By way of a break from the refurb posts; I was recently part of a conversation on FaceBook regardig a coulple of 48K Rubber Key Sinclair Spectrums. Both were proved to have faulty Upper RAM problems by issuing the command ‘CLEAR 65535’. Both machines reported ‘M Ramtop no Good’.
Ian Gledhill, of Mutant Caterpiller fame, was helping Francesco Morandin diagnose the problems. Ian went on to request Francesco issue the commands POKE 50000,85 then PRINT PEEK 50000 and POKE 50000,170 then PRINT PEEK 50000. By comparing the bit patterns expected to those that were reported, the individual ICs at fault could be found.
In this instance, Francesco got 85 for the first test and 186 on the second test for the first machine, then 213 for the first test and 170 for the second test on the second machine. So, working through:
If we then break the values down into binary:
I have underlined the bits that aren’t reporting correctly. After a quick exchange with Ian, I drew up a diagram to show what bits correspond to what ICs in the 48K Spectrum.
By finding the IC responsible for the mis-reporting bit, you can assertain whch one requires replacement. So, for the first of Francesco’s machines, IC19 is indicated as being faulty, and the second is IC22.
I have made a PDF of the diagnosis table which can be freely downloaded from HERE.
Co-incedentally, I have a 48K Spectrum with the same problem so I had the chance to use this technique for myself. I did the POKE/PEEK test and recieved 85 for the first test and 171 for the second test. Looking at the binary pattern for 171 (10101011) showed that bit 1 was reporting incorrectly. This indicates that IC15 is the culprit in this case. I have plans for this faulty machine to become a test bed so I’m not repairing it straight away.
A big thanks goes to Ian Gledhill for his insights and Francesco for posting his results.