/ documentation / repair notes /

Repair notes for various computer equipment

AlphaServer 1000A Storage Bay Door Button

The button of the storage bay door is broken and now the storage bay door is constantly open.

The "hinges" of the bay door button are two small plastic pins and are not very strong. After some time pushing the bay door button the pins broke and the button can no longer hold the storage bay door in its place (i.e. swings open every time). So, I made holes in the face plate where these plastic pins used to be and inserted an iron pin which would hold the bay door button in place and thereby the storage bay door.

DECpc Trackball Eject Mechanism (1)

The DECpc laptops have a funny mouse pointer assembly (it's actually called the Trackball). The trackball assembly is mounted on a metal sheet which is curved upwards at the end of the trackball assembly. When the trackball is unused it is retracted in a small cavity of the system casing. In order to use the trackball, the user pushes the eject button near the trackball assembly causing the assembly to pop out of the cavity. The trackball is then positioned along the side of the laptop casing.
Unfortunately, handling the eject button to hard may break a plastic pole of the eject mechanism. Attached to this plastic pole is a spring which (indirectly) holds the trackball assembly in the aforementioned cavity. If the pole breaks, there's no tension to the spring and the trackball assembly contineously pops out. If you still have the broken piece of the plastic pole, you can try glue to fix it. If the pole is gone, you can try to make a small cut in the stub of the pole. Next, you can hook the spring to the stub and ensure the spring aligns with the cut you made. There's enough room for the spring to hook to the stub. The eject mechanism should now be working again.

DECpc Trackball Eject Mechanism (2)

Someone has been rocking the trackball assembly too hard, so now the metal strip attached to the trackball no longer ejects properly. The metal strip is held in position by a plastic cover. This plastic cover is attached to the base chassis by a number of tiny nails sticking through small holes in the base chassis. The nails are made of plastic and they seem to be molten to the plastic cover. Strong enough for normal use, but not always.
So, I drilled some holes in the plastic cover, matching the aforementioned small holes in the base chassis, and used tiny screws, washers, and nuts to tighten the plastic cover to the base chassis. The screws and nuts need not be too tight else the metal strip of the trackball assembly won't move smoothly.

HSG80 Resurrection

I used the CONFIGURATION RESET command on two HSG80 controllers in an ESA12000 array. Apparently you shouldn't do that because it renders the maintenance port inoperative and you loose the ability to configure the controller. Admittedly, the HSG80 CLI Reference Guide does mentions the command disables communication between host and controller.

The controller does seem to be operational given its flashing Reset button/LED, but you just cannot access it. So, I picked another HSG controller which was working but had a slightly higher firmware version. I used the FRUTIL utility and tried to remove & re-install the "dead" HSG controller into the ESA12000 array. Too bad, the firmwares were not compatible (OCP Code 3D, "NVPM structure revision greater than image's."). I guess this error relates to some data structure in the non-volatile memory of the controller. Suffice it to say that the "dead" controller was now really dead.

Clear NV-memory and Restore Maintenance Port

There's a coin cell on the PCB board of the controller which gives you the opportunity to clear the NV-memory. All you need is a long, thin, plastic strip, and stick it between the coin cell and its retainer clip. Wait maybe 15 sec. and then retract the strip. The coin cell is located behind the reset buttons, but you don't need to remove the top cover of the controller. If the plastic strip is long and flexible enough, you can reach the coin cell through the GLM access door. Like this. The controller should be operational again, but the maintenance port is still not accessible. Using the FRUTIL utility on the surviving controller to re-install the "dead" controller should bring it back. You may need use the SET FAILOVER command or the SET MULTIBUS_FAILOVER command to get the controller configuration correct.

I guess you don't have to clear the NV-memory as long as the firmware of your HSG controllers is compatible in which case a re-install of the "dead" controller should be enough.

Next, it could be that the resurrected controller now has lost its serial number and its hardware revision number. I'm not entirely sure about this, but the controller may signal an Error 4090 ("Module has invalid serial number."). In this case the serial number has been reset to "CX20100000" while the hardware revision is set to "0000". A command exists to correct these numbers. I've used the following commands on the resurrected controller to restore its serial number:

  3. DANGEROUS X SERIAL="<serial-number>" PRE
  4. DANGEROUS X HARDWARE=" <hw-rev>"
  5. Restart OTHER controller.
NOTE: I forgot to correct the hardware revision, but the DANGEROUS command above should do the trick. The hardware revision can be found on a label on the PCB. Also, before using the command to set the serial number, you can postfix the command with a question mark; there are some more options.

Replacing Coin Cell

I've also encountered an OCP Code 30 ("The JSRAM battery is bad."). It appears this refers to the coin cell mentioned earlier. The coin cell is a 3V lithium cell (CR2032) and in order to replace it, you need to remove the top cover of the HSG controller. You can follow these steps:
  1. Remove the GLM access door from the top cover as well as the controller's program card. Strictly not necessary, but removing these two items makes removal of the top cover somewhat easier.
  2. Remove the left and right handle of the controller. You can use something like a small nail and a pipe wrench to push the captive pin out of the handle.
  3. Remove the 5 screws holding the PCB and the top cover.
  4. Gently lift the top cover such that the thermal pad of the i960 policy processor comes loose.
  5. Slide the top cover off the PCB.
  6. Replace the coin cell which is located behind the reset buttons.
  7. Re-assemble the controller and use the FRUTIL utility to re-install the controller (see above).

Dead MGE Pulsar Evolution 500

Two Pulsar 500 UPSes died after power outage, completely dead. It turned out that the contacts in the battery connector doesn't always connect properly. If this happens the UPS won't come up. I've read somewhere that the batteries must hold some power to start the inverter circuit of the UPS. If the contacts in the battery connector don't connect, it would explain why the UPS remained dead. So, I checked the contacts and twisted them a little bit hoping they would make proper contact. Powered up the UPS and... success! Both UPSes are working again.