What does single supply really mean, when referring to op-amps?

Half way right. Unfortunately, the terms have some historical baggage. No one would name today's gasoline formulation "lead free" unless older formulations had contained lead.

First there were dual supply opamps that were not called that, because all opamps were about the same, with input common mode range and output swing not getting very close to either supply rail. Usually these were operated with dual supplies, but would work with single supplies if the signals were not close to ground or the positive rail. I remember when analog systems almost always had +_12 or +-15 volt rails.

Then TTL logic came along, and the +5 volt supplied system became the most common one. Some people wanted to add a few analog features to TTL systems, and it added a lot of cost to add +- 15 volts to a 5 volt system, just to add a couple comparators or opamps, so the low voltage capable, ground rail inclusive designs were invented (the LM324 series of opamps and the LM339 series of comparators, especially). These let you add analog features to a 5 volt system i.e. the inputs and output stayed between +3.5 volts and ground. The advertising code words to distinguish these new variants of opamps and comparators was "single supply", since they could be shoe-horned into 5 volt systems.

There were no rail-to rail designs for years, until CMOS processes matured for quite a few years and were refined to be used for analog chips. The wasted 1.5 volts of the supply range was not a serious problem for many designs, till the supply started dropping from 5 volts to 3.3 and below. Even though "single supply" designs were improved to function better down to 3.3 volts, the remaining 1.8 volt or so swing was getting to be a real problem, that was pushed back by coming up with parts that really worked with input and output voltages over the full supply voltage, based on low voltage CMOS processes. The advertising code words that have been used to distinguish these new features from the older two types was "rail-to-rail" inputs and outputs. There are also some versions that have only "rail-to-rail" inputs or outputs but not both. This is a gotcha you have to watch out for, when you pick an opamp that is advertised as just plain "rail-to rail".

As far as I can remember, the original dual supply opamps only started being labeled as such only after "rail-to-rail' designs became available, and there was a need to distinguish them from the newer designs by their power supply head-room requirements.

Not a bit upset.

As far as thinking things over, a 3-supply-opamp could make sense. All

John

As in my 1964 design...

Without a high impedance node, or a PNP mirror, rather good CMRR and PSRR performance.

Still sold today...

...Jim Thompson

Thank you. Now, that is a real dual supply opamp!

Good, but not the whole story. There are single-supply op amps (LM301 and LM307 come to mind) that tolerate near-(V+) but not near-(V-).

It is difficult to get common-mode rejection for near-power-rail signals with most linear components, so when manufacturers got their op amps to work well when the inputs were near a power supply rail, they advertised the achievement. That part is called 'wide CMR range', or 'common mode range includes negative supply', or somesuch.

Later they got the output impedances to stay low when the outputs also strayed near one power rail, and they called this 'single supply operation' because the inputs and output could get close to one power supply. It became possible to use the op amp for simple operations (like a comparator) with AC-coupled signals and a resistor to (GND aka V-) biasing an input. Nowadays, some op amps can tolerate inputs near either power supply, and outputs swing close to either power supply, and those are 'rail-to-rail' op amps.

Few folk really like using single-supply amplifiers, though; your inputs can be AC (above/below the rail) but the outputs can only be of one polarity. You can AC-couple the output, of course, but the situation isn't ideal. The ideal operational amplifier allows wide-ranging inputs and the output has low impedance and can maintain a range that is bipolar (range above and below the GND). To get that, conveniently, you still use two regulated power supplies.

It could be worse. Vacuum tubes used three (filament, bias, and plate) and early transistor amplifiers (does anyone else remember NIM?) benefited from up to five regulated supplies.