Note that some of this information is incomplete. Check the Grasp manual for clarification on e.g. the script-file commands. If anybody would like to merge all relevant documents together, that would be nice. These documents were passed to me by Martin Fong, fong@erg.sri.com Eli Brandt eli@smectos.gang.umass.edu 32@4351 WWIV ======================================================================== The formats of GRASP animation files. By George Phillips Distribute this freely, but give credit where credit is due, eh? Version: Jan. 19,1991 GRASP is an animation system particular to the IBM PC world. It consists of a program to create animations and a run-time environment for displaying them. The most common form these animations take is ".GL" archives which may be displayed on an IBM-PC with a program called GRASPRT.EXE. This document describes what I have been able to decipher about the format of ".GL" archives and the files contained within. It should be useful to those attempting to write ".GL" animation players on other platforms. A ".GL" file is simply an archive file which contains images, fonts and a command file which tells GRASPRT what to do. These various files have standard extensions to denote their contents: ..txt - A command file; usually there is only one of these per archive. ..pic - An image. ..clp - An image but without a colour map. ..set or .fnt - A font containing character glyphs. It should be noted that the GL archive is of no particular importance; all the archived files could exist as ordinary files and the animation should still work. Any GL player should be able to operate both from an archive or from ordinary files. File Formats Most of the data in GL files can be adequately described as a stream of bytes which is practically universally understood. Some fields contain 2-byte and 4-byte integers. I'll refer to these as "words" and "long words" and they are all stored in little-endian format. So if we have 4 consecutive bytes, b1, b2, b3 and b4, the word at b1 is (b1 + b2 * 256) and the long word at b1 is (b1 + b2 * 256 + b3 * 256 * 256 + b4 * 256 * 256 * 256). Since this information was gathered by example, the purpose of some header fields and commands may not be known. I've marked unknown fields with question marks and have tried to put question marks and other warnings about descriptions which are guesses. GL Archives (.gl) A GL archive begins with a directory listing the files in the archive which is followed by the data for each file. +-- Directory Header | dir length (word) number of bytes in the directory header | +-- File Entry (17 bytes per, (dir length) / 17 of them) | | offset (long word) Position of file data as an offset from | | the beginning of the archive | | name (13 bytes) File name, null padded. | +-- +--- File data area | +-- File Data | | length (long word) Size of the file | | data (bytes) the file's data (surprise!) | +-- +--- Font Files (.fnt or .set) These are very simple; first a short header describing the size of the characters in the font and what byte values correspond to each glyph followed by the glyph data. +-- Font Header | length (word) length of the entire font file | size (byte) number of glyphs in the font file | first (byte) byte value represented by the first glyph | width (byte) width of each glyph in pixels | height (byte) height of each glyph in pixels | glyphsize (byte) number of bytes to encode each glyph +-- Glyph Data | glyph first | glyph first + 1 | ... | glyph first + size - 2 | glyph first + size - 1 +-- Each glyph is stored almost exactly as you would expect a raw PBM file to contain it except that a '0' bit means black and a '1' bit means white. In other words, row major order, each line padded to end on a byte boundary, most significant bit is leftmost. Image Formats (.pic and .clp) These consist of a header containing the usual image information followed by blocked, run-length encoded image data. +-- Image Header (17 or 19 bytes) | magic? (byte) magic number? Always is 0x34 or 0x12 | width (word) width of image in pixels | height (word) heigh of image in pixels | ???? (4 bytes) unknown | bpp (byte) bits per pixel (only seen 1 or 8) | type (byte) image type, either 'L' or 'C' | flags (byte) if (flags & 4) then image has colourmap | ? (byte) unknown | extend (byte) extended header byte (if != 0, header | has 2 more bytes) 1/2? | ? (byte) unknown | ?? (2 bytes) header extension if extend != 0 +-- Colour Map ((1 << bpp) * 3 bytes, only if flags & 4 == 4) | +-- Colour Map entries (as many as indicated by bpp) | | R (byte) red intensity, 0 - 63 \ | | G (byte) green intensity, 0 - 63 + entry 0 | | B (byte) blue intensity, 0 - 63 / | +-- | ... +-- Image Data | blocks (word) number of blocks of data | +-- Data Block (blocks of them) | | length (word) length of data block, including header | | bufsize (word) buffer size needed to hold all the | | uncompressed data in this block | | esc (byte) the escape code in this block | | data (length - 5 byte) run-length encoded data | +-- +-- The run-length encoding is byte oriented and follows these rules: - characters other than "esc" (see