Encouraged by my recent repair of the TS450 that had languished without computer control for a couple of years, I decided to reach back even further in the time stream and to pull my trusty Compaq nc6000 out of the vault. That laptop was my main computer for several years, starting in Sri Lanka right after the tsunami, then back to Bangladesh, a year in Virginia, and in Belgium up to 2008. In that year, it developed an annoyingly intermittent failure that slowly increased in frequency until the machine was unusable — it would just power itself off. It got to the point that it would turn off in only a few seconds, so my initial thought that it was heat-related didn’t hold up for long.
I noticed that if I pushed on a specific place on the upper edge of the keyboard, just to the left of the fan, I could convince the computer to remain powered up for as long as I held pressure. I though it might be a bad motherboard edge connector or perhaps something to do with the fan itself, so I tightened down the fan. At the time, HP (which had recently acquired Compaq) was of no help, but over time, users consistently reported the same failure mode and kludgy solution. Finally, someone must have looked at the computer under a microscope and diagnosed that these failures related to microfractures in a single chip: the maxim 1987, a 48pin QFN chip that controls CPU power. If you google that chip, you’ll find scores of reports of its failure, either intrinsically or due to poor soldering connections. For my model, most of the speculation is that stiff hinges flexed the motherboard and caused mechanical failure, but I wonder if the failure is more related to repeated heating and cooling of the chip in operation.
The fix for this sort of thing is to resolder the chip, so that’s what I did. Unfortunately, the chip is on the underside of the motherboard, which is buried deep in the laptop. There’s nothing to do (aside from dremeling through the case) but to disassemble every module in the computer, completely remove the main board, flip it over, solder, and put it all back together. Torx screw drivers are not optional — there are a lot of torx screws of varying sizes, plus a few tiny philips head screws.
I used a sparkfun hot air rework station 303D set to 385C and toasted the chip for 5-10 seconds per side. I could see that the solder melted as I did so, but had to take it on faith that the chip was sitting correctly and that all the connections were good (direct visualization is not possible).
In disassembling the laptop, when I removed the heat sink fins near the exhaust port, I found a cat-like fuzzy creature composed of many years of hair, link, fur, dust, and whatever else has been in the air that I breathe. I looked much worse than the air filter that they force you to inspect every time you take the car to JiffyLube. I did a general cleaning as I went through the computer, dusting out cooked ants, moth wings, and other animal life remnants. Since I didn’t want to cook the CPU, which is not far from the max1987, but on the other side of the board, I removed the heat sink and the chip itself while I was soldering. I cleaned off the chip, removing dry and questionably effective heat sink grease, and reapplying some of my own industrial grade heat sink compound.
I was surprised as anyone when the machine went back together with nary a left over screw. After five years out of action the LiON battery was dead flat, but I put in back in position anyhow. When I powered up, I got the bios splash screen followed by the Windows XP/SP2 screen. Windows did not fully load, but got to the point where the screen turned light blue and I got a usable mouse cursor. Program manager and task manager weren’t working, so I couldn’t get much further.
I suspected that the drive had been corrupted by erratic shutdowns, so I ran SpinRite from CD-ROM at level 2 and it detected and repaired some flawed sectors. On the next boot: success, 2008-style. I was staring at the same screen that I had used in 2008, now oddly dated looking. I fired up a few programs, and everything ran correctly. I was not, however, able to get the computer on my home network despite being able to physically activate the wifi module and to detect local networks.
I plugged the machine directly into my router and spent the next couple hours watching netflix (on my other computer), while windows update did its thing. Many were the times that I heard the windows boot sound. I enjoyed watching the patches of yesteryear strut their time across the stage: NET framework, Windows “Genuine Advantage”, and finally, the 800 pound gorilla: service pack three.
At the end of all this, I found that my home wifi didn’t work, but I could connect to an open network, so it was a problem with encryption. I realized that WPA2 wasn’t out when the machine was manufactured, and that my home router would not roll to a lower standard. I was pleased that the Intel wifi card was supported by an update that came out in 2009, and which enabled it to connect to modern wireless access points with good security.
So, the nc6000 seems to have made a full recovery. I’ve promoted it to “back up logging computer” for the station, which is great timing, since I now have two computer-controllable radios. A new battery (a bit less than $40) has been ordered, and will complete the update.