HPAVC

home *** CD-ROM | disk | FTP | other *** search

/ HPAVC / HPAVC CD-ROM.iso / pc / CIS_GAME.ARJ / QSBSB1.THD < prev next >

Wrap

Text File | 1993-07-01 | 34.1 KB | 721 lines

________________________ Subj: Playing Sound on PC ________________________ Fm: Mark Betz/GD SL 76605,2346 # 181875 To: Intergalactic, Dev., Inc 76356,2172 (X) Date: 27-Jun-92 13:08:37 A couple of choices, Ezra: You can roll your own. This will involve creating a low-level driver (probably driven off the timer interrupt). Or, you can purchase a library, such as John Ratcliff's DIGPAK, which supports a variety of sound devices. Adlib .ROL, and CMS .CMF files are essentially MIDI files with restricted MIDI event/message recognition. The files are parsed, and the driver will turn notes on and off at the proper times. I've done low-level programming to the FM chip, but haven't had occasion to parse MIDI files yet. I imagine it's not trivial programming, as the timing must be fairly precise. --Mark ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 181904 To: Sarwan Narine 76675,164 Date: 27-Jun-92 14:52:53 Hi, Sarwan. There are several routes you can go. The DSP chip on the Soundblaster (and TB) can be driven via DMA, freeing the CPU for other business. Or it can be "pumped" using the timer to send data to it at the correct speed. Whichever route you take will be fairly complex and require low-level programming to the sound card hardware. A third alternative is to use the supplied drivers, which can be loaded and accessed by your program, and a fourth is to purchase a sound devide library like DIGPAK. For alternatives 1, 2, and 3, get yourself a copy of the Sound Blaster Developer's Kit from CMS. It's not great, but you can use it to get the job done, and there is good information on the low-level interfaces to the hardware. Creative Labs, Inc. 2050 Duane Ave. Santa Clara, CA 95054 Sorry! I don't have the phone number at hand. --Mark ........................................................................... Fm: Rasch H Young 70511,2043 # 182020 To: Sarwan Narine 76675,164 Date: 27-Jun-92 22:24:51 Sarwan, I have both the Sound blaster SDK and DIgPak. I believe DIGPAK pro is a better product at a lower price $69.95. The Audio Solution 1-314-567-0267. Rasch ________________________ Subj: Sound card detection ________________________ Fm: Sarwan Narine 76675,164 # 185506 To: All Date: 08-Jul-92 02:47:07 How do you detect the presence of a sound card in a PC system? ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 187160 To: Sarwan Narine 76675,164 (X) Date: 12-Jul-92 19:54:16 Hi, Sarwan. Here is a function (and support functions) for detecting the presence of the FM music chip. If DetectFM() returns true then the chip is present in the system. This tells you that you have either an SB, Adlib, or one of the SB clones. You can further distinguish between the SB and Adlib, but more on that later. The first function we need is a function to send data to the FM chip registers. Here's one in assembler... _SetReg proc far push bp mov bp, sp push ax push cx push dx xor ax, ax mov dx, 0388h ; FM Address Register Port mov al, byte ptr [bp + 6] ; get the register into al out dx, al ; write it to the Address Reg. mov cx, 7 ; need to wait 3.3 microsecs here... wait1: ; which we'll do by reading the in al, dx ; status reg 7 times loop wait1 mov al, byte ptr [bp + 8] ; get the data byte into al inc dx ; point dx to the Data Register Port out dx, al ; write the data mov cx, 35 ; need to wait 23 microsecs here... wait2: ; read the port 35 times in al, dx loop wait2 pop dx pop cx pop ax pop bp retf _SetReg endp The techniques for this come right out of the Sblaster Developer's Kit, which I recommend. It's not the job I wish it was, but it's all that is available as far as I know. I haven't seen any books on the topic. Anyway, the SBDK is adequate if you're willing to go mining yourself for the stuff it doesn't give you. Call CMS at 408-986-1461 if you're interested. Now that we have the function to set the registers we need the the function that detects the card. This writes to some registers, which are referred to in the calls to SetReg(). I won't explain what they do, as