data block header) are literal - esc n c means repeat c n times (1 <= n <= 255) - esc 0 len(word) c means repeat c len times If bpp=1, then the resulting data stream is interpreted as it is with font glyphs (i.e., msb is left, pad to bytes, row first, etc). If bpp=8, then each byte in the data stream is an index into the colour map. If no colour map is available, the map to use can only be discovered by running through the command file. I've only seen images with bpp=1 and bpp=8 and they it always works out that either bpp=1 and type=C or bpp=8 and type=L. The type=C corresponds to CGA graphics which are mostly monochrome and 640 x 200 (so the aspect ratio is funny). Type=L is colour graphics, prob. VGA and usually 320 x 200. Notice that the colour maps have only 6 bits, the same as VGA's digital to analog converters. ".pic" files always have colour maps, ".clp" files never do. It seems that you can be lazy with your run-length decoding code; I've never seen a full sequence appear across a data-block boundary (encoders should probably not let that happen). The amount of uncompressed data in a block never seems to exceed 8192 bytes. Much of the header information is mysterious. Note that the header extension field is a guess and that there are other consistent possibilities (e.g., the extension field is a length byte or even part of a length word). Only type=C images seem to have the extension. Maybe the extra information is supposed to be used in video mode operating system calls on the PC? What made this part easier was the existence of a PC-based program which converts ".pic" files into GIF files. Its called "cvt2gif" and can be found on wuarchive.wustl.edu:/mirrors/msdos/gif/cvt2gif.zip. Those wishing to enhance the format descriptions would do well to get a copy. I did notice that bpp=1 images are not necessarily black and white but could be black and some other colour as selected from the CGA pallette. I doubt the distinction will make much difference to the animation, but if you really want to do it right... Command File (.txt) The command file looks like a typical script file with the lines delimited by carriage returns, line feeds or both. Any text following ';' on a line is a comment. Text followed by a colon is used to indicate a label (much like most assemblers). Commands consist of a keyword followed by a list of comma separated arguments. The input is case-insensitive except for arguments containing text to display (which are in double quotes). The basis of the command language seems to be what I call picture and clip registers, of which there are 16 of each. A few commands will load a picture (or clip) from a file into a register. Other commands then reference the register numbers to display the pictures or get colour maps from them. It seems that the colour map from a picture (.pic) is installed into the hardware and this is where the colour maps for the clips (.clp) come from. I assume that I am missing a lot of commands, but most notably I believe there should be more primitive drawing commands. Many of the commands seem to have a delay argument associated with them. This seems reasonable as control over time in an animation is important. I may have been over-zealous in looking for delays. The actual time units of the delays is unknown. They are typically numbers < 100, so milliseconds are a likely candidate. Hundredths of a second are possible as well. Here is a list of commands. Optional arguments are enclosed in []. Ranges are possible in arguments (I've only seem them in fly) and take the form "n,-,m", (e.g., fly 0,0,10,10,1,1,1,-,16). * box x1,y1,x2,y2,colour? Draw a box with corners (x1, y1) and (x2, y2) in the colour given by the colourmap entry number. * cfade x,y,delay,img,[,?,?] Display a clip image img at (x, y) and wait for delay time units before proceeding. * cfree n Free up any memory associated with clip register n. * clearscr Clear the display (to the currently selected colour or black?). * cload name,num[,?] Load a clip image "name" into clip register num. If name does not have a .clp extension, it will be automatically appended. * color n Set the current colour to n. This at least seems to affect the text displaying commands. * exit Terminate the command file. * fload name Load the named font which becomes the font to be used when displaying text. ".fnt" is appended to name if necessary. * float x1,y1,x2,y2,step?,delay?,num Move the clip image (num) by displaying it at (x1,y1) and erasing it and displaying it every step pixels until (x2,y2). Delay delay time units in between steps. Or maybe something completely different, but the x1,y1,x2,y2 and num arguments are probably coordinates and a clip number. * fly x1,y1,x2,y2,step?,delay?,clip list Successively display the clip images from (x1,y1) to (x2,y2) with delay time units in-between. The clip list is just a bunch of clip numbers separated by commas (i.e., fly is varags). A range is likely to appear in the clip list. Often (x1,y1) == (x2,y2). * fstyle ?