Project Logs

Collapse

• Complete BASIC program

  danjovic • 12/20/2016 at 00:43 • 0 comments

  I have wrote a complete BASIC program for loading the hook code using the MSX and used "LLIST" command to transfer the program to the PC using the hook code itself.

  This version was named "complete" because it has instructions on how to change the baudrate and how to use parity

  10 REM
  20 REM Joy232 hook code
  30 REM danjovic 2016
  40 REM version 1.31 19/12/2016
  50 REM
  60 EI=&HFAF5:SIZE=63
  70 FOR A = EI TO EI+SIZE
  80 READ B$: POKE A,VAL("&H"+B$)
  90 NEXT A
  100 REM
  110 POKE &HFFB8,&HFA
  120 POKE &HFFB7,&HF5
  130 POKE &HFFB6,&HC3
  140 REM
  150 DATA F3,F5,C5,E5,6F,3E,0F,D3
  160 DATA A0,26,FF,A7,CB,15,CB,14
  170 DATA 06,0B,DB,A2,CB,87,CB,1C
  180 DATA CB,1D,CE,00,D3,A1,0E,11
  190 DATA 0D,20,FD,10,ED,E1,C1,F1
  200 DATA A7,FB,33,33,C9,CE,00,D3
  210 DATA A1,0E,06,0D,20,FD,10,ED
  220 DATA E1,C1,F1,FB,33,33,C9,00
  230 REM
  240 REM To change baudrate
  250 REM
  260 REM POKE &HFB27,value
  270 REM
  280 REM Baudrate  value
  290 REM 1200      170
  300 REM 2400      83
  310 REM 4800      39
  320 REM 9600      17
  330 REM 14400     10
  340 REM 19200      6
  350 REM
  360 REM ----------------------
  370 REM To use parity
  380 REM
  390 REM POKE &HF96D,value
  400 REM
  410 REM Parity    value
  420 REM NONE        0 (or 1)
  430 REM EVEN        2
  440 REM ODD         3

  
  

• Testing the baudrates

  danjovic • 12/19/2016 at 23:55 • 0 comments

  I have used a BASIC program to test the baudrates.

  10 REM test transmission baudrates
  20 BDR = &HFB27
  30 REM Baudrate Value
  40 REM 1200     170
  50 REM 2400     83
  60 REM 4800     39
  70 REM 9600     17
  80 REM 14400    10
  90 REM 19200     6
  100 REM
  110 POKE(BDR),17
  120 LPRINT "THE QUICK BROWN FOX JUMPS OVER THE LAZY DOG 01234567"
  130 LPRINT "the quick brown fox jumps over the lazy dog !@#+-()."
  140 GOTO 120

  The code worked in all speeds with the given values for the delay loop. but I have tested some values around and verified that above 9600 bauds such value is very critical.

  
  

• New fluxogram

  danjovic • 12/19/2016 at 23:07 • 0 comments

  New fluxogram to reflect the last version of the code. The colors from latter fluxogram have been kept to ease the comparison.

• In Soviet Russia.... BIOS uses your software.

  danjovic • 12/19/2016 at 22:39 • 0 comments

  This post is to remember the contest judges that my code does not use the BIOS from MSX computers to print the character, but the opposite! BIOS use my software.

  I may use BIOS to load the code to memory but this is the same circumstance of the Arduino or the Parallax bootloader :)

  Spasiba!
  

• Added parity control

  danjovic • 12/19/2016 at 22:19 • 0 comments

  Added configurable parity control. Parity can be NONE, EVEN or ODD, selectable by contents of address 0F96DH which RS232 Queue Putback flag that is not used by MSX BIOS.
  The hook code size is 63 bytes and can still fit within unused RS232 Queue space (64 bytes). The entire binary takes 93 bytes.
  

  Parity was checked using oscilloscope and a BASIC program .

  10 REM test transmission with parity
  20 CNF = &HF96D
  30 REM
  40 REM 0/1: no parity   (N)
  50 REM   2: even parity (E)
  60 REM   3: odd parity  (O)
  70 REM
  80 POKE(CNF),2
  90 LPRINT "A";
  100 GOTO 90

• Code profile

  danjovic • 12/13/2016 at 04:44 • 0 comments

  The enhanced code takes only 75 bytes, from which 7 are the binary header file, 23 are for transfering the hook code to the unused RS232 queue. The hook code itself takes only 44 bytes.

• Further enhancement

  danjovic • 12/12/2016 at 20:37 • 0 comments

  After optimization the size of the loadable binary file shrinked to only 65 bytes being 7 for the file header, 14 for hook code installer and 44 for the hook code itself which is small enough to fit whithin the unused 64 bytes reserved for the rs232 queue in 'system variables' area.

  ;
  ; Install bitbang routine on printer HOOK
  ; Borrowed from the book "+50 dicas para o MSX"
  ;
  HLPT:  EQU 0FFB6H ; Printer hook entry
  RS2IQ: EQU 0FAF5H ; RS232 queue, 64 bytes
  
  ;
  INSTALL:
  LD HL, HOOK_PRNTJ232           ; beginning of hook code
  LD DE, RS2IQ                   ; destiny, unused rs232 queue
  LD BC, HOOK_END-HOOK_PRNTJ232  ; block size
  LDIR                           ; transfer hook code to its new location
  
  
  LD HL,RS2IQ     ; Write the execution entry point for printer hook
  LD [HLPT+1],HL
  LD A,0C3H       ; then write the CALL instruction (0xC3)
  LD [HLPT],A
  RET

  
  

  The downside is that now we need 10 extra bytes of instructions to move the hook code from the load address to the rs232 queue area. Another necessary modification was on the bit delay loop which used an absolute address jump to simplify the timing (JP NZ,xxxx takes 11 cycles either jump or no jump) . It was replaced by a relative jump (JR NZ,xxxx) .

