conversion from another flavour of basic

Previous topic - Next topic

Gary

I am looking to write a simple converter from a sound eprom from a fruit machine which is encoded for an oki M6876 chip in 4 bit adpcm format into wav files.

I have found some source code that does the job apparently (funny enough the guy who wrote it bought an unknown board from an electrical shop and was curious to know what was on the sound proms on there. The board is the one I want to do the decoder for, so the code is written already :) ). The problem is I don't know what BASIC it is written in, it wont compile with any of the VB editions I have and thats about the only other one I know.

I'm wondering if any of the more knowledgeable members could have a look and let me know if they recognise what BASIC it is and also any pointers to converting it to GLB would be great. Everything looks fairly straight forward but if there are any particular commands that don't have a GLB equivalent it would be good to know before I start.

Cheers
Gary

Code (glbasic) Select



' ADPCM decoder. this code decompresses the contents of a speech chip sound ROM set that
' is 4 bit ADPCM encoded. A table of addresses at the start of the ROM is used to find
' the start of the individual samples, the end, and the sample length, is calculated from
' the sample block sizes. this version saves the decompressed samples as eight bit 16ksps
' mono .wav files with the hex start address of the sample as the file name

' 05/04/06

SCREEN 12 ' 640 x 480 x 16
CLS ' clear it

DEFINT A-Z ' all integers except where defined

DIM stepadj(7) ' step index adjust, 0 to 7
DIM steptable(48) ' step size table, 0 to 48
DIM wavaddr&(127) ' sample start addresses
adrcount% = 0 ' number of sample addresses found

RESTORE stepadj ' read the step index adjust table
FOR i = 0 TO 7
  READ stepadj(i)
NEXT

RESTORE steptable ' read the step size table
FOR i = 0 TO 48
  READ steptable(i)
NEXT

OPEN "snd_2.rom" FOR BINARY AS #1 ' open input file

i& = 3 ' skip unknown start bytes

DO
  GOSUB GetLong ' get address longword
  wavaddr&(adrcount%) = llong& ' save address
  adrcount% = adrcount% + 1 ' increment address count
  PRINT RIGHT$("00000" + HEX$(llong&), 6); "  "; ' display address
LOOP WHILE llong&

PRINT
PRINT  adrcount% - 1; "addresses found";
adrcount% = adrcount% - 2 ' correct for zero term and zero
' count start

FOR wavfiles% = 0 to adrcount% ' for each sample
  filename$ = RIGHT$("00000" + HEX$(wavaddr&(wavfiles%)), 6) + ".wav"
  OPEN filename$ FOR OUTPUT AS #2 ' open output file
  PRINT
  PRINT "File "; filename$; ' show file name
  GOSUB GetLength ' get the sample length
  GOSUB WriteRIFF ' write the header chunk
  GOSUB WriteFORMAT ' write the format chunk
  GOSUB WriteDATA ' write the data chunk
  CLOSE #2 ' close output file
NEXT

CLOSE #1 ' close input file

PRINT
PRINT "Done. Any key to exit." ' show file name

DO
LOOP WHILE INKEY$="" ' wait for a key

END


' get the sample size. the sample is arranged into blocks, the first byte of the block
' is the byte count for the remainder of the block + $80. the sample is terminated by
' a block count of zero

GetLength:
  sslength& = 0 ' reset length
  i& = wavaddr&(wavfiles%) ' sample start address
  DO
    GOSUB GetByte ' read block length
    byte% = byte% AND 127 ' mask block length
    sslength& = sslength& + byte% ' add this chunk length
    i& = i& + byte% ' index to next chunk length
  LOOP WHILE byte% ' zero length is end marker
RETURN


' write RIFF header chunk

WriteRIFF:
  PRINT #2, "RIFF"; ' always there
  samplength& = sslength& * 2 ' length of output in bytes
  llong& = samplength& + 36 ' length of rest of file
  GOSUB WriteLongword ' write file length
  PRINT #2, "WAVE"; ' always there
RETURN


