The (brief?) era of single-voltage computing

My first computer ran with a single 5V supply to the chips - it was a 6502-based Compukit, an unlicensed copy of Ohio Scientific’s Superboard. It had 4K or 8K of SRAM, in the form of 512x4 bit chips. This was the late 70s.

It struck me recently that the single-supply computer was a relatively limited phenomenon. Earlier than my machine, it was common to need a 12v supply or a negative rail, or even both. (Even earlier, the valve/tube based computers had quite a spectrum of supply voltages.) Later than my machine, there would commonly be a negative rail for audio purposes, or a 12v, or both. Even later, we start to see 3V or even lower voltages, often with 5V as well.

So: which early or late machines needed and used only a single supply rail? (If you like, discount any higher voltages needed for a built-in display.)

Here’s a Compukit - you’ll notice it was sold as a single board, you had to make your own case and PSU:

4 Likes

What an interesting observation! I hadn’t thought about it, but it is certainly true. Even when the logic was all single voltage, disks and I/O often required multiple voltages.

I think many of the true single voltage computers of the mentioned era were effectively embedded designs (certainly in the MOS, Intel, and Zilog spaces, the 6502, 8085, and Z80 were designed for that market). You can still find this today in embedded designs; it is pretty easy to build a universally 3.3 V embedded ARM device with significant computing power. Interfacing with the outside world still, as it did then, typically requires additional voltages.

1 Like

The first CPUs needed multiple voltages, but even after 5V-only CPUs and logic started to appear there was still a need for plus and minus 12V - for RS-232, at least. I guess that’s one reason my Nascom-1 needs both plus and minus 12V. Though it also needs -5V (for e.g. the MCM6576P character generator) in addition to +5V.
So, if you wanted proper RS-232 it’s easier to ask for + and - 12V from the PS instead of synthesizing the voltages (easy today. Though they tend to cheat and make only ± 5V, which may or may not work with RS-232 devices).

Good point about RS-232. My initial thoughts were mainly about RAMs (and ROMs) needing multiple rails. As did the 8080.

Even the ZXSpectrum’s 4116 DRAMs needed three rails.

The Apple 1 used the single-rail 6502, but the advert mentions ‘four fully regulated power supplies’ - again, the DRAMs need a 12V supply as well as 5V, and some 24 chips need a -5V rail.

2 Likes

I agree a lot of systems are “signed” by the usage of 4116 ram chips. Luckily several systems can be converted to a single 5V rail with the replacement of the DRAMs using a modern SRAM (and some additional logic)

Ed,

The Motorola 6800 of 1974 boasted that it only required a single 5V supply.

However, they employed an on-chip voltage doubler circuit to provide the higher voltage rail.

Provided that the 6800 was using static RAMs (the 1K x1 Intel 2102 appeared in 1972) it would give a single rail system.

However the EPROM was still the problem - only the Intel 2716 introduced in 1977 used a single rail supply.

This is why early machines used bipolar fusible link PROMS for their non-volatile storage. Monolithic Memories was a principle manufacturer making devices up to 1K x 8, See page 22

http://www.bitsavers.org/components/mmi/_dataBooks/1980_MMI_Bipolar_LSI_Databook_2ed.pdf

The larger devices were in industry standard packages with pinouts that made them compatible with the 2708 EPROM. They were also a lot faster than the contemporary EPROMS with sub-100nS access times.

Elsewhere, Intel introduced the single 5V rail 8085 cpu in 1976, in order to simplify the power supply, and supprt chip requirements of the 8080. They were also under a lot of competition from the newly introduced Z80.

2 Likes