Why do linear anything at all when the SMPS can do it in the first place?
There are a few differences:
1. We use a different architecture, since the most common flyback design doesn't have a wide output range;
2. The iron transformer is fine by itself, I mean we'll probably want to save its weight and cost, but a two stage system isn't that bad, either;
3. The SMPS output filtering is necessary, and therefore the impedance at high frequencies is defined by that.
The consequence for #3 is that, concerning short-circuit current limiting, or open-circuit voltage overshoot, as transients (output-shorted or breaking-load conditions) both are fundamentally defined by the filter, then the control loop, in that order. We can't have perfectly limited current, or perfectly stable voltage -- well, in general we can't, regardless of design method -- but almost certainly not to the same frequency range / transient duration we could with a well made analog supply. That is, the filter only approaches whatever respective CC/CV characteristic the control provides, for frequencies well below the cutoff frequency, and at best approaches a resistance at higher frequencies (but more often, a complex combination of L and C, being the values in the filter itself, and wiring strays and etc.).
This doesn't seem to be much of an issue in practice, as most linear supplies are happy to load their outputs with fuckoff massive capacitors anyway. They surely don't need 470uF+ hanging out there.......but many do it anyway!
And if you're unaware of both facts [transient impedance, output cap] -- then, keep this very well in mind when using any bench supply as a current source. They're mostly voltage sources primarily, with current limiting as a bonus feature. So the cap on the output is a common sight. This makes it rather destructive to, say you're testing LEDs -- obviously you should be able to set voltage for anything higher than Vf, and current equal to the desired If. But what you get is, you connect the LED and the cap discharges into it, blowing it up most likely. If in doubt, always set desired current limit, set voltage to zero, connect the load, then turn up voltage until the current limit hits; or set desired maximum voltage, short the output to get it in current limit at ~0V, connect the load, then remove the short, letting the voltage rise to the load voltage (Vf) in current limit.
Anyway, for design purposes, #1 basically involves adding an aux supply, so that the main controller can keep running -- normal flyback supplies are self-powered, and they need say 10-20V for themselves, so the output likewise can only vary over about the same range. No good for a bench supply, obviously. An aux supply removes that restriction. Also, we prefer changing the type to a forward converter, as primary side current and voltage (times PWM) set the output, making tighter control possible this way. (Conversely though: we might be interested in the natural constant-power characteristic of the flyback, capable of delivering fairly high currents at low (primary side) duty cycle. This is not at all a hard rule, many options are possible, maybe with more work in some cases to get the controls stable and accurate.)
#2 is basically, even if we stick with the iron-core transformer, we can take its output and buck it down to any desired output voltage and current within range; a buck converter is relatively simple to design and build, so this is a surprisingly good option. And the transformer serves in part as a line filter, so we have less noise emissions to worry about. We can improve upon it by using an offline SMPS (fixed output voltage) to replace the transformer and rectifier, and still buck the output to implement our control. It's still simpler than making the full-range, adjustable, offline SMPS ourselves -- safer and less noisy too, as a lot of testing goes into making them run safe and quiet.
Notice I've hardly mentioned noise. Noise is mostly a non sequitur. I've made radios with SMPS in them; I have test equipment with SMPS in them; I've made SMPS with <= 100µV noise output; it's just a matter of good filtering and shielding. 100µV is a pretty reasonable level I think, for not too much effort being required. Below this level, you may need to employ more aggressive measures (metal shields over PCBs, shielded enclosures and cables, etc.), including active measures like linear postreg, capacitor follower, etc.
And most of all: few users actually need low noise. Noise isn't just something you say; well...often, sadly, it is; you see it claimed a lot in marketing blurbs. What does it mean there? Absolutely nothing. Noise is measured. No measurements, no calculations? No mind! 10s of mV of noise is more than enough for average benchtop use; you'll barely see it on the scope at that level, most of the time. Single to fractional mV is enough to behave around radios; such is roughly the level of FCC Part 15, etc. emissions limits. Fractional mV is enough for audio purposes, a bit lower still for very high gain circuits (phono/tape preamp?) or wide dynamic range (professional equipment).
Tim