[,?] Presumably set up some parameters on how a font is displayed. * goto label Force flow of control to the given label. * loop Denotes the end of a mark loop. Continues the loop at the most recent mark if the loop hasn't finished. * mark n This pairs with the loop command and begins a for loop from 1 to n. One assumes that the interaction of mark, loop and goto is the same as for, next and goto in BASIC. That is, loops are dynamically scoped and you can jump in and out of them. Mark simply pushes a loop start onto the stack and loop examines whatever is on the top of the loop stack. * mode ? Modify the current video mode in some way. I haven't seen this often. * note freq,delay?,duration Play a musical note of the given frequency and duration and delay for delay time units afterward. * pallette n Make the colour map from picture register n be the one to use. This probably installs it into the hardware so that when a clip is loaded there is no colour map to change. * pfade effect,pict[,delay?[,?,?]] Display the picture numbered pict on the screen. The effect number indicates what sort of special effect is used to display it. What the numbers mean I have no idea, but I know some of the effects. Each pixel loaded randomly, every even line then every odd line and so on. The delay parameter seems to make sense, but not always. The extra parameters could be those needed for some effects. Often they are large numbers. * pfree n Free up any memory associated with picture register n. * pload name,n Load picture "name" into picture register n. ".pic" is appended to name if necessary. * putup x,y,n Display clip register n at (x,y). * set retrace [on|off] Set is probably a general internal control variable changing command. What retrace is I have no idea, but it was set off then on around a fly statement. * spread ?,? Who knows, but the numbers used are probably picture register numbers. Maybe some kind of colourmap changing? * text x,y,"text",[delay?] Display the given text (enclosed in double quotes) at (x,y). The extra parameter is probably a display, but it could be the display colour or the background colour. Probably the display colour is that given by the color statement. * tran [on 0|off] No idea. Was used around some cload and float statements. * video mode Set the display mode to 'C' or 'L' (remember the image format types?). Usually the first statement in a command file. C almost certainly refers to CGA which is 640 x 200 monochrome and L almost certainly to VGA which (in their case) is 320 x 200 x 256. * waitkey [[delay[,label]] Wait up to delay units for the user to press a key (or forever if no delay time is given). If the user presses a key and the label argument is present, transfer control to that label. * window x1,y1,x2,y2,? Some kind of display control. Probably a clipping window with appropriate coordinate translation (i.e., (0,0) becomes (x1,y1)). This document was created by looking hard at a number of GL files, using cvt2gif to help decipher the image file format and looking at 1 or 2 animations on an RS-6000 running a PC emulator and using grasprt. cvt2gif was very useful; grasprt under the PC emulator was painfully slow at times and didn't help my understanding much. I've never even gotten close to a copy of the program for creating and editing GL files. If you find out more about GL files, send me the changes so I can extend this document. Feel free to include this as supplementary documentation if you write a GL player. Finally, here are some projects which could help find out more about GL files: - Get cvt2gif and feed it small variations on .pic files to decipher the meaning of the missing header fields. I may do this. - Alter control files on some animations and see what effects they have. Something easy would be to change the effect number on pfade statements (if that's what it is). I don't have the hardware to do this. - Look at the GRASP animation package and intuit what the commands mean by what control you have over generating animations. This is probably the easiest way to get information. I don't have GRASP, I don't know where to get it and I don't has a PC good enough to run it on. ======================================================================== GRASP/Pictor Font format description 09/06/87 ------------------------------------ -------- For convenience, we have chosen to adopt the IBM ROM font format for data, but to keep things manageable, we have added a 7 byte header which describes the font. The seven byte header is defined as follows: WORD number of bytes in character data, plus this 7 byte header. BYTE number of characters in set. 1-255 or 0 if 256. BYTE ascii value of first character. BYTE x size of character in pixels. BYTE y size of character in pixels. BYTE number of bytes in a single character. As you can see from this header data, these limits apply: 1) Maximum number of characters in set is 256. 2) Maximum character size is limited as: xsize/8 * ysize <256. 