Have you seen this being done in actual equipment, where there is provided a controlled amount of current at a controlled voltage to the otherwise isolated side?
Sure. A simple way is an isolated DC-DC converter driven by a current- and voltage-limiting supply. You can find such even in USB high-speed isolators for the downstream port, limiting the current to 100mA until USB connection has been formed and the actual current limit desired discovered from the USB descriptor; then usually limited to 500mA.
For sensors and medical equipment, you often have a filter and a linear voltage regulator followed by a secondary current limiter (based on a dedicated chip; or an instrumentation amplifier, shunt resistor to measure the current, and a high-side P-MOSFET or low-side N-MOSFET to cut the power altogether) so that if the current draw is exceeded, instead of just voltage sagging the power is cut completely.
(That said, I'm only somewhat familiar with scientific sensors and such, and never worked on real medical equipment myself.)
I think most with such isolator may still accept the entire power supply voltage and current at the isolated input.
Most isolators do not provide any current to the isolated side, and only deal with the signal. This includes USB isolators like ADuM3160, ADuM3166, ISOUSB211, and others; and isolation amplifiers like ISO122.
Voltage and current is typically provided by an isolated DC-DC converter, which uses a high-frequency transformer to pass current over the isolation gap. Some input power (from a couple of percent to twenty percent, depending on the transformer; up to half for low-voltage ones like 3.3V and 5V ones) is lost as heat, but the rest is coupled to the other circuit. The number of turns in the transformer dictates how the voltage over the primary side reflects on the voltage induced on the secondary side; and since the output power matches input power minus losses, the current similarly. That is, by controlling the current and voltage on the primary side with a transformer having known number of turns (ratio), you can tell approximately how much current and voltage is induced in the secondary side. It is easy to regulate to a fixed voltage with a linear regulator, and the current drawn by the rest of the circuit can be measured and connection stopped if it exceeds a set limit.
Also it is not hard to create a module which can work with all equipment? The module would simply provide isolation to the 2 differential inputs and one ground/common,reference? This means the voltage coming in would be the same voltage in the output. No amplification, only sure isolation with that very tiny amount (less than 10mA) of current controlled just to make the isolator operate.
It is not that simple, essentially because there is no transformer suitable for DC, nor an universal one-fits-all transformer for AC. The geometry depends on the primary side current and voltage, as well as the turns ratio needed. Plus, if the isolated side is connected to the main ground, it is no longer isolated: current can flow through to ground, making the isolation useless.
This is the difference between class I and class II isolation: class I has a safety capacitor between the grounds, reducing EMI generated during switching, and class II has the powered side completely isolated. In both cases, the potential difference between the zero-volt rail on either side can be hundreds of volts without any extra current passing, but in class I the capacitor means that some leakage is possible, and can lead to "tingling" sensation.
Medical equipment uses MOPP (Means of Patient Protection) AKA IEC60601-1 standard compliant, class II isolated power supplies. They're not that expensive; Mean Well ones seem to be common. For example,
Mean Well RPS-30: 30 watt class II isolated AC/DC supply, with several submodels providing one 3.3V, 5.0V, 7.5V, 12V, 15V, 24V, or 48V output). Basically, if you short the output to ground, only 5-10mA will flow over the short. This is not enough to stop the human heart. (In pulses it could disrupt the rhythm of the heart causing arrythmia, though.)
Of course, if you put a human within the circuit, for example connect that power supply 0V to a needle in one hand, and the +V to a needle in the other hand, the current will pass through the human and possibly kill them. We are fragile bags of conductive salty water, mostly, with a rubber-like somewhat resistive outer covering; and easy to kill.
Is there already a commercial unit that does this?
No magic bullet exists, no.
But the components needed to make your supply and your equipment thus safe are extremely common and cheap.
For example, you could start with an extremely common USB wall wart. (Of course, for medical equipment you get something better and safer, especially with careful EMI checks so that it does not generate interference to other devices nearby.)
That gives you about 5V and ample power. You then add current limiting that cuts the circuit for a set duration (say, one second), if the current exceeds some set limit. They exist for USB (at 100mA and 500mA selectable limits), but you can construct one yourself using an instrumentation amplifier or analog current sense IC, a shunt resistor (under 1Ω –– the voltage drop over the resistor corresponds to the current over the resistor, U = I R, and the amplifier boosts that by a fixed factor), a comparator, a MOSFET, and some capacitors and resistors. Not too complicated.
You then add an isolated DC-DC converter. Because the voltage and current is limited on the primary side, the isolated secondary side is already limited to how much power it has available. I would filter the output (using a Pi filter or capacitive multiplier) followed by a linear regulator (dropping a volt or two) to get a really nice, stable voltage, necessary for many sensor applications, and might even add a secondary current cut-off limiter (the same thing as on the primary side) for safety.
Of these, the isolated DC-DC converter is the most "expensive" component; at Mouser in singles, they cost between 5€ and 15€ apiece. In the above cost-optimized scheme, I might use RECOM power RM-0505S, whose 5V output is unregulated (which is fine, because I'd filter and regulate the output to 3.3V anyway, assuming 3.3V needed for the medical electronics). It itself is limited to 50mA on the isolated side, which is ample for powering signal isolators, but it has no minimum load, so it should work for the targeted 10mA max on the isolated side.
The main cost would be the AC-DC and isolated DC-DC converters, with the rest of the components being 'jellybeans' as Dave calls them.
In commercial medical use, the cost of safety and standards compliance testing would probably be much higher!
What is the schematic for this circuit say using the ISO122 (capacitiv isolator)
TI ISO122 is a precision isolation amplifier. On both sides of the isolation barrier, you have a positive and a negative supply, and a ground reference. On the isolated side, an analog input voltage (referenced to its own ground) causes a corresponding output voltage (referenced to its own ground), within the valid range; the input resistance is about 200kΩ, so very little current is drawn from the input pin. On both sides, it may draw up to 7mA of current from the supply on that side.
You could use the power supply circuit I outlined above to power this (except with a dual-output isolated DC-DC converter, to get both rails; or with a circuit to generate the negative rail). Then, all the microcontroller stuff would be on the side between the DC-DC converter and AC-DC supply, and use as much power as it needs (as long as the AC-DC supply can provide it). Assuming you set the current cutoff on the isolated side to say 8-9mA, and on the non-isolated side (before the DC-DC converter) to say 17mA, you could prove that even if the ISO122 part was directly grounded, less than 10mA would flow to/from ground.
To pass medical safety tests, you'd of course have to do more design work, and consider and test for all kinds of fault conditions, which is what tends to make medical-grade electronics so much more expensive. Without the testing and verification, their designs and implementation aren't
that much more expensive, really.