were not an issue.

Fault Isolation:

The popular 7805 single-IC linear voltage regulators have short-circuit protection built in. This means that if there's a Vcc to ground short:

  1. The fault current is limited to 1..2A; less likely to fry traces, connector pins, and parts.
  2. Find which board the short is on by sniffing / feeling for the hot regulator, or unplugging boards.

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.