3) All character data plus 7 byte header must be <64K in size We use the following structure when writing programs that use fonts. Note the additional words at the end of the structure which allow you to keep the actual character data in a far segment. struct chs { /* character set structure */ unsigned int numchbyts; unsigned char numchars; unsigned char ascoff; unsigned char chxsize; unsigned char chysize; unsigned char chbytes; unsigned int chsseg; /* segment of character data */ unsigned int chsofs; /* offset in segment of character data */ }; So....A 256 character 8x16 font's header would look like: numchbyts = 4103 256 chars X 16 bytes/char + 7 bytes for header numchars = 0 0 to represent 256 ascoff = 0 start with 0 character chxsize = 8 8 dots wide chysize = 16 16 dots high chbytes = 16 1 byte wide x 16 dots high and a 96 character 11 X 18 font whose first character is SPACE's header would look like: numchbyts = 3456 96 chars X 36 bytes/char + 7 bytes for header numchars = 0 0 to represent 256 ascoff = 32 start with 'SPACE' character chxsize = 11 8 dots wide (this takes 2 bytes!) chysize = 18 16 dots high chbytes = 36 2 byte wide x 18 dots high ======================================================================== PCPAINT/Pictor Page Format Description Format by John Bridges. Document by Microtex Industries, Inc. Revision Date: 2/9/88 Global Notes: ------------ PCPAINT 1.0 - Revision 1.0 was developed for Mosue Systems in 1984 supported only BSAVE files in CGA 4 color mode. In the space between the scan buffers was a string that read PCPAINT 1.0 followed by 2 bytes which were the pallete and border information for that picture. PCPAINT 1.5 - Revision 1.5 was the same as 1.0 except that it contained larger than screen images and also had a primative packing format. This was sold for so short a time that it won't be covered here. PCPAINT 2.0 thru Pictor 3.1 - This document describes these formats. The file description is identical for all revisions in this range. However, in PCPAINT 2.0, the bit-planes were packed together so that the pictures resembled a PCjr picture, or 4 bits per pixel, 1 bit plane. Starting with Pictor 3.0, the files were saved with the bitplanes separated. This takes a little more memory in some cases, but the speed in loading and saving was a desireable consideration. NOTE TO PROGRAMMERS: A good PCPAINT/Pictor file decoder will use the variables in the header to decode the image and thus be compatible with all formats since the October, 1985 release of PCPAINT 2.0. Also please note that PCPAINT/Pictor are stored from the bottom up. This is opposite that of most of the screen adapters it supports. This really causes no problem, but be aware that you should use a Y table to look up scan lines. In all PCPAINT/Pictor pictures, the scan lines are continuous. If a picture is to be displayed on a particular adapter, the programmer is responsible for using a y-table to properly interleave the lines if necessary. Also note that Pictor was designed for speed, so no inter-mode loading is possible. If you are writing applications that create Pictor images that you want to load into Pictor, you must remain mode dependent. Header - A full description of the file header information. offset type name description ------- ------- ------- ----------------------------------------------------- 0 word marker marker that is always 01234h 2 word xsize x size of page in pixels 4 word ysize y size of page in pixels 6 word xoff x offset into page where lower left hand corner of viewport is located (default of 0 is ok) 8 word yoff y offset into page where lower left hand corner of viewport is located (default of 0 is ok) 10 byte bitsinf bits 0-3 is the number of bits per pixel per bit plane and bits 4-7 is the number of bit planes (so 4 color cga mode would be 02h and 16 color ega would be 31h and plantronics 16 color would be 12h) 11 byte emark marker that is always a 0ffh 12 byte evideo single uppercase letter indicating which video mode this picture was created in, can default to 0. 0 - 40 col text 1 - 80 col text 2 - mono text 3 - 43 line text A=320x200x4 cga B=320x200x16 pcjr, stbplus, tandy 1000 C=640x200x2 cga D=640x200x16 ega E=640x350x2 ega F=640x350x4 ega G=640x350x16 ega H=720x348x2 hercules I=320x200x16 plantronics J=320x200x16 ega K=640x400x2 AT&T or Toshiba 3100 L=320x200x256 vga M=640x480x16 ega plus(video 7, tseng, paradise), vga N=720x348x16 Hercules InColor O=640x480x2 vga 13 word edesc extra information descriptor defines what is in the extra information that follows this header, 0=nothing 1=pallet (single byte) border (single byte)[CGA] 2=pcjr or non ECD 16 color registers (0-15), 1 byte each 3=EGA with ECD 16 color registers (0-63) 1 byte each 4=VGA 256 color info - 256 colors, 1 byte each rgb gun. 15 word esize size of extra information in bytes 17 byte edata[] the actual extra data the size which is defined by esize (at offset 15). 