Instruction manual

to be tested by adding the high order ad-
dress and the low address
of each loca-
tion
to a "counter" byte. After all loca-
tions to be tested have been filled
with
their calcuiated data byte, the routine
then recalculates the
data byte that
should be stored in each location and
checks
it against the actual contents of
the location. If the data in memory is dif-
ferent from the calculated value, then
the location and offending bit pattern
are printed. As previously mentioned,
there can be
differences due to "stuck"
bits or interaction of memory locations.
Each time that
the routine is successful-
ly executed,
it will print a "pius" on the
system terminal.
To completely . test
the memory (adding all 256 possible
"counter" byte combinations to the ad-
dress),
it is necessary to have 256
"plusses" printed on your terminal. The
program
listing is exhaustively com-
mented and should be pretty much self
expanitory for even a novice machine
language programmer.
To
test 4K of memory occupying
hex
locations 200 to 2FFF, enter 00 at
0080, 20 at 0081, 00 at 0082, and 30 at
0083 (end address plus
1) and run at
0010. If no errors are detected, you
will
get a string of plusses on your terminal.
Remember
that 256 plusses are required
to
complete the test. An example of an
error would
be a carriage return line feed
on the terminal, a four
digit address (in
hex), a space and a
two digit number.
The
two digit number represents the bad
bit pattern. Now convert
the "bad bit"
pattern to its binary equivalent. Each "1"
in the binary pattern represents a bad bit
at the memory location printed.
If
23A840 was printed on your terminal, it
would mean that bit 6 was bad at loca-
tion 23A8. By reference to the memory
board
documentation, you should be
able
to determine which chip on the
board is faulty.
An
interesting observation was
made during the developement of the
program. My machine is a homebrew
S100 bus, dual processor system. I have
a 6502 and a 6800 on an S100
prototype
board, each sharing all of the system ex-
cept for a
little PROM which is unique
for each microprocessor. The system
clock is derived from the clock generator
in the 6502
(1 MHz.). An equivalent
SUMTEST program for the 6800
would
cycle through my 24K of memory with no
errors detected. The
6502 SUMTEST pro-
gram would
consistantly catch several
bad bytes. Apparently there is a few
nanosecond's
difference in the timing of
the two microprocessors, and that was
just enough for some of the memory to
fail. All of the memory that tested bad on
the
6502 was purchased from one vendor
as 450 nanosecond memory. So
be
aware that a few nanoseconds can make
a big difference, and purchase your
memory from a reputable supplier.
19:74
Usting 1
0UI0
SUMTST OR3
$0100
ASSEM8LE IN STACK PAGE
REMEMBER THAT THE ROUTINE DESTROYS THE CONTENTS
OF
THE MEMORY
TESTED,.
0100
8GNADL ,
$lJ080
START ADDRESS OF MEMORY TO 8E
TESTED
0100
8GNADH ,
$0081
011.10
ENDADL ,
$lJ082
END ADDRESS &1 OF MEMORY TO 8E
TESTED
",Ul0
ENDADH ,
$IJ",83
//l1",0
COUNTR ,
$0084
COUNTER AND SEED FOR TEST
"'1",0
TMPADL ,
$//l085
WORKING ADDRESS POINTER
"'1",0
TMPADH •
$IJ",86
"'Ief)
TMPY
$lJ087
TEMPORARY
STORAGE 0, Y
KIM ROM ROUTINES USED
"'1",0
CRLF
$IE2F
CARRIAGE RETURN - LINE 'EED
"'1",0
OUTCH
$IEA"
OUTPUT ASCII CHARACTER
"'UI",
PRTBYT ,
$IE38
PRINT I HEX 8YTE AS TWO ASCII
"'1",,,,
OUTSP
$IE9E
OUTPUT BLANK
el"''''
2",
2F
IE
IN IT
JSR
CRLF
PRINT CRILF
"'1",3
A", ",IJ
TEST
LDYIM $08
INITIALIZE INDEX REGISTER
"'11J5
2",
4A
01
JSR
SETEM
CREATE WORKING ADDRESS POINTER
IJHI8 21J
53
"'1
LOOPA
JSR
SUMUM
CALCULATE TEST DATA BYTE
1J11J3 91
85
STAIY
TMPADL
STORE THE TEST BYTE
010D 20 58
IJI
JSR
INCPTR INCREMENT THE WORKING POINTER
", II'" DIJ F6
SNE
LOOPA
MORE TO 8E TESTED?
