Pages
21
0100 1-OCT-67
On write to the DISK the third word of the buffer will occasionally have the second word OR'ed into it. Sometimes it is being given bad parity, to boot.
Test 12, random number generator, seems to leave parity errors in about half the disk. I have left the disk in that state.
John Sauter
0420 1-OCT-67
While running a new address tester the disk made a funny sound. It turned on the "sector address error" light, and, a few minutes later, "cabnet warning temperature" and "stator temperature". It also turned off "unit A operative" and "unit A". I tried to get 'hold of Dutch or Leo, waking up people and operators
22
from here to LA with no success. In the absence of anything better to do I will do nothing until somebody shows up in the morning.
John Sauter
1725 1 Oct 67
Tested Read errors that we write on disk. Found the pattern[s?].
word 1 111111 222222 word 2 333333 444444 word 3 555555 666666
all 0 other: words: 0
Will consistently produce permanent read errors.
SR Russell
1745 1 Oct 67
Note that the check bits that fibrascore writes are the number of
23
zero bits [moulo?] 4 in each sector. The self-tester is based on a 16-bit shift register that shifts once an 18-bit byte time. THus it will [??way] shift the pattern 4 times in each sector, and so the check character will always be its initial value. This is also true of almost all test patterns, such as counting, etc.
Writing the pattern:-
word 1: 200000 00000 all others 0 gives permanent read errors on secter 3 of TRK 0
the pattern: word 1: 577777 000000 all others 0 gives no read errors.
Dutch found 2 bad modules in check bit computation circuit. Replaced them. Now above patterns work. Tried also
777777 00000 | 777777777777 |
---|---|
0 | 777777000000 |
777777777777 | 777777777777 |
0 | 777777777777 |
24
They all seem to work. This may explain the mysterious "Timing" errors, but I don't think it does completely.
Dutch also set timing on clock to us on write to be 50 ns earlier than it was previously.
SR Russell
1900 1 Oct
Looked some for word 2 + 3 [oring?] but didn't find any evedice. Wasn't a very thorough test, though.
SRR.
0040 2 OCT 67
DATA MISSED AND NON EX MEM LIGHTS IN 176 ARs OUT. LAMPS ARE OK. CONI GIVES CORRECT BITS
SRR
25
0125 2 OCT 67
Put disk in self test.
SRR
0220 2-OCT-67
Began running diagnostic. Disk is on line.
Bit 28 is being dropped sometimes. It happened three times while address testing, then the address tester ran all the way through without error. The three times were : record 231267 word 10, record 231663 word 14, record 231665 word 20.
'Random DATA' found garbage on BAND= 128+4+2+1, TRK=0, but re-running found good data there but SAME garbage (651703) on BAND=64+32+8+1 TRK=1. On rerunning it happened on BAND=64+32+16+8+2, TRK=1. While writing on this band (I didn't notice whether TRK=1 or 0) the SAR stopped