17+ esize word numblks the number of packed blocks in this file. if this is a zero, then data is unpacked. Structures - These C structures describe the header information. struct head { unsigned int mark=0x1234; /* marks begining of a page file */ unsigned int xsize; /* x size of page */ unsigned int ysize; /* y size of page */ unsigned int xoff; /* current x offset into picture of viewport */ unsigned int yoff; /* current y offset into picture of viewport */ unsigned char bitsinf; } struct extra { unsigned char emark=0xff; unsigned char evideo; unsigned int edesc; unsigned int esize; } int edata[esize]; unsigned int numblks; If the file is packed then what follows is a multi block packed file, otherwise (if the file is not packed, numblks=0) the actual data follows. Bit planes follow each other in the file and when packed each bit plane must start in a new packed block. Packed Block Description Packed block header PBSIZE dw ;Packed block size. The size of this block BSIZE dw ;Unpacked block size MBYTE db ;Unique marker byte. This is a byte that does not ; exist in the current unpacked block. If no unique ; byte exists, then pick one that is used rarely ; to avoid too much redundancy. Packed block data - variable size depending on whether 16 bit run is needed. MARKER db ;mark a run (this is where MBYTE goes) LENGTH db ;length of run. if 0, then look at BIGLEN BIGLEN dw ;16 bit run count (only exists if LENGTH==0) DATA db ;byte to fill run with Example 1 - a 320x200, 4 color, packed page file, of a white screen. dw 0x1234 ;marker dw 320 ;x size dw 200 ;y size dw 0 ;x offset dw 0 ;y offset db 02h ;2 bits per pixel and 1 bit plane db 0xff ;extra info flag db 'A' ;vidmode dw 1 ;extra area descriptor (pal and bord) dw 2 ;bytes in extra area db 2,0 ;pallet and border (extra information) dw 2 ;number of packed blocks ;first block dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0xff ;byte to fill run with ;second block dw 5+5 ;packed block size dw 7808 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 7808 ;16 bit run count db 0xff ;byte to fill run with Example 2 - a 640x350, 16 color, packed page file, of a red screen (color 4). dw 0x1234 ;marker dw 640 ;x size dw 350 ;y size dw 0 ;x offset dw 0 ;y offset db 31h ;bits per pixel and 1 bit plane db 0xff ;new extra info flag db 'G' ;vidmode dw 3 ;extra area descriptor (pal and bord) dw 16 ;bytes in extra area db 0,1,2,3,4,5,14h,7 db 38h,39h,3ah,3bh,3ch,3dh,3eh,3fh dw 16 ;number of packed blocks ;block 1 of first bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 2 of first bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 3 of first bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 4 of first bit plane dw 5+5 ;packed block size dw 3424 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 3424 ;16 bit run count db 0 ;byte to fill run with ;block 1 of second bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 2 of second bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 3 of second bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 4 of second bit plane dw 5+5 ;packed block size dw 3424 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 3424 ;16 bit run count db 0 ;byte to fill run with ;block 1 of third bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0xff ;byte to fill run with ;block 2 of third bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0xff ;byte to fill run with ;block 3 of third bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0xff ;byte to fill run with ;block 4 of third bit plane dw 5+5 ;packed block size dw 3424 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 3424 ;16 bit run count db 0xff ;byte to fill run with ;block 1 of fourth bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 2 of fourth bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 3 of fourth bit plane dw 5+5 ;packed block size dw 8192 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 8192 ;16 bit run count db 0 ;byte to fill run with ;block 4 of fourth bit plane dw 5+5 ;packed block size dw 3424 ;unpacked block size db 0 ;marker byte db 0 ;mark a run db 0 ;a 16 bit run count follows dw 3424 ;16 bit run count db 0 ;byte to fill run with Example 3 - For more detail lets consider a block that isn't all the same. Say the data consists of 30 2's, and 8, a 4, and 300 1's. ; the block would look like this dw 5+10 ;packed block size dw 332 ;30 + 1 + 1 + 300 bytes as above db ff ;what to mark a run with, ; because there are no ff's in our example. db ff ;mark a run db 30 ;8 bit run count db 2 ;byte to fill run with - 2 db 8 ;not a run marker, so must be data db 4 ;not a run marker, so must be data db ff ;mark a run db 0 ;means 16 bit run count follows dw 300 ;run count db 1 ;byte to fill run with - 1 The actual unpacked data that resides in memory consists 2 seperate sections. 1. The control structure: contains x size, y size, x offset, y offset, segment of bit mapped data, number of bits per pixel and number of additional bit planes. this information is kept in pcpaint's data segment. 2. The actual bit mapped data: contains the actual page image, mapped from bottom left (so bottom scan line is first). The data is contiguous within each bit plane, so scan line 1 follows scan line 0 directly. the page can and does cross segment boundires (a bit plane can be larger than 64k). each bit plane follows the previous but starts on a paragraph boundary, the printer driver will be passed the offset in paragraphs between bit planes and the number of additional planes. The bit planes start with bit 0, each additional plane is the next bit. .