Recently continued work on an 12V powered ATX power supply, using ideas and parts suggestions from here. Now have all the required parts to make a full ATX (rather than a partial ATX as he did), but before that it would be a good idea to know the actual power requirements for the motherboard. The tests were a Celeron 1G with Geforce MX440, network card and 380 Meg of PC100, , a Celeron 400 with an old SiS gfx card and 64 meg of PC100 and a VIA C3 533, in a all in one Via Advance 6T with on board LAN and graphics, this was measured at both 533 (133 FSB) and 266 (66 FSB). In the past I've measured a Pentium 166 with on board video, sound card and 56 meg of SIMMs. The current/power requirments were:
3.3V(A) 5V(A) 12V(A) 5VSB(A) -5V(A) -12V(A) W Machine/state
.36 4.73-4.75 0.19 30 uA 0 30 mA 28 P166 - full load
- 3.48-3.52 0.19 30 uA 0 30 mA 21 P166 - playing MP3
.33 2.76-2.77 0.19 30 uA 0 30 mA 18 P166 - idle
4.1 4.8 0.2 0 0.2 0.04 41 C 1G - full load
3.7 4.3 0.2 0.09 0 0.04 37 C 1G - idle
1.1 2.6 0.1 0.01 0 0.01 18 C 400 - in BIOS
0.9 2.6 0.1 0.01 0 0.01 17 C 400 - Fail boot
0.7 2.5 0.1 0.01 0 0.01 16 C 400 - Fail boot, no gfx card
1.6 2.0 0.2 0.09 0 0.01 18.3 C3 533 - in BIOS
1.4 1.3 0.2 0.09 0 0.01 13.9 C3 266 - in BIOS
Using Linux, full load created with gzip < /dev/urandom > /dev/null, mpg123 was used to play an mp3 on the C400, mpg321 was used on the C3.

If you are interested in power requirements for a UPS, then some figures on power usage as measured from the mains side are:
0.02A Monitor Off, Computer Off
0.04A Monitor on standby, Computer Off
0.27A Monitor Off, Linux in Console and Idle
0.42A Monitor On, Linux in Console, find / > /dev/null
0.42A Monitor On, Linux in Console, both HDs sleeping
0.45A Monitor On, Linux in Console, dd if=/dev/hdc of=/dev/null bs=1k
0.47A Monitor On, Linux in Console, Idle
0.485A Monitor On, Linux in Console, gzip < /dev/urandom > /dev/null
0.525A Monitor On, Windows 98, running Unreal Tournement
0.53A Monitor On, Windows 98, Idle
0.54A Monitor On, Linux, xdm login idle
This was for a Celeron 400 in a complete system. TNT2 gfx, sound card, network card, cdrom, 5400 rpm hd and 256 meg of PC100. The monitor was a 17" low end of the market (1280x1024 at 60 Hz, rather than at 85 Hz).

The 12V to ATX power supply is now complete and tested. The board layout is available here (eps), which implements the schematic found at the above link (which themselves came from datasheets) and also includes a full ATX style ON/OFF from this(eps) schematic. In making the board I spotted some problems with this layout. The ground wires are too hard to get to, the -5V bits should have been moved further right to make more room. Forget the Power Good signal, turns out there was enough room on the bottom left of the board to add the capacitor and resister that the reference used. Would have be better to use the bottom right. While the power supply is on, powering the Celeron 400, the 5V regulator gets really hot and needs a large heatsink. By itself it is too hot to touch in 20 seconds. For the Celeron 400, all other parts ran cool - even the 3.3V regulator. Suggests there is a cut off point above which the regs will need a good heatsink - even though the worst case showed the 3.3/5V regs would give off around 7 W each. So, the layout of the regs I used isn't the most ideal for easy heatsinking. The heavy diodes all ran cold and so are fine as they are. In the BIOS the whole system was drawing 2.5A from the 12V supply. Testing on the Cyrix 3 533, running at 266 Mhz, the 5V reg didn't get quite as hot, and the 3.3V reg was hotter than before. At around 2A on the 12V supply this generally drew less current so it was to be expected. Putting the system into sleep mode it draws 1A, but as the CPU fan was still running, I suspect it can go lower. In the end, this heat didn't really matter as the PSU case was then made of 1/8" aluminium plate, folded to cover three sides and then another peice covering the fourth large side. As the pictures below show:

This PSU is then bolted onto the 'case' that encloses the miniATX motherboard. Also on the case is the 2.5" HD, the cut out for the CPU fan and a PSU controller circuit that switches on the PSU and computer based on the ignition. Here is the complete unit and with the PSU removed, from the front and the back. Notice the round base of the case, the height of the whole unit is low enough to fit entirely inside the steel rim of the spare wheel. When fitted, all you see is a near comlete circle of perspex.

The system is now working, in the future will need to talk about: the user interface, to a PDA; the modification of the stardard in-car radio cassette to accept the audio from the computer; and the PSU controller and timer. Could also do with talking about the build of the case and current problems.

Last modified: 3/09/03

Could also now talk about the addition of infrared remote control, which now gives instant wireless control without the PDA. This was done almost entirely with information found here at Linux Infrared Remote Control - with the modification that the receiver was powered via a motherboard usb header, which was wired to a unused wire that had come from the serial motherboard header. Making that 9 pin D type serial port very non-standard but properly powering the receiver all the same. Incidently, the receiver still needed a capacitor next to its power supply pins.

Last modified: 1/03/04