"'112 20 4A //l1
JSR
SETEM
REINITIALIZE WORKING POINTER
11115 21J 53 IJ I
LOOPS
JSR
SUMUM
RECALCULATE THE TEST DATA BYTE
//l118 51 85
EORIY
TMPADL CHECK MEMORY WITH CALCULATED
TEST BYTE
III IA D0 flF
8NE
ERROR
GO TELL IF TEST FAILED
IJ II C 20 58
01
RETURN
JSR
INCPTR INCREMENT THE WORKING POINTER
011F Dill F4
SNE
LOOP8
MORE TO 8E TESTED?
1J121 A9 28
ONCE
LDAIM
'&
PRINT A "PLUS" TO INDICATE
SUCCESS
0123 21J
A0
IE
JSR
OUTCH
PRINT ASCII
0126 E6 84
INC COUNTR
SET UP NEW PATTERN
0128 4C
1J3 fl1
JMP TEST
TEST UNTIL MANUAL RESET
012B 84 87
ERROR
STY
HIPY
SAVE Y
012D 48
PHA
SAVE THE SAD BIT
PATTERN
",12E 21J
2F IE
JSR
CRLF
PRINT CR/LF
0131 A5
86
LOA TMPADH
GET HIGH ADDRESS
OF ERROR
0133 2//l
38
IE
JSR PRTBYT
PRINT IT
//l136 A5
85
LOA TMPADL
GET LOW ADDRESS
OF ERROR
1J138 21J 38
IE
JSR PRTBYT
PR INT IT
elI3B 2fl
9E
IE
JSR
OUTSP
PRINT A SPACE
//l13E 68
PLA
RESTORE THE 8AD BIT
PATTERN
1J13F 20 38
IE
JSR
PRTBYT
PR INT IT
0142 2//l 2F
IE
JSR
CRLF
PR I NT A CR/LF
0145 A4 87
LDY
TMPY
RESTORE Y
1J147 4C IC
IJI
JMP
RETURN
CONTINUE WITH
THE TEST
SUBROUTINES
1J14A A5 8fl
SETEIII
LDA
BGNADL GET BEGINNING
ADL
1J14C 85
85
STA
TMPADL MAKE A COPY
elI4E
A5
81
LDA
BGNADH GET BEGINNING
ADH
1J151J
85
86
STA
TMPADH MAKE A COPY
l!l152 60
RTS
0153 18
SUIIIUM CLC
GET READY TO ADD
1J154 A5 86
LDA TMPADH GET WORKING POINTER ADH
l!l156 65
85
ADC TMPADL ADD IN WORXING POINTER ADL
0158 65 84
ADC COUNTR ADD IN COUNTER
015A
60
RTS
RETURN WITH CALCULATED TEST DATA BYTE
IN A REGISTER
815S E6 85
INCPTR INC
TMPADL
INCREMENT WORK POINTER ADL
0150
D0
02
BNE CMPADL PAGE NOT CROSSED
elI5F E6 86
INC Tl'IPADH
INCREMENT WORK POINTER ADH
1J161
A5 85
CIIIPADL
LOA
TMPADL GET ADL OF WORK POINTER
0163
C5 82
CMP
ENDADL SEE IF END OF' MEIIIORY TO BE
TESTED
0165 DIJ 114
BNE RTN
RETURN IF NO MATCH
0167 A5 86
LOA
Tl'IPADH GET ADH OF END 0' MEMORY TO BE
TESTED
//l169 C5 83
CMP
ENDADH SEE IF ADH'S MATCH
1l16B 61l
RTN
RTS
RETURN WITH ~ESULTS OF CMP IN Z FLAG
MICRO -- The 6502 Journal
December, 1979