This is something I've thought about quite a bit, that seemed to make too much sense to ignore: Having extra-low voltage DC circuits within a building, produced by a few beefy power supply units in suitable locations.

Motivation
Apart from heating, air conditioning and refrigeration, the lion's share of other devices in a modern household have relatively low power requirements. And most of those, being electronic in nature, also natively work at low DC voltages, therefore requiring a power supply unit to convert the mains to whatever the device itself needs.

What this idea basically amounts to, then, is to move the mains-to-DC conversion from individual small PSUs at the point-of-use, to a few big PSUs that several devices can share the output of. So, what advantages does this confer? Let's find out...

Why small PSUs can be a nuisance
To convert the mains (240VAC or whatever) to a suitable voltage for the device, with safety isolation, there are two basic methods:

In decades past (and still to this day if you want a low AC output), you used linear power supplies (a mains-frequency step-down transformer, followed by a rectifier to convert the lower AC voltage to pulsating DC, a capacitor to smooth those pulses into a continuous output, and maybe a regulator if a precise, constant voltage is needed). While reliable, they tend to be relatively inefficient, heavy, and bulky, rendering them often unattractive.

The more modern approach is to use switching power supplies. In grossly simplified terms (as they are quite complex), a basic SMPS consists of a mains-voltage rectifier and smoothing capacitor, then the resulting ~340VDC is "chopped" into pulses at high frequency (typically on the order of 100kHz) which are fed into a much smaller, ferrite-cored transformer. The secondary of this transformer is fed through a high-speed rectifier (usually a Schottky type, or even MOSFETs switched on and off in time for lower loss), into a smaller output capacitor (which should be of a low ESR type). Regulation of the output voltage is achieved by adjusting the duty cycle of the "chopped" pulses.

So, let's look at some of the problems with such small switching supplies (many of them explained in detail in this article, questioning the wisdom of the Australian ban on "inefficient" DC-output supplies):
  • As they are often built down to a price, their performance and reliability will not be the best possible. It is very common, for example, that when low-quality electrolytic capacitors are used, they will dry out prematurely and cut the life of an SMPS short - and this is often exacerbated by using capacitors with an insufficient ripple current rating for the job, resulting in internal heating. Surge protection, if present at all, will be limited to 1 or 2 rather small metal-oxide varistors. Most of the smaller units don't have PFC circuitry, either, resulting in harmonic distortion of the input current (which can't be removed easily once present). Even over-voltage protection may be wishful thinking. Size constraints don't help with any of this, either.
  • The high switching frequencies present necessitate additional capacitors and inductors to reduce RFI to acceptable levels. While the inductors are quite innocuous, the same cannot be said for the "class Y" capacitors present between the primary and secondary circuits of most "off-line" SMPS; while they are rated to (hopefully) not pose a danger to humans, they can still hold enough charge to threaten signal inputs and outputs of devices, when the circumstances allow them to (quite readily, given that these supplies are generally of "double-insulated" construction, with no protective earth connection).
  • Internally, considerable "creepage" spacing is mandated by safety standards, between the primary and secondary circuits of such supplies (usually >6mm for a supply used on nominal 230V mains) - but it is still not inconceivable that some failures could result in a breach, particularly if an electrolytic capacitor blows its top, spraying electrolyte and/or scattering shredded foil and paper throughout the unit. Of course, many of the shady no-name Chinese supplies don't even bother to comply with the standards, and I don't need to tell you that those are death traps waiting to go off... Also, in keeping the size down, creepage spacing within the primary circuits of small SMPS is often made narrower than ideal, increasing the risk of tracking across (how many actual failures does this cause, I'm not in a position to judge, but I thought it worth mentioning anyway).
Many devices will, of course, still require a DC-DC converter to step down whatever voltage we choose for the DC branch circuits, to the lower voltages they use. These will have fewer problems than the small "off-line" supplies, though (power factor is irrelevant to DC, only minimal surge protection should be necessary, isolation if still needed will be much easier than at full mains voltage), and can be made smaller.

