Computers Overview
Commodore PET
        RAM (Extended)
Sinclair ZX80
Sinclair ZX81
BBC Micro
Commodore 64
Sinclair ZXSpectrum
Memotech MTX
Memotech CP/M
Tatung Einstein
Atari ST
Commodore Amiga
DEC 3000 AXP
Raspberry Pi



The Commodore PET (Model : CBM 8096)

Repairs - RAM


When the machine was powered on, it started up with an encouraging CBM 8000 series "chirp" and the screen displayed the "basic 4.0" welcome screen, but only reported "15359 bytes free".

A CBM 8096 should report "31743 bytes free", i.e., there appeared to be a RAM fault on the main board. (The 64K additional RAM is not reported at start-up.)

This is the 32K main system board, with the 64K mezzanine board removed.

My 8032 uses Hitachi HM4716AP-4 RAMs, these are 16K x 1 bit, installed in two banks, identified as UA5 to UA19 on the board - located below the heat sink for the voltage regulators.

The lower 16K is provided by the odd numbered chips, UA5, UA7, . . . UA19 and the upper 16K is provided by the even numbered chips, UA4, UA6, . . . UA18.

15359 bytes free equates to a missing 16K (31743-16384), suggesting a fault in the upper bank of RAM.

I wrote a simple BASIC test program to poke values into a range of memory addresses, read them back and compare them. When "0" values were written, they were read back correctly.

When "255" was written, all addresses returned "253", i.e., data bit 1 (2) was stuck low. The PET schematic on show that bit 1 in the upper 16K block is stored in UA16*.

With UA16 removed, you can see that the separation between the solder pads and adjacent tracks on the component side of the board is tight, but is adequate to allow a socket to be fitted in place of UA18.

* Thanks to Dave Curran of Tynemouth Software for this tip, check out his vintage computer blog pages.

New socket installed in position UA16 and replacement 4116 RAM fitted - a Mostek MK4116N-3IRL

With the RAM chip replaced, the upper 16K is now visible and the system correctly reports "31743 bytes free"
The output from my memory test program running a test over the 16K boundary, i.e., 15700 to 16400.

I now have a working PET 8096, or at least the 32K on the main board is good, the additional 64K on the expansion board requires additional software to test it.

Update : see here for expansion board RAM testing


mailto: Webmaster

 Terms & Conditions