this is a topic for a long magazine article, or chapter in a book. // DetectFM() returns true if the FM chip is detected in the system, or // false if it is not. bool DetectFM() { byte stat1; byte stat2; byte result = 0; SetReg( 0x1, 0 ); // initialize test register SetReg( 0x4, 0x60 ); // reset both timers SetReg( 0x4, 0x80 ); // enable timer interrupts stat1 = inportb( 0x388 ); // read the status port SetReg( 0x2, 0xff ); // write ff to the timer 1 count SetReg( 0x4, 0x21 ); // start timer 1 Wait( 90 ); // wait 90 microseconds stat2 = inportb( 0x388 ); // read the status port SetReg( 0x4, 0x60 ); // reset both timers SetReg( 0x4, 0x80 ); // enable timer interrupts stat1 &= 0xe0; // stat1 should be zero after if ( !stat1 ) { // ANDing it with 0xe0 stat2 &= 0xe0; // stat2 should be 0xc0 after if ( stat2 == 0xc0 ) { // ANDing it with 0xe0 result++; // if so, return true } } return( result ); } Oh, and I see we have one other support function required. Wait() simply waits for the specified number of microseconds, approximately. You can substitute your own function if you like, but here is the technique used in the SBDK. I would probably implement it as a more accurate timer-based function, instead of reading the bus the way they do, but I haven't had the time to re-do it. // Wait() approximates a wait in microseconds by reading the bus. Called // only by the DetectFM() function. void Wait( int howLong ) { byte dummy; int reads = howLong / 10; reads *= 16; do { dummy = inportb( 0x388 ); reads--; } while ( reads ); return; } One last thing. I mentioned that you can tell whether it's a Sound Blaster FM chip you're detecting. This is because the SB maps the chip to ports 0x2x8 and 0x2x9 (where x is the base port address of the SB). The Adlib standard is to map the chip to 0x388 and 0x389. To maintain compatibility the SB decodes the chip here as well. So if you can find the chip at both locations it's a Sound Blaster. If not it's an Adlib. ________________________ Subj: Sound card I/O ports ________________________ Fm: Sarwan Narine 76675,164 # 187955 To: Mark Betz/GD SL 76605,2346 (X) Date: 15-Jul-92 02:46:26 Mark, thank you for that informative and complete example of how to detect a sound card. On my ThunderBoard sound card I can set the I/O port addresses to 210h, 220h, 230h, 240h, 250h, or 260h. How do you find out where the I/O ports are? The documentation I have says that only ports 2n6, 2n8, 2n9, 2nA, 2nC, and 2nE are used, where n is 1,2,3,4,5,or 6. What are these ports? A brief description of each port would be appreciated. Thank you. --Sarwan- ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 188069 To: Sarwan Narine 76675,164 Date: 15-Jul-92 15:12:08 Hi, Sarwan. According to my reference the SB uses the addresses from 2n0H to 2nFH where n is in the range 1 to 6 inclusive. Here is a map of the ports: PORT WHAT IS IT WHAT CAN I DO TO IT ------------------------------------------------------------------ 200H-207H ANALOG JOYSTICK READ/WRITE 2n0H C/MS MUSIC VOICE 1-6 DATA PORT WRITE ONLY 2n1H C/MS MUSIC VOICE 1-6 REGISTER PORT WRITE ONLY 2n2H C/MS MUSIC VOICE 7-12 DATA PORT WRITE ONLY 2n3H C/MS MUSIC VOICE 7-12 REGISTER PORT WRITE ONLY 2n6H DSP RESET WRITE ONLY 2n8H FM MUSIC STATUS PORT READ 2n8H FM MUSIC REGISTER PORT WRITE 2n9H FM MUSIC DATA PORT WRITE ONLY 2nAH DSP READ DATA PORT READ ONLY 2nCH DSP WRITE DATA OR COMMAND WRITE 2nCH DSP WRITE BUFFER STATUS (BIT 7) READ 2nEH DSP DATA AVAILABLE STATUS (BIT 7) READ ONLY The C/MS music voice ports are available only on SB cards with these chips installed. The FM music ports can also be accessed at ports 388H and 389H (for Adlib compatibility). --Mark ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 189758 To: Sarwan Narine 76675,164 Date: 20-Jul-92 16:21:03 >> ...how do you find the value of n Well, there are a couple of options. You can ask the user, which is the route most apps take. You can rely on Creative's SBLASTER environment variable (NOT!). Or you can perform the test I outlined for each possible port. --Mark _________________________ Subj: Playing .VOC files _________________________ Fm: Sarwan Narine 76675,164 # 191206 To: All Date: 24-Jul-92 04:12:35 How do you