Another arrangement, used for low-current supplies that don't need isolation from the mains, is the capacitive dropper. They have two main failings - a horrendous leading power factor (though not usually that big of a deal in the overall scheme of things), and many of the capacitors used have a bad habit of losing capacitance, resulting in unreliable operation of the appliance (or no operation at all if they get bad enough).

What plug?
In Australia, as has been mentioned before, we have a plug (Clipsal 492/32 or equivalent) designed specifically for ELV supplies, with two pins in a "T" configuration. Construction is similar to our mains-voltage plugs and it's rated for up to 15A, at 32VAC nominal (and 110VDC?). Unfortunately, there's a disagreement between Victoria and the rest of the nation over which polarity it should be wired with, so if we go with it, the appliances should definitely be protected against reversed input.

As for the rest of the world, I don't know...

What voltage to use?
Obviously, this is a compromise between safety and efficiency (keeping losses down), but I'm thinking a nominal 48V or failing that, 24V (either of which is reasonably "standard"). I would then design the appliances to handle at least between 90% and 110% of the nominal (43.2-52.8V, or 21.6-26.4V respectively). (As for individual tolerances, I would allow perhaps -0% to +5% at the PSU, and a drop of 5% of the nominal in the circuit wires. The remaining 5% under that can be taken as allowing for any further drops in the appliance flex, etc.)

Not being entirely confident that separated extra-low voltage (SELV) can be maintained indefinitely, I would rather go with protected extra-low voltage (PELV), where the negative conductor is earthed at the PSU. (A "crowbar" circuit could then be installed, to shunt positive to negative if a grossly excessive voltage somehow ends up there.)

I'd probably go with 15A per circuit, that being the rating of the Australian ELV plug.

So I'll do a quick bit of math to check that these are reasonable demands: Let's take 2.5mm2 conductors, which will have a resistance of approximately 8 milliohms per metre per conductor (allowing a modest temperature rise), so will lose 240mV per metre of cable (counting both ways) under the mentioned 15A load. So with the allowable 5% drop (2.4V out of 48V, or 1.2V out of 24V), we can run the cable for up to 10m from the PSU with 48V, or 5m with 24V. That seems workable enough to me (if you need to extend it a bit further, you could use 4mm2 or maybe even 6mm2 for a proportional increase in length).

Switching, fusing
One of the problems with DC, apart from being rather impractical for large-scale power distribution, is that it's much harder on opening mechanical contacts than AC of the same magnitude (which is why lots of switches, for example, are marked "AC only"). Mains-rated (250VAC) relays, for example, seem to be typically rated to switch only 30VDC. Switches are similarly affected (though some with reasonably respectable DC ratings can be found). The obvious solution, of course, is to use electronic switching (which is quite a bit easier with DC than with AC) - though the fact that semiconductor failures tend to be as short circuits may present hazards in certain situations.

Fuses suitable for 24VDC, and even 48VDC, seem easy enough to find. A 48VDC rating also seems typical of (AC type) DIN rail MCBs (when specified); and there is an ample range of purpose-made DC MCBs, in any case.

The big PSU(s)
These would be switching supplies, but of course much larger and more powerful than those for individual devices. I'll give an approximate calculation of the power you could get out of one: Let's start with a 10A input, for example (the rating of an IEC 60320-1 C14 inlet, or the basic Australian/NZ outlet), and we want to work down at least to 20% under the nominal voltage; if we take that as 200V (used for high-power appliances in Japan - and of course there's 208V between two phases of a 3-phase supply with 120V from each phase to earth, as used in North America - so I've covered two bases at once here, as well as European 230V), then that gives 160V we need to work down to. Multiply that by the 10A (not accounting for power factor, but the unit should have active PFC to begin with) and we have 1600W available; multiply that by an efficiency of (say) 90%, and we can put out 1440W of DC (30A at 48V, or 60A at 24V altogether).

