A linear regulator will essentially dump the excess voltage as heat. So, for example, if you take a 1A 5V linear regulator (ie an LM78CT05), and wire it across a 12V line, you are getting out (at full load) 1A @ 5V as usable voltage (5 watts), but the regulator is taking 1A @ 12V (12 watts) to produce it. The excess 7 watts is dissipated as heat, so one needs to provide for facilities to get this heat away from the circuitry before something horrible happens, ie a heatsink. The same regulator run from a 16V power supply would dissipate 9 watts, and from 6V only 1 watt. You get the idea.

So, to minimise the wastage (or conversely increase the efficiency), ideally one wants to run the regulator input as close to it's output voltage as possible. The problem is that there is a certain minimum amount of headroom required to allow the regulator itself to work, usually a couple of volts. Some more modern low-dropout regulators can be run down to only about 0.2V above the output, and are normally short-circuit and overtemperature protected as well. Most linear regulators will work at up to a 32V differential between input and output.

Now, in the case of the empeg main power supply, a switchmode design was used, which is much more efficient. I won' t get overly involved in the technical aspects, but essentially this works by means of a resonant LC circuit which stores energy at one voltage and releases it ad a different one, which can be either higher or lower than the input level. The higher frequency that the circuit uses the more efficient it gets, and the smaller the inductors can be. The chip used in the empeg is the MAX1631, which runs at 300kHz and gives typically 95%+ efficiency. It will actually accept voltages up to 30V, but there are other components in the player that are limited to about 18V maximum, hence the 16V limit (with an overhead for safety). It will also work down to about 8V, although external amps generally won't, which sets the lower limit.

The SMPSU produces the +3.3V and +5V rails, which power the majority of the circuitry. The CPU, memory, etc, all run off 3.3V, while the drives and a few other things run from 5V. The power supply can produce about 3A on the 5V line and 2A on the 3.3V one.

Although switchmode power supplies are very efficient, they tend to be electrically noisy due to the high frequency and heavy current switching going on in the inductors, so they're not ideal for analog equipment. The audio sections of the player are thus powered by a 5V and 10V linear regulator from the main 12V input, but at very low current so efficiency and heat aren't major problems.

The display board 60V supply is produced by yet another SMPSU (MAX770), which is configured as a step up (or boost) controller. This runs from the main 12V input, and produces a nominal 60V, which is wound down to about 40V to dim the display.

Boost SMPSU chips are also very efficient, although usually somewhat less so than step down (or buck) converters. The 770 runs at about 82% efficiency. I designed a +5 and +12V psu board for my UAV project using some of the things, which runs from a single cell. It produces about half an amp on each line, and can do amusing things like spinning up a hard drive from one small lithium cell. The cell doesn't last very long, of course!

Does this help?

pca