output an 8-bit sample of a .VOC file through the SoundBlaster or AdLib sound card? Also, is there any on-line information on using a DMA channel or timer to playback sound? I'm looking for assembly source code. Thank you. --Sarwan- ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 191420 To: Sarwan Narine 76675,164 Date: 24-Jul-92 20:02:33 Hello again, Sarwan. If you have the manual for the Sound Blaster you will find a simple reference to the functions in the CT-VOICE.DRV driver supplied on disk with the card. This driver is the simplest way to pump data to the dsp chip, since it provides you with some high level control constructs. Alternatively you can write directly to the chip. The DSP chip is actually very simple, being sort of like a pipe with one end attached to memory, and the other to the speaker system. All you really have to do is pour data into it at the proper rate. You can do this either by pumping it through with the timer interrupt, or using the DMA channel 1. I strongly suggest contacting Creative Labs and buying the SB developer's kit. If you simply want to get sound support, then I suggest the DIGPAK library. Leave mail or a message here for John Ratcliff 70253,3237. The have a lib for sale at a very good price which supports midi and digitized sound on just about every device you can imagine, if half of what I've heard is true. --Mark ........................................................................... Fm: James Mayes 71151,3037 # 203245 To: Mark Betz/GD SL 76605,2346 (X) Date: 21-Aug-92 18:04:08 I also have another question. I recall your recently discussing with someone else your experience programming the Soundblaster by loading the driver into memory and then making calls to the driver with the correct registers initialized properly. I have always used the SB Development Kit by Creative, but I am curious about the direct method and I was wondering about a few things. (BTW, I am strictly talking about digitized voice and the CT-VOICE driver) 1. The SBK says you need to load the driver at offset 0 of some segment, load the registers and make a far CALL to that seg,off. What would be the syntax for making this call from C given a far pointer to the buffer containing the driver? 2. The CT-VOICE driver doesn't include any functions for detecting the SB card or scanning for the correct interrupt number. Are you aware of any good sources for this info? 3. Do you know if there would be a problem with distributing the CT-VOICE driver file with a game that used this approach? (I don't think there should be, but those kind of questions always worry me.) ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 203304 To: James Mayes 71151,3037 (X) Date: 21-Aug-92 19:47:00 Hi, James. I'm not sure what you mean by "the direct method". Is there some other way to access CT-VOICE.DRV explained in the SDK? Basically you have to get the driver into memory first. That requires allocating memory for it, and reading it in. The twist is that you have to allocate an extra paragraph, or 16 bytes, so that you can force the offset to 0. the new keyword and alloc() functions usually return an offset of 4, and in any case you can't guarantee that the offset will be 0. Allocate an extra paragraph, and do a little segment/offset arithmetic to normalize the pointer (make sure to keep a copy of the original for use with free() or delete). Once you get the driver read into memory you can set the IRQ and port address directly. The offsets of these values are listed in the SDK, but anyway, they are right near the beginning of the CT-VOICE driver file; maybe offset 40h or thereabouts. As for finding out which values to use, you either have to ask the user, or test all possible combinations. From that point on it's a simple matter of casting the driver buffer pointer to a pointer to function, and then making calls to it. void* buf; // pointer to buffer void (*drvr)(); // function pointer buf = new byte[sizeof_driver]; // alloc driver memory drvr = (void(*)()) buf; // not sure about this cast! ........................................................................... Fm: Sarwan Narine 76675,164 # 203407 To: All Date: 22-Aug-92 02:06:32 Well, I've posted this question before, with unsucessful results. So, I'll post it again -- because I like to aggravate people <g>. OK, here it is: How do you output an 8-bit sample to the SoundBlaster card? Assume the SB is installed at 220h. Thanks. ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 203510 To: Sarwan Narine 76675,164 (X) Date: 22-Aug-92 13:39:16 Sarwan, if you're talking about driving the output from your application, then you are into several complex areas of hardware programming, including the interrupt controller, dma controller, and the dsp chip on the SB. I'll give you skeletal procedures for doing it with the driver, but you'll have to dig in your references for the details. Using the driver: 1. Allocate ram for the driver on the heap. The driver needs to be loaded at offset 0 of a segment, so grab a paragraph more than you need, and do pointer math to force the offset to 0 and the seg to the largest possible value. 2. Read the driver (CT-VOICE.DRV) into memory. 3. Set a function pointer on the first byte of the driver NOTE: to call a driver function set BX = the function no., and set any other registers needed, then call the driver. 4. Set the base I/O address. BX = 1, AX = base I/O addr. 5. Set the dma interrupt. BX = 2, AX = dma interrupt no. 6. Initialize the driver. BX = 3. Returns error code, AX. AX = 0: driver init successful AX = 1: bad driver version AX = 2: I/O read/write failure AX = 3: DMA interrupt failure 7. Load the voice data into memory (see block structure) 8. Turn the speaker on. BX = 4, AL != 0. 9. Start voice output. BX = 6, ES:DI = voice data, returns error code, AX. AX = 0: output successful AX != 0: failure 10. Turn speaker off. BX = 4, AL = 0. 11. Terminate driver. BX = 9. Note that the voice data you output must be in CMS .VOC format. This format uses a block structure, with each block of data preceeded by a header which identifies it to the driver. The only block-type you're concerned with at this point is the Voice Data Block. It looks like this: byte 0 : type field = 01 1-3: three bytes data block length (actual length - 2) 4 : time constant = 256- (1000000 / sampling rate) 5 : compress. (0: raw, 1: 4-bit, 2: 2.6-bit, 3: 2-bit) Make sure one of these blocks is at the front of your voice data and initialized correctly. Once you start the voice output you'll get control of the processor back, and your program can continue. ........................................................................... Fm: James Mayes 71151,3037 # 203802 To: Mark Betz/GD SL 76605,2346 (X) Date: 23-Aug-92 10:45:17 Thanks a lot, Mark. By "direct method" I simply meant writing my own code to access the driver rather than calling the ctvm_XXXX routines supplied with the SDK. The longer I program, the worse my English becomes <g>. The info you gave me was exactly what I was looking for. I think I will give it all a try very soon. Thanks for your help!! ........................................................................... Fm: Sarwan Narine 76675,164 # 207584 To: Mark Betz 76605,2346 (X) Date: 02-Sep-92 02:51:10 Hi Mark. I haven't had much success with playing back a .VOC file thru the SoundBlaster. Here's what I've got (assume SB installed at 220h): ; reset DSP OUT 226h, 1 OUT 226h, 0 ; turn on the speaker OUT 22Ch, 0D1h ; playback loop PLAYBACK_LOOP: OUT 22Ch, 10h OUT 22Ch, SAMPLE DELAY LOOP PLAYBACK_LOOP ; turn off the speaker OUT 22Ch, 0D3h Hey kids, don't try this at home. The above is "pseudo-code", don't try it on your assembler <g>. OK, Mark, the above _occasionally_ outputs something resembling the .VOC file contents, although it sounds like it's played-back in slow-motion. Most of the times I get no output. Any suggestions? ........................................................................... Fm: Mark Betz/GD SL 76605,2346 # 207964 To: Sarwan Narine 76675,164 (X) Date: 02-Sep-92 23:34:56 Hi, Sarwan. The procedure outlined in the manual says that you must wait 3 microseconds between the two steps of the DSP reset process. It could be that this is mandatory, and it could be intermittent because certain variances caused by a cache hit/non-hit or something are causing you to get a >= 3 