Whiteley Research Inc., 456 Flora Vista Avenue, Sunnyvale CA 94086 wrcad.com
|
|
|
Whiteley Research Inc., 456 Flora Vista Avenue, Sunnyvale CA 94086 wrcad.com |
This page may evolve into a semi-blog containing descriptions of
miscellaneous technical problems that I have managed to stumble
through, posted in the off-chance that this information will save
someone a few hours of head-banging.
DSL drop-out with incoming/outgoing callsFor 10 years or so I have had a problem with DSL de-syncing whenever there is an incoming or outgoing call. I've been through three modems, and many hours spent with technicians, both on the phone and house calls. Nothing worked.I finally found the solution here. There is a small thingy called an "MTU" in the gray box where the phone lines enter the home, one for each line. It lives in the little sub-box for the line, which snaps out (see pictures in the reference), and is a circuit board covered with black encapsulant. These are no longer used by the phone company. You can either call the phone company and have it removed, or do what I did and clip it out and splice the wires. Radio Shack has official telephone "splicing peanuts". Motherboard Re-cappingIt seems that there was an epidemic of bad electrolytic capacitors manufactured in the far East around 2004. These were used on motherboards, causing huge trouble.Symptoms: the computer starts to crash in non-repeatable ways, it gets worse over time, eventually it won't boot. Dell is well aware of the problem, but my computer was out of warranty, and Dell was of absolutely no help. The model sc1420 was discontinued, and Dell could not supply replacement parts. The internet provided info about the capacitor problem. Inspect the motherboard. Look for electrolytic capacitors with a rounded or bulging top, a good cap will have a flat top. The bulging caps are very-likely defective. These can be replaced, but it takes a bit of work. Each of my motherboards, the original and the "new" surplus replacement, were made in 2004 and each had 31 suspect caps. Each cap was 2200uF/6.3V. I bought 100 replacement caps from Mouser Electronics ($30 + shipping). I found that the solder used on the motherboard was not standard solder, and melts at a higher temperature than normal solder. So, my existing precision soldering iron was useless. I bought a temperature-adjustable Weller soldering iron at Frys ($100). It was time-consuming, but I managed to remove all of the caps from one motherboard, and clear the solder from the mounting holes. I soldered in the replacement caps, noting the polarization. The motherboard was completely dead before I relaced the caps, but to my shock, actually (since I was sure that I had probably damaged something in the process) it booted right up. (Note added: still running fine a year later). Tale of Woe, and Sun Disk Drive ReplacementNote: This is really old, but sort of entertaining. Sun Solaris is no longer a supported platform.The Sun workstation which has been used to support the Solaris distributions has had a hardware failure, which delayed posting of the last Solaris updates. The ancient SparcStation LX machine has two disks, the boot disk that contains the operating system, and a second disk that contains the user login space. The second disk was having problems: after boot-up, large parts of the file system would not be accessible, and attempting to access these areas would cause hardware failure messages to appear on the console. When this first happened, I tried the experiment of reformatting the drive, after which it seemed to work fine, so I built new file systems and reinstalled everything. This cycle was repeated two or three times over intervals of a few months. Lesson #1: When a disk drive does this, get rid of it. The disk has worked for several months, but on update-build night the problem reappeared. This time, the format failed, too, so it was clearly time for a new disk. A replacement was on-hand, and installed. The system would not boot. Efforts to make it boot (i.e., powering down the second disk, booting, and powering it up) succeeded in corrupting the boot disk. Lesson #2: Don't do this! After reinstalling the operating system on the boot disk, all traces of the /dev entries of the second disk were gone. The system would still not boot with the second disk powered up. Since the machine does not have a cdrom, installation was via another machine on the local subnet. This was relatively simple and worked well, but is an unbelievably slow process on an LX. I did a "probe-scsi" from the "ok>" prompt. The result was bizarre. Every scsi address was connected to the new drive. I knew that the drive worked, since it had been used before. The problem turned out to be as follows. The drive was installed in an external box. Rather than use jumpers to set the scsi address, there is a connector and cable that lead to a switch on the back of the box that allows setting the scsi address. Apparently the inductance/resistance of the cable was causing this problem. Using jumpers fixed the problem. Lesson #3: Always do a probe-scsi after connecting a new scsi device. This time, the system started into the boot sequence, then paused with a "bad magic number" warning from the second disk. This was not unexpected, since Sun places an identifying number in the disk label, which wasn't there yet as the drive was a commodity drive. The system paused for a long time, with the indicator led for the boot drive on. Resetting or turning the system off at this point would probably corrupt the boot disk again. Just when I thought that it was hopelessly hung, the scsi bus timed out and the system booted up. Lesson #4: Scsi timeouts are your friend. Don't give up, wait for the timeout. How to get the /dev entries for the new disk? Simple if you know the right commands. These are drvconfigGiving these commands allowed the new disk to be accessed. The new disk was repartitioned and labeled with the "format" command, and everything was fine. All that is left to do is reinstall and rebuild 500Mb of program files. |