     ; delay 1 bit time
     LD C,T9600     ;   8     8      poke here to change the baud rate
     DELAY_C:
     DEC C          ; ( 5)
     JR NZ,DELAY_C  ; (12)(/7)
                    ;  (12+5)*(TDelay-1)+(7+5) = 17*TDelay - 17 + 12 = (17*Tdelay-5)
  
  DJNZ SEND_BITS    ; 14(/9)         send next bit

  
  

  The relative jump made the hook code relocatable (now it can be copied and executed from wherever address) and the new timing due to the relative jump ended by enhancing the error rate for 14400 and 19200 baud are now both under 2% of error (it was not intentional though).
  

  Clock = 3.575611 MHz
                   
  Baud   Tdelay  Cycles RealRate  Error %
  1200    170    2971   1203,50    0,29
  2400     83    1492   2396,52   -0,14
  4800     39     744   4805,92    0,12
  9600     17     370   9663,81    0,66
  14400    10     251   14245,4   -1,07
  19200     6     183   19538,8    1,76
  

  
  

  I could have loaded the hook code directly in the rs232 queue along with the installation code (14+44 bytes) , but I am planning to use the remaining 20 bytes to allow extra stuff like parity.

• Squeezing the code because 85 bytes is too many...

  danjovic • 12/10/2016 at 14:17 • 0 comments

  I have started to squeeze the hook code. This can be done by merging the start, stop and data bits into a 16 bit register and use the same loop to transmit all the bits from the RS232 character (i.e. 10 bits). It also has a positive side effect that now all the bits will last exactly the same time (the stop bit may linger but it doesn't really matter as long as it takes at least 1 bit time.

  Then the bitbang code turns into:

  SND232:
  ; Inputs:
  ; L: Byte to be sent  
  ; 
  ; Changes: AF, BC, HL
  
  
  ; Select PSG Register 15 
  LD A,15     
  OUT [PSGAD],A  
  
  ; Prepare stream of bits
  LD H,255; Add stop bits
  AND A   ; clear carry flag
  RL L    ; Add start bit and send 7th bit to carry
  RL H    ; Send 7th bit to H register
          ; HL register is now   
          ; ------------ H -----------    ----------- L ---------
          ; 16 15 14 13 12 11 10  9  8    7  6  5  4  3  2  1  0
          ;  1  1  1  1  1  1  1 stp b7   b6 b5 b4 b3 b2 b1 b0 start
       
  ; Now loop through 10 bits [1 extra to compensate time at the end of the loop]
  LD B,10+1  ; poke here with 12 to send two stop bits instead of only one 
  
  SEND_BITS:
     ; prepare mask ; Cycles  Accumul
     IN A,[PSGRD]   ;  14     24      save the present state of the bits from PSG register 15
     RES 0,A        ;  10             clear bit 0
  
     ; 16 bit rotate
     RR H           ;  10     20      H.0->Cy (bit 0 from H goes to carry) 
     RR L           ;  10             Cy->L.7, L.0->Cy    
  
     ; set TXD line state
     ADC A,0        ;   8     20      A.0 now equals Cy
     OUT [PSGWR],A  ;  12             write bit to output. Here starts the cycle counting 	
  
     ; delay 1 bit time
     LD C,_DLY      ;   8     8      poke here to change the cycle count	
     DELAY_C:
     DEC C          ; ( 5)
     JP NZ DELAY_C  ; (11)
                    ; 16*_DLY     
  
  DJNZ SEND_BITS    ; 14(/9)         send next bit
  
                    ;Total:  8 + 16*_DLY + 24 + 20 + 20 from Start bit to beginning of stop bit
                    ;        72 +16*_DLY cycles
                    ;
                    ; after the last bit we have:  8 + 16*_DLY + 9 
                    ;                              17 + 16*_DLY
                    ;
                    ; we're missing 72-17 = 55 cycles that's why we added 1 to ensure that
                    ; at least 1 stop bit time has passed before we return
  
  RET

  
  

• Test of cleanup code

  danjovic • 12/09/2016 at 19:50 • 0 comments

  The USB to serial adapters are now more common than computers with real serial ports and it makes the interface between the MSX and the PC far easier. It takes only the USB-Serial adapter and the DB-9 female connector.

  The code have been cleaned up to remove some useless instructions and now takes 85 bytes including the 7 header bytes for binary type files, but it can be loaded with a BASIC program as well.

  The video below shows the code running after being typed from BASIC prompt.

  The wafeforms below shows the character "A" (0x41) being transmitted at 9600 baud.

  The next waveform shows the character "U" (0x55) being transmitted.

  The second video shows the code being loaded from a binary file. An application test was written to read the memory locations where the code sits in and print them as hexadecimal values.

• Cleanup and repository

  danjovic • 11/26/2016 at 15:12 • 0 comments

  I have made a cleanup in the code, eliminating the selection of the multiplex for PORTB which was useless for only writing to the joystick ports.

  Restructured (slightly) the code by moving the instructions that select PSG register to the subroutine where they are used. Now the structure of the program follows the diagram below.

View all 11 project logs