microsecond delay on some calls, and not on others. I'd specifically place the delay in the function. As for slow playback, that is almost certainly related to the speed with which you're sending the data to the DSP. ........................................................................... Fm: Hans Peter Rushworth 100031,473 # 214082 To: Mark Betz/GD SL 76605,2346 (X) Date: 17-Sep-92 10:08:04 >> The volume can be controlled by altering the sampled data in certain mathmatically precise ways that I do not know <g>. It is much simpler than you think! If the samples are simple signed amplitudes, then just multiply each 8-bit sample by a constant from 0 to 256, and use the top 8 bits of the result. If they are unsigned, convert to signed by subtracting 128 first, and adding 128 afterwards. (you must use a signed multiply or equivalent shift operation). The output volume is not linearly related to the multiplier. A range of 0 1 2 4 8 16 32 ... (ie use shifts!) will give a more or less linear progressive increase in volume (ie half the amplitude is NOT half as loud). With low multipliers the quantisation error (hence noise) will be high, unfortunately. Only 8-bits are a bit limiting I'm afraid (pun intended), so for a finer variation in volume you may need find intermediate values for the multiplier. For example, the multiplier between 128 and 256 is 181, (not 192) which is found by taking 2^7.5 (2^7 is 128 and 2^8 is 256, so 2^7.5 is half way in between). You shouldn't simply add values as you suggest, since this will introduce high levels harmonic distortion, not what you want I suspect. _______________________ Subj: Sound Blaster Effects _______________________ Fm: Jesse 76646,3302 # 267987 To: all Date: 26-Dec-92 15:32:45 Here's a thought, though it's either easy to do and has already been done, or it can't be done and hasn't [g] Instead of worrying about .vocs for small sound effects like gunfire, explosions, engine noise, etc, couldn't we just create an instrument for each sound and dynamically "play" them as the game goes on? I've only peeked at the FM-Organ thing that came with the SB, but I seem to recall it letting several instruments play at once. ........................................................................... Fm: John W. Ratcliff 70253,3237 # 268036 To: Jesse 76646,3302 Date: 26-Dec-92 17:43:38 Jesse, >>just create an instrument for each sound and dynamically "play" them as the game goes on? Yes it can be done. That is in fact how all sound effects are made in Ultima Underworld and a number of other games. However, it is very difficult to do, has no analgy on non-FM synth devices, and digital sound effects are easier and sound better. Plus I'm selling DigPak and people should definatly be doing digital effects from my point of view. You can do all of you music/FX that way if you use the Audio Interface Library from Miles Design Inc. However, AIL has a rather overwhelming API and is out of the price range of the small developer. John. ........................................................................... Fm: Mark Betz/Ass't SysOp 76605,2346 # 268883 To: Jesse 76646,3302 (X) Date: 28-Dec-92 11:26:54 Hi, Jesse. I don't know how many people use digitized sound for engine noises. The engine is one area where a synth chip can do a fine job, so my approach has been to create the engine noises on the synth, leaving the sampled sound to be used for specific effects. _____________________ Subj: Sound Blaster Programming _____________________ Fm: Everett Kaser (Sherlock) 70673,1547 # 310695 To: Vu Truong (Siliconis) 70242,3015 (X) Date: 08-Mar-93 22:10:28 There's a file available from an FTP site on internet (and may be somewhere here on CIS) written by Jeffrey S. Lee for the Soundblaster Freedom Project. It's file name is usually SFP.ZIP. It contains lots of good reference information regarding the Soundblaster (and Adlib) registers. I have a couple of other files that I've snarfed off of internet that contain file formats for SBI files and CMF files. If they're not already available in one of the libraries, and someone would point out the appropriate place for them, I'd be glad to upload them (on Tuesday, for a hamburger