' write FORMAT chunk

WriteFORMAT:
  PRINT #2, "fmt "; ' always there
  llong& = 16 : GOSUB WriteLongword ' length of chunk, always 16
  llong& = 1 : GOSUB WriteWord ' always 1
  llong& = 1 : GOSUB WriteWord ' mono
  llong& = 16000 : GOSUB WriteLongword ' sample rate
  llong& = 16000 : GOSUB WriteLongword ' bytes/second
  llong& = 1 : GOSUB WriteWord ' 1 byte/sample
  llong& = 8 : GOSUB WriteWord ' 8 bits/sample
RETURN


' write DATA chunk

WriteDATA:
  PRINT #2, "data"; ' always there
  llong& = samplength& : GOSUB WriteLongword ' data chunk length
  sstart& = wavaddr&(wavfiles%) ' sample start address
  send& = wavaddr&(wavfiles% + 1) ' sample end address
  GOSUB Decompress ' decompress the sample
RETURN


' write llong& as word to file, little endian format

WriteWord:
  PRINT #2, CHR$(llong& AND 255&); ' write low byte
  llong& = llong& \ 256&
  PRINT #2, CHR$(llong& AND 255&); ' write high byte
RETURN


' write llong& as longword to file, little endian format

WriteLongword:
  GOSUB WriteWord ' write low word
  llong& = llong& \ 256&
  GOSUB WriteWord ' write high word
RETURN


' decompress ADPCM sample data to .wav data. first get the blocksize, if it's zero then
' all done so exit else for each byte in the block umpack the high then the low nibbles
' into 12 bit samples

' This routine could be greatly simplified but is given in this form as it more closely
' represents how it would be implemented in assembly code or directly in hardware

Decompress:
  stepindex = 0 ' reset decompression variables
  stepsize = 16 ' initialise index
  sample& = 2048 ' set 50% for unsigned value

  i& = sstart& ' set start of sample
  eflag% = 0 ' clear end flag
  DO
    GOSUB GetByte ' get this sample blocklength
    blklength% = byte% AND 127 ' mask and copy block length
    IF blklength% THEN ' if there are samples
      DO
        GOSUB GetByte ' get two compressed nibbles
        FOR nib% = 0 TO 1 ' for each nibble
          IF nib% THEN
            nibble% = byte% AND 15 ' second sample is low nibble
          ELSE
            nibble% = byte% \ 16 ' first sample is high nibble
          END IF

          newsample = nibble% AND 7 ' compressed sample magnitude

          difference = stepsize \ 8 ' effectively add 0.5 to difference

          IF newsample AND 4 THEN
            difference = difference + stepsize ' + stepsize / 4 * 4
          END IF
          IF newsample AND 2 THEN
            difference = difference + stepsize \ 2 ' + stepsize / 4 * 2
          END IF
          IF newsample AND 1 THEN
            difference = difference + stepsize \ 4 ' + stepsize / 4 * 1
          END IF

          IF (nibble% AND 8) THEN ' if -ve compressed sample
            sample& = sample& - difference '  subtract the difference
          ELSE ' else compressed sample was +ve
            sample& = sample& + difference '  so add the difference
          END IF

          IF sample& > 4095 THEN ' if > than unsigned 12 bit max
            PRINT " +clip "; '  print a warning
            sample& = 4095 '  clamp to 4095
          ELSEIF sample& < 0 THEN ' if < than unsigned 12 bit min
            PRINT " -clip "; '  print a warning
            sample& = 0 '  clamp to 0
          END IF

          PRINT #2, CHR$(sample& \ 16&); ' scale and save decompressed sample
          stepindex = stepindex + stepadj(newsample) ' new decompress step index
          IF stepindex > 48 THEN ' if > than step table max
            stepindex = 48 '  clamp to step table max
          ELSEIF stepindex < 0 THEN ' if < than step table min
            stepindex = 0 '  clamp to step table min
          END IF

          stepsize = steptable(stepindex) ' get step size for next sample

        NEXT nib% ' do other nibble
        blklength% = blklength% - 1% ' decrement count
      LOOP WHILE blklength% ' loop if more to do
    ELSE
      eflag% = -1% ' else set end flag
    END IF
  LOOP UNTIL eflag% ' loop until end
RETURN


