I used to require my computer to be on standby (just PSU connected to mains, not the hibernate-like thing) for a couple of reasons:
When I was studying I found it useful to be able to turn my computer on remotely (WOL via my router which had a port forwarded to the LAN broadcast address). I mostly used it if I had been working at home and forgot to commit my work to our SVN repository; it allowed me to power on my home machine from the university, log in via SSH, commit my work and power off when I was done.
Later I started to use my computer for automated TV recording which required it to be able to start up by itself (scheduled via ACPI).
Now that I have moved my MythTv backend to my server which is always on (using approx. 10w) it is time to switch my desktop computer completely off when not used.
First a few words on my original setup. In Denmark the "electricity-saving-foundation" has developed and promotes so-called "USB power saving rails" (directly translated from danish, I don't know the english word for these devices or if they even exist outside of Denmark). Anyway, they are basically a power strip with a relay attached to a USB cable and the idea is that you plug the USB cable into your computer and let your peripheral devices (monitor, printer, speakers etc.) be powered from the strip and hence they are only connected to mains when the computer is on. The goal is to reduce the amount of energy wasted on standby. I have used a setup like in the figure below for many years.
In my new setup I will also power the computer from the "power saving rail". I have seen this done with laptops, which is straight forward as they can power up from the battery, on stationary computers it is a little more tricky. My idea is to insert a small circuit between the computer and the "power saving rail". The circuit will charge a capacitor while the computer is on (besides powering the relay in the power strip). When the computer is off it will be possible to energize the relay using the charge of the capacitor and if the computer BIOS is configured to power up after power loss, the computer should be able to provide 5v for sustained operation before the capacitor is discharged too much to hold the relay.
Simply charging a capacitor to 5v will not be enough as the voltage would quickly drop below the threshold for the relay and self-discharge may even mean that it would not be able to activate the relay shortly after power is cut, at least it would require charge pump to utilize most of the charge. Instead I charge the capacitor through a voltage multiplier to get around 22v (5 x 5v - 5 x Vf). Discharge is done through a linear 5v voltage regulator.
If you take a look on the circuit schematic below (click to enlarge) you may notice that I have added a relay parallel to the activation switch. The reason for this is that, while testing the circuit on breadboard, I found that I often released the switch too soon because I was fooled by my monitor powering up (with a logo) straight away while my computer apparently waits a few seconds before it powers up. The added relay, which feeds itself, ensures that the capacitor is discharged for as long as possible. The zener-diode across the coil let current pass, when the coil voltage has reached ca. 5v. This prevent exceeding the rating of the relay and could give a higher input voltage to the voltage regulator (I haven't done any measurements of the actual circumstances, though).
I made a board layout which can be implemented on prefboard and which fit in a project box I purchased for the purpose. Both schematic and board layout was done in CadSoft eagle and you can download them here and here. I haven't put much thought into the component values; the storage capacitor is 2 x 3300uF (the largest that would fit), the multiplier capacitors are all 10uF, the oscillator capacitor is 100 nF and the zener-diode is a 5v1.
The assembled circuit and project box with cuts for the USB sockets and a hole for the switch wire.
Bottom view of the circuit board in the project box. I haven't bothered to screw the circuit board to the bottom of the project box as it sits quiet tight when the box is closed. I actually had to remove some of the mounting studs to make it fit.
The assembled box.
The standby circuit and a "USB power saving rail".
The full setup mounted on/under my desktop. Besides saving some energy, it is much more convenient to power on the computer from a switch on the desktop instead of the switch on the front of the cabinet. So far, the longest the system has been without power was for four days and during those four days the capacitor charge had dropped from 22 to 19v and the system could power up without any problems.