2065 MAINTENANCE APPROACH 
(Cont) 
Some points to keep in  mind while gathering information for troubleshooting are: 
1. 
Does 
the 
problem involve a broad area 
of 
the machine, or 
is 
it 
localized?  Is 
the 
systelT' 
really  dead or getting mUltiple errors? 
(It 
could be a missing clock pulse  or voltage.) 
2.  Always suspect the last thing 
that 
was done 
to 
the system (Ee, program change, 
equipment 
change, etc.). 
3.  Don't trust indicator lamps.  Many CE's have  been misled by burned 
out 
indicators. 
LAMP 
TEST' 
4. 
Is 
power 
up 
on 
all  units? 
Finding The Trouble 
1. 
There are five  general ways 
to 
shoot 
troubles in the Model 65. 
a.  Ripple storage 
b.  ROS test 
c.  FLT'S (CPU troubles) 
d. 
Diagnostics 
e.  "Brute force" (manual method) 
If 
you have a red-light CPU error, run ROS and FLT first.  For other errors, running 
FL T's first may still 
be 
the best way since it takes so little time to run  them. 
Use 
your 
own judgment. 
2. 
As 
you troubleshoot you will  have  your own 
"pilot" 
errors to contend with in 
additklfl to the original trouble.  Some typical 
"pilot" 
errors at the system console are: 
a. 
Rate switch in  wrong position.  (Ever try to load  FLT's with the rate switch in 
single cycle?) 
b. 
Specification error in the instruction you arc trying to repeat or have 
in 
the  data 
keys. 
c. 
Assuming good parity in  GPR's and FPR's. 
d. 
The following sort 
of 
thing will  happen to you 
at 
least once.  Suppose 
you 
have just 
stored a small program in core that 
is 
supposed to loop on itself.  You set 
Ie 
to the 
beginning location, set Rate switch to Process, and activate Start.  Instead 
of 
running 
your program, the machine  does nothing.  You left the wait bit on in  the 
PSW. 
After clearing the wait bit and resetting the 
Ie 
you again activate Start.  Now instead 
of 
running your program the machine hangs up in  a tight loop with the 
Ie 
at 1 
or 
2. 
This time; your program has something wrong that caused an interrupt.  The 
interrupt sequence stores the current 
PSW 
and pulls 
out 
a new 
PSW. 
Since you 
probably cleared core in  the beginning, the new 
PSW 
is 
all zeros.  This means that 
the machine 
will 
start executing instructions after the interrupt, starting at location 
zero (which also 
is 
cleared) and will cause a program interrupt (zero op code). 
It 
then goes hung up in a tight interrupt loop. 
e. 
If 
you try 
to 
do any store class 
of 
instruction 
(anyone 
that puts something in 
storage - e.g., ST, STH, CYD,  RDD, most SS, and others), 
it 
will bomb 
out 
if 
the 
PSW 
key 
is 
not 
zero (unless you are  lucky enough to get a match). 
S/360 
MODEL 65 
53