Some experiences to share with power supply problems in a A928 i-Buddie desktop of ECS / Elitegroup and with loss of HDD connectivity.
problems observed on a version 4 board:
1) erratic and unpredictable start behaviour; sometimes no power light and nothing at all, sometimes powerlight but no fan noise nor bios, and sometimes OK. Disconnecting power cord and reconnecting it also seemed to have influence.
2) after some time of operation loss of connectivity with the harddisk HDD. In windows this obviously causes the blue screen of death and perhaps corruption of HDD data. In BIOS the HDD is not detected anymore, the CD-R/DVD is still seen. BIOS sees that there is a device as it dwells some time on trying to detect it but doesn't manage to communicate with it to even establish its identity. Upon removal of the HDD the BIOS sees that there is nothing there and shows the CD-R/DVD drive instantaniously.
Remedies tried on 1):
(unfortunately cannot tell which of all these things is exactly the remedy but in the end the desktop started to function as it should with suprisingly little heat being developed)
Have no service docs of the A928 but found on the net the service doc of the Mitac 8500 and Mitac 8677 which to a large extent use the same keycomponents in the power supply (Maxim's max1632 and Intersil's hip6301) and north/south bridges (Sis650, Sis961, or XP4's HT7650 and HT7961).
There is much written in this thread about melted power plugs and receptacles and following damage to the i-Buddie.
My guess is that in either of the power supplies (the max1632 does 5V and 3.4 V, the hip6301 does the processor core voltage like 1.5 V) then a classical switch-mode power supply problem has occurred that in any of the pairs of power transistors-in-series both transistors for some time are conducting although it is the intention that it is always either one or the other. This in fact causes a momentary short circuit over the input power and obviously high dissipation. Not enough to make the built-in smd fuses trip but enough to melt the plug and perhaps damage the power transistors and their drive circuitry.
It seems that ECS did not follow intersil's instructions to have independent decoupling of the power supply to each of the three converters that the hip6301 is driving in the A928. All three are in parallel and decoupled only with 6 electrolytic capacitors in parallel. So there is very poor isolation between the three stages and havoc may occur.
Also obseved that the MAX1632, although creating the voltages it should, sometimes thinks it is not and signals the rest of the circuitry not to start up (only power light situation) via its reset output.
Actions:
* 1 uF, 5V tantalum capacitor parallel to C458, next to the max1632
* 3 pcs. 0.22 uF capacitors in positions C200, C201, C202 (were not present; to be found at the back of the six electrolytes)
* mount C152, 86 uF, 22V, plus towards processor
* check BIOS and update to latest (20 aug 2003) with a flash/eeprom programmer
* check cmos battery ( > 3 V) and renew if necessary
tip: remove the cmos battery holder, connect it to the previous position via thin isolated wires soldered at the ends, make it end up in the space between the keyboard connector and the touchpad connector, put battery in and isolate it with tape. It can fit snugly under the keyboard and the battery can now be replaced without taking the entire computer apart.
This made my i-Buddie react normally to power up and down commands.
The plug stays cool. There are clearly issues in the board lay-out and in the chosen circuit and in some components not having been placed or having aged that cause transient voltages and erratic bahaviour of the sensitive switch-mode power control ICs and the power switching circuits that they control. My desknote now runs for several hours from the external ecs battery (2.4 GHz P4, 512MB DDR), the fans run at very low level and are hardly audible, and it certainly doesn't draw enough current to melt a plug. Most noise comes from the little fan in the mains power supply that connects to the desknote or to the battery. If you see heating of the power plug and receptacle only replacing those won't help; the cause is elsewhere in the circuitry, the heating of the plug is just a symptom.
Another tip: clean the heatsink surface to shiny copper, clean the P4 metal surface and put some fresh heat conducting paste on the P4. Before fastening the 4 screws gently move the heatsink in tiny 8 movement until you feel the heatsink stick to the P4 as the paste has evenly been distributed. Now fasten the screws, diagonally opposite in stages, while exerting pressure on the center of the heatsink so that it doesn't get undone from the P4 by uneven pulling by the screws. The P4 will report lower temperature that way (the heatsink is actually of excellent quality) and the fans will make less noise as they are turned lower by the system.
Remedies tried on 2)
rather brute; take motherboard out, 961 chip that does IDE facing up (next to mini PCI modem board), remove modem and plastic/rubber isolation material, take piece of alu foil and cut a rectangle out slightly bigger than the 961. Put foil in position with the 961 sticking through the opening and alu foil touching the PCB so little air can get under it. With a hot air gun heat the 961 chip to melting point of solder (close to 200 C). Let it gently cool down. Don't exert force on the chip, it's just to make the many bond pads reflow once more. The foil is to protect the surrounding area. Remove foil, reassemble the desktop to the stage that IDE function can be tested. In my case the loss-of-HDD problem had disappeared so it must have been a bad contact to the 961 when the chip heated to its normal operating temperature.
All the above is clearly in the spirit of hobby solutions and I take no responsability for the reader following any of these hints.