today <G>). Everett ........................................................................... Fm: Dan Corritore 70243,1110 # 310714 To: Vu Truong (Siliconis) 70242,3015 (X) Date: 08-Mar-93 22:36:29 I have the format here, but I'd rather not type the long descriptions that are on it. If you want more description, buy 'The Sound Blaster Book'--very useful, especially in the file-formats category. Bytes Description 0-3 File ID-- 'SBI' followed by 0x1a 4-35 Instrument Name--name of instrument (ASCIIZ form) 36 Modulator sound properties 37 Carrier sound properties 38 Modulator volume 39 Carrier volume 40 Modulator attack/delay 41 Carrier attack/delay 42 Modulator sustain/release 43 Carrier sustain/release 44 Modulator waveform 45 Carrier waveform 46 Synthesis mode and phase shifting 47-51 Future use Sorry if the above does not appear in columns.. my software doesn't give a 'WYSIWYG' interface in messages. And also sorry that the above weren't very descriptive! _Dan ........................................................................... Fm: Everett Kaser (Sherlock) 70673,1547 # 313151 To: Everett Kaser (Sherlock) 70673,1547 Date: 13-Mar-93 15:07:10 OK, here's the internet ftp sites for soundboard related stuff that I had at work and their directories: ftp.brad.ac.uk misc/mod saffron.inset.com pub/unix/sound_drivers (access limited to 3:30pm-5:30am) ftp.ulowell.edu msdos/Sound/AdlibSB wuarchive.wustl.edu mirrors3/garbo.uwasa.fi/sb snake.mcs.kent.edu pub/SB-Adlib ccosun.caltech.edu pub/heathh/sb Good luck! I hope you find what you want. Everett ........................................................................... Fm: Everett Kaser (Sherlock) 70673,1547 # 313152 To: Vu Truong (Siliconis) 70242,3015 (X) Date: 13-Mar-93 15:07:30 Ok, I was unable to track down the file itself (that documents the .SBI file format) in my CAREFULLY arranged and METICULOUSLY indexed disk library <G>, so I typed in from a printed copy the section pertaining to the SBI format: ----------------------------------------------------------------------- Sound Blaster Instruments File (SBI) Format: ---------------------------------------------- The SBFM driver programs the registers of the FM chips to synthesize the different instrument sound. Such register information is stored in the SBI file with one instrument per file. The Sound Blaster IEDIT program allows users to define new instruments and store them in the SBI format. The format of the SBI file is as follows: Offset (hex): Description: ------------- ------------------------------------------ 00 - 03 File ID (ASCII string "SBI", ends with 1A hex) 04 - 23 Instrument Name (NULL terminated ASCII string). 24 Modulator Sound Characteristic 25 Carrier Sound Characteristic Bit 7 : Pitch Vibrato (AM) Bit 6 : Amplitude Vibrato (VIB) Bit 5 : Sustaining Sound (EG-TYP) Bit 4 : Envelope Scaling (KSR) Bit 3-0 : Frequency Multiplier (MULTIPLE) 26 Modulator Scaling/Output Level 27 Carrier Scaling/Output Level Bit 7-6 : Level Scaling (KSL) Bit 5-0 : Output Level (TL) 28 Modulator Attack/Decay 29 Carrier Attack/Decay Bit 7-4 : Attack Rate (AR) Bit 3-0 : Decay Rate (DR) 2A Modulator Sustain Level/Release Rate 2B Carrier Sustain Level/Release Rate Bit 7-4 : Sustain Level (SL) Bit 3-0 : Release Rate (RR) 2C Modulator Wave Select 2D Carrier Wave Select Bit 7-2 : All bits clear Bit 1-0 : Wave Select (WS) 2E Feedback/Connection Bit 7-4 : all bits clear Bit 3-1 : Modulator Feedback (FB) Bit 0 : Connection 2F - 33 Reserved for future use ------------------------- that's all folks! ------------------------------ Everett _______________________ Subj: VOC Compression Method _______________________ Fm: Bryant Bunderson 71742,1076 # 378740 To: All Date: 18-Jun-93 15:43:46 I am trying to write a program to compress my PCM sound files to the 4 bit ADPCM format used in the Sound Blaster VOC files so they don't take up so they don't take up so much space. I have only been partially successful. Does anyone out there have a specification or code fragment describing the compression algorithm used in 4 bit compressed VOC files? I have seen a lot of technical know how in this forum and thought I would ask here. Any help would be