In order to facilitate easy replacement (as like with any SMPS, it's inevitable that some will fail), I would like the units to be mounted on standardised brackets, with pluggable connections in order so that anyone (able to obtain a replacement PSU) can do it. To keep the outgoing cables to a sensible length, the PSU should (obviously) be installed in the same general area as the DC outlets it feeds.

These big PSUs could also include "bells and whistles" that would be impractical for little plug-pack/"wall wart" units; from almost any fault protection you can name, to a display showing various status attributes. Given how much they would cost, of course, you wouldn't want to skimp on build quality, either.

Lighting and alarms
LED illumination is another obvious candidate for use on a DC system, given that LEDs natively require DC (albeit current limited rather than at a fixed voltage). Running the lights off 24/48VDC instead of the mains, also leaves us with one less hazard to contend with in the roof-space.

Alarm systems (for fire, intrusion, etc.) are also great candidates for use with this type of system, which segues us neatly into...

Battery back-up
Providing back-up power to mains operated devices is a relatively complex affair (it needs to be converted to DC and stored in a battery bank, which then has to be converted back to AC when needed). For DC loads, it's a bit easier as they could be simply run at the battery voltage with little more than a fuse or circuit breaker in between (plus maybe a cut-off to prevent the battery from discharging too far and being damaged). (These loads will probably need wider voltage tolerances than -10 to +10% to accommodate the discharge curves of various battery types, though.)

Further isolation (at the loads)?
It may be useful in various situations, for functional reasons (e.g. avoiding ground loops) and/or extra safety (e.g. in wet environments).

Where the purpose of the isolation is solely functional, that'd be almost trivial to meet; a dielectric withstand of 500V or so should pretty much cover it. There may be a small capacitor or two needed across the isolation for RFI suppression, which I might rate for 1-2kVDC conservatively (to withstand any common-mode surges that may be induced into the wiring). I would also consider it prudent to add a high-value resistor (i.e. 100k-1M) from the input negative to the output, to bleed away any static charge that might otherwise build up.

When added for safety, it may be stronger, perhaps even up to the equivalent of a single-level mains voltage isolation. (You shouldn't ever have mains voltage on the ELV wiring, of course, but that's not to say some buffoon won't put it there at some stage...)

A short-list of various viable loads
  • Lighting, particularly LED
  • Brushless DC fans
  • Some power tools
  • A/V equipment, such as TV sets, media players, radio tuners, preamplifiers, headphone (and smaller power) amplifiers
  • Video game systems
  • PCs (less power-hungry configurations, at least), monitors, and peripherals for them
  • Certain lower-powered heating devices, such as soldering irons (and maybe an electric blanket?)
  • Many novelty items
Basically, this includes most loads up to perhaps two or three hundred watts.

In conclusion
So as far as I understand it, the advantages of this kind of set-up would be:
  • Safety - less mains wiring to worry about
  • Easy battery back-up (for lights, alarms, etc.)
  • Avoids at least some of the problems that plague small PSUs
That leaves us with the following disadvantages:
  • A bit of extra voltage drop in the ELV wiring under heavier loads
  • If the PSU fails, it brings down the power to several items instead of just one (that's not to say you couldn't implement redundant PSUs like in high-end servers, of course)
  • If you want to use an appliance specifically made for the ELV DC system in a building that hasn't had one put in, you'll have to bring along a portable version of the PSU (but I think I can live with that)
In a way, we've already made a small step in this direction, with the availability of wall outlets with built-in USB charging ports. But I'm not overly keen on the idea of hard-wiring a small off-line SMPS (or several) into the electrical installation, having discussed that they already have enough problems. rolleyes Adding USB charging sockets with their required 5V derived from the 24V/48V/whatever by small DC-DC converters would, of course, be quite straightforward.

Feel free, now, to discuss how well this would work out, if I've missed anything, or if you would do it differently.