' get an indexed byte from the input stream. return the value as a string and a number

GetByte:
  byte$ = " " ' one byte long string
  i& = i& + 1& ' increment the index
  GET #1, (i&), byte$ ' get LEN(byte$) bytes
  byte% =  ASC(byte$) ' get numeric byte value
RETURN


' get a big endian longword from the input stream. return the value in llong&

GetLong:
  GOSUB GetByte ' get MS byte
  llong& = byte% ' copy it
  GOSUB GetByte ' get next byte
  llong& = llong& * 256& + byte% ' convert and add it
  GOSUB GetByte ' get next byte
  llong& = llong& * 256& + byte% ' convert and add it
  GOSUB GetByte ' get next byte
  llong& = llong& * 256& + byte% ' convert and add it
RETURN


' dialogic modified IMA tables

steptable:
DATA 16,  17,  19,  21,   23,   25,   28,   31,   34,  37
DATA 41,  45,  50,  55,   60,   66,   73,   80,   88,  97
DATA 107, 118, 130, 143,  157,  173,  190,  209,  230, 253
DATA 279, 307, 337, 371,  408,  449,  494,  544,  598, 658
DATA 724, 796, 876, 963, 1060, 1166, 1282, 1411, 1552

stepadj:
DATA -1,-1,-1,-1,2,4,6,8


I have attached the source file as well as a sound prom file to decode

[attachment deleted by admin]

okee

Gary, It seems to run in QBasic, I get a message

000000
0 Addresses found
Done Any key to Exit

I happened to have a copy of QB64 on my computer.
http://www.qb64.net/
Android: Samsung Galaxy S2 -  ZTE Blade (Orange San Francisco) - Ainol Novo 7 Aurora 2
IOS: 2 x Ipod Touch (1G)

Gary

Thanks for the quick reply Okee, I can at least run in that and see how good the conversion is then compare the output to the GLB version when done

Gary

btw is you put snd_2.rom in your mail QB64 folder, or move the exe file generated to the same folder as the rom it will decrypt it

Gary

apologies for the questions but do you know if a variable has an & at the end of it does it signify anything special? Im just removing them for now

okee

I think it signifies a Long Integer, so you should be ok
Android: Samsung Galaxy S2 -  ZTE Blade (Orange San Francisco) - Ainol Novo 7 Aurora 2
IOS: 2 x Ipod Touch (1G)

Gary

got the majority converted to GLB but couple of sticking points

this section wont compile

Code (glbasic) Select

//' get a big endian longword from the INPUT stream. RETURN the value IN llong&

SUB GetLong:
READLONG 0,llong
ENDSUB



it gives the error

Code (glbasic) Select

C:\Users\Gary\AppData\Local\Temp\glbasic\gpc_temp0.cpp: In function `DGInt __GLBASIC__::GetLong()':
C:\Users\Gary\AppData\Local\Temp\glbasic\gpc_temp0.cpp:1194: error: invalid initialization of reference of type 'DGNat&' from expression of type 'DGInt'
C:/Program Files (x86)/GLBasic/Compiler/platform/Include/glb.h:1124: error: in passing argument 2 of `void __GLBASIC__::READLONG(DGNat, DGNat&)'


and is there any way of jumping bytes in  file so you can go straight to the 4th byte in the file and skip reading the first 3? the only way I can see to do it is to load the whole file into an array and interrogate it from the array but that seems long winded

MrTAToad

The second parameter for READLONG must be an integer, and not a floating-point variable.

Gary

For anyone that is interested I have attached the converted program. The outputted files are not totally identical to the QBASIC output (seems to go slightly off track half way through decoding a sample) but the outputted wav file sounds exactly the same.

Its not pretty and is slower than the QB original (partly because I have to load the rom into an array before I start to do the conversion and partly because I have to check the rom size first so its loaded twice) but it does work :)

Thanks to everyone for the help. And for the that are curious, the sound roms that can be converted are from Barcrest MPU4 fruit machines

Gary


[attachment deleted by admin]