A straightforward solution would be to install several (at least two) PWM controllers into the power supply, but you wouldn’t build a PSU of a reasonable size and cost this way, since you have to put at least two pairs of power transistors and at least two power transformers… Thus, the classic one-PWM-controller design still holds – but with additional regulation of the output voltages!

A linear regulator, one that dissipates the excessive power on the control element (transistor), would be the simplest device for this additional regulation. Alas, but the efficiency factor of linear regulators is very low – the very principle of their operation implies that there must be that surplus power. In other words, the input voltage of such a regulator is always higher than the output one, and the power corresponding to the voltage difference is dissipated as heat. Best samples of modern linear regulators are capable of working at an input-output voltage difference of 0.5 volts. It means that each such regulator in a PSU would dissipate over 20 watts of heats under a high load. These linear regulators would require additional and quite large heatsinks and forced air cooling, and thus they are not acceptable.

It would also be problematic to add a dedicated switching regulator to each of the output rails. Although, placed at the output, it wouldn’t have a high-frequency part and related problems as the main regulator has, but each switching regulator outputs pulsating (rectangular, to be exact) voltage, so a smoothing LC filter has to be put after it. One LC filter already stands at the output of the main regulator, but it’s impossible to put additional regulators before it: they would receive the pulsating voltage from the main regulator. The auxiliary regulator’s own pulsation being superimposed on it, there will be low-frequency, practically non-filterable fluctuations at the output. And if we attach the auxiliary regulator after the main regulator’s LC filter, there will still be rectangular pulsations requiring a second filter. Thus, the use of auxiliary switching regulators would call either for doubling the number of filters at the PSU’s output (and they take up quite an amount of space) or for complicating the already sophisticated circuit with synchronization of the auxiliary regulators with the main one.

The most reasonable way out of this situation is the use of the so-called magnetic regulator or magnetic amplifier, which employs an inductive element as a controlled switch. While the main regulator generates rectangular pulses of a certain frequency, adjusting their length as necessary, the magnetic amplifier cannot generate anything, but can reduce the length of the pulses generated by the main regulator.

The picture below shows you a simplified schematic of a magnetic amplifier taken from the materials of a Unitrode Corporation seminar on the topic:

In this figure Ns is the secondary winding of the main regulator’s power transformer, on which there are rectangular voltage pulses with an amplitude of, say, 10 volts (the oscillogram of the voltage V1 in the figure), a period of 20 microseconds and a duration of 10 microseconds. Without the choke L1 and the voltage source Vc (6v) the voltage coming to the load would be 5 volts.