After RESET the baudrate is determined by measuring the length of the start bit of an incoming serial character. This means any character is usable where the first data bit is the opposite of the start bit. The KIM-1 User manual suggest RUBOUT ($7F 1111111) but ENTER ($0D 0000 1101) also works fine. SPACE ($20 0010 0000 for example does not work!

0612   1C2A A9 FF               LDA   #$FF       ; COUNT START BIT
0613   1C2C 8D F3 17            STA   CNTH30     ; ZERO CNTH30
0614   1C2F A9 01               LDA   #$01       ; MASK HI ORDER BITS
0615   1C31 2C 40 17    DET1    BIT   SAD        ; TEST 
0616   1C34 D0 19               BNE   START      ; KEYBD SSW TEST
0617   1C36 30 F9               BMI   DET1       ; START BIT TEST
0618   1C38 A9 FC               LDA   #$FC
0619   1C3A 18          DET3    CLC              ; THIS LOOP COUNTS 
0620   1C3B 69 01               ADC   #$01       ; THE START BIT TIME
0621   1C3D 90 03               BCC   DET2
0622   1C3F EE F3 17            INC   CNTH30
0623   1C42 AC 40 17    DET2    LDY   SAD        ; CHECK FOR END OF START BIT 
0624   1C45 10 F3               BPL   DET3
0625   1C47 8D F2 17            STA   CNTL30
0626   1C4A A2 08               LDX   #$08
0627   1C4C 20 6A 1E            JSR   GET5       ; GET REST OF THE CHAR, 
0628   1C4F             		         ; TEST CHAR HERE

What happens here:

– bit 7 (PB7) is tested until it becomes 0 (BIT SAD and BMI DET1 loop)
– the time is counted and kept in CNTH30 and CNTL30
– bit 7 is tested for becoming 1 (LDY SAD and BPL DET3)
– the rest of the character is read in by jumping into GETCH , the actual character received is not tested.

1006   1ED4             ;		
1007   1ED4             ;       DELAY 1 BIT TIME   
1008   1ED4             ;       AS DETERMINED BY DETCPS
1009   1ED4             ;
1010   1ED4 AD F3 17    DELAY   LDA   CNTH30     ; THIS LOOP SIMULATES 
1011   1ED7 8D F4 17            STA   TIMH       ; DETCPS SECTIONS AND WILL DELAY
1012   1EDA AD F2 17            LDA   CNTL30     ; 1 BIT TIME
1013   1EDD 38          DE2     SEC   
1014   1EDE E9 01       DE4     SBC   #$01
1015   1EE0 B0 03               BCS   DE3
1016   1EE2 CE F4 17            DEC   TIMH
1017   1EE5 AC F4 17    DE3     LDY   TIMH
1018   1EE8 10 F3               BPL   DE2
1019   1EEA 60                  RTS
1020   1EEB             ;		
1021   1EEB             ;                          DELAY 1/2 BIT TIME   
1022   1EEB AD F3 17    DEHALF  LDA   CNTH30     ; DOUBLE RIGHT SHIFT OF DELAY                              
1023   1EEE 8D F4 17            STA   TIMH       ; CONSTANT FOR A DIVE 2 
1024   1EF1 AD F2 17            LDA   CNTL30
1025   1EF4 4A                  LSR   A
1026   1EF5 4E F4 17            LSR   TIMH
1027   1EF8 90 E3               BCC   DE2
1028   1EFA 09 80               ORA   #$80
1029   1EFC B0 E0               BCS   DE4

The actual delay routines use the same logic as DETCPS. IN DEHALF the delay time is divided by 2 and jumped into DELAY.

KIM usernotes vol 06

Jim mcClahanan notes:

The PAL-1 (just like the KIM-1) uses a ‘soft UART’ or ‘bit banger’ for its serial I/O. I’m not a fan of this approach, but at the same time it demonstrates what could be accomplished with a minimial amount of hardware. The PAL-1 automatically figures out the appropriate delay between bits of the serial character when you press enter after a reboot. I have found that decreasing the value actually significantly improves the odds of an error-free load of larger punchtape format files. Below is a table for values found and suggest for $17F2.

Baud Found New
300 $EA $E8
1,200 $37 $35
2,400 $1A $18
4,800 $0B $0A
I haven’t tried to optimize the delay values. Right now I’m using 5 ms between characters and 500 ms between lines when doing 8K transfers and with the modified values I usually am successful. With the default values, it seemed like even with longer delays things would slip out of synchronization at some point more often than not on large transfers.