In the log Capacitively Coupled Pre-Biased Relay Logic, I have explained the discovery that capacitive coupling with 100µF can effectively replace the 22 Ohms liaison resistor. The working voltage range is increased. This also reduces the power draw because a current path is removed, the average required power is lower.

Why do I bring thing again ?

Well, I tried to drive several coils from one relay. With resistive coupling, things didn't work well.

The 22 Ohms resistor is great for driving one relay.

For two relays, the current paths become more complex but it still works somehow. With 3 coils loads, things are not good. Some relays barely switch. This is because the voltage swing is not symmetrical anymore, as seen on the 'scope.

The CCPBRL can solve this. But this creates a new type of problems...

My test circuit worked but it is considered as a "fluke" because electrolytic capacitors don't like being backwards-biased.

Electrolytic capacitors are chosen for their cost and capacitance density. Non-polarised capacitors are larger or too expensive. So there is a need to bias it properly.

The "solution" is to use more than one supply voltage rail. For example, +3.3V and -3.3V.

For a single output, this works as expected : the signal drives the slave coil(s) through a capacitor that is connected to the opposite rail.

For 3 coils, there is a new possibility : tie the 3 coils in series and use the higher voltage (between -3.3V and +3.3V).

It's pretty easy but the power supply becomes a different story.

The funny part is that I encountered this situation while trying to create additional voltages (-3V and +6V) using a charge pump. The thing more or less works but is not worth continuing.

Higher voltages are required at least for the RAM system. A higher charge should provide a stronger "kick"...

Trying to "stay true" to antique design methods and using technology that my grandfather has used, the natural approach is to use a dual-windings transformer (torus if possible) followed by a couple of diode bridges and a crazy amount of capacitance (I'm covered on that part).

I'm not covered on the transformer part though. Most voltages I find are at least 9V but I need 2×3V.

At 3.3V working voltage, add the diode bridge drop : 4.7V. Then the transformer output should be around 4.7/1.4=3.35Vac. The lowest secondary I can find is +6V/-6V. Maybe I'll need an autotransformer in front (a Europe-USA converter, 240-120V), but I'll loose efficiency and room...

So for now I'll simply use a couple of cheap switching PSU.