appreciated. Thanks, Bryant Bunderson ........................................................................... Fm: Bryant Bunderson 71742,1076 # 379544 To: Andrew Amess 100141,1567 Date: 19-Jun-93 23:25:44 Yes, the ADPCM method does lose information. It stands for Adaptive Delta Pulse Code Modulation. Basically instead of saving all the bits of a sample (PCM) it stores the difference or delta between two samples using fewer bits. Generally the fewer bits you use to represent the deltas the less the resulting wave looks like the original. I need my compressed sound files to be compatible with the Sound Blaster VOC format so I really should use thier method of ADPCM compression but thanks for the response. Bryant ........................................................................... Fm: Patrick Reilly 71333,2764 # 379632 To: Bryant Bunderson 71742,1076 Date: 20-Jun-93 03:20:04 Bryant, I though the Sound Blaster used a modified form of ADPCM (I know that's what *my* code used <g>) - a sentinel value is used to allow the following word to be full (in this case) 8-bit. For example, using 3-bit encoding, 0-6 are delta values; a 7 indicates that the decoder should grab the following full 8-bits and use this as an absolute value instead of a delta. If the system is designed right, you can get nice size savings; I had some 8-bit sound files that ended up being (average) about the same size using 2-bit; for 3-bit, they averaged about 2/3, etc. In fact, it varied so widely from file to file that I ended up using a file header and allowing different bit sizes. The decode program (read in .WAV files) would go through and try 2, 3, 4, 5, and 6-bit encoding; whoever came out smallest won; if they were all bigger (never happened) used a special header signature to show that it was a raw 8-bit data file... Worked pretty well with PC-speaker sound library... -Pat- ........................................................................... Fm: Kirk Bateman 100112,71 # 379643 To: Bryant Bunderson 71742,1076 Date: 20-Jun-93 05:57:22 I wouldn't use the ADPCM routine SB style to compress your SB data because the loss in quality is too great, look in the CLMFORUM under the Data Compression Section there is some interresting work there. ---Kirk/UK ........................................................................... Fm: Bryant Bunderson 71742,1076 # 380900 To: Patrick Reilly 71333,2764 (X) Date: 21-Jun-93 18:45:14 I will look for a "sentinal bit" change in the data stream and see if I can determine if it is being used. I had come to the conclusion that they were doing something like that but didn't see a pattern. At least not a byte aligned pattern. Thanks for the help. Bryant P.S. Do your internal speaker routines work pretty well? I have pretty much crashed and burned every time I try to get intellagible playback through the internal speaker. ........................................................................... Fm: Patrick Reilly 71333,2764 # 381245 To: Bryant Bunderson 71742,1076 (X) Date: 22-Jun-93 02:15:00 Bryant, Sentinel value - I know that for *my* encoding/decoding, what I use for the sentinel value is the 'odd man out'. For example: with 4-bit encoding, this means that you have 1 sign bit and 3 'value' bits, for a range from +7 to -8. In this case, -8 is the sentinel value, and only +7..-7 are delta values (-8 is 1000b). Yeah, my speaker lib worked fine; I only ran it in the foreground though (actually, it ran in the background, but I used a while-loop to do nothing until the buffer finished playing). I would seriously think that there would be some files here on CIS that have a speaker driver... -Pat- ........................................................................... Fm: John W. Ratcliff 70253,3237 # 381154 To: Bryant Bunderson 71742,1076 (X) Date: 21-Jun-93 23:30:31 An algorithm called ACOMP is in the public domain. It was published in the July 1992 issue of DDJ. It uses a special type of ADPCM that features a user selectable amount of loss, supports silence squelching, and provides extremely fast decompression speed. Go DDJFORUM to find the source. John. ...........................................................................