Atari/Atari Games Memos and Status Reports 1982 Jed Margolin ___________________________________________________________________________ To: Lynn Nyberg From: Jed Margolin Re: Composite Negative Sync for Wells-Gardner 19K4603 and 19K4633 color monitors. The Wells-Gardner 19K4603 and 19K4633 color monitors will accept only positive sync. Testing games like DIG DUG with composite negative sync will require either MTO'ing new monitors or replacing the interface boards in the 4603/4633's with P317's (the one in the 4625). The horizontal centering control on the new board will not function with the older chasis, but that is the only difference. The P317 interface board costs $34.70; all inquiries should be directed to Brenda Thompson. cc: Darl Davidson Fred McCord Rick Moncrief Dave Stubben Bill Woods _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Speech Status Dt: 2/11/82 In this memo I will discuss TI's new text-to-speech system, their coming LPC development system, and their offer to sell us vocabulary. Also, I will mention GI's continuing lack of interest. On Wednesday, February 10, I attended a demonstration at the Texas Instruments Regional Technology Center, where I was shown a text-to-speech system. The basic system consists of a TMS-7000, 2K of RAM, 10K of language rules, 3K of LPC Allophones, and a TMS-5220 speech synthesizer IC. The input consists of serial ASCII text. When the IC's are available (a few months) a complete set would cost about $30 in quantities of 100K. The quality was easily the best of the phoneme/allophone systems I have heard, but it would definitely not be confused with straight LPC. The only application I can think of is where a game is called upon to speak a name or phrase entered by the player. For example, if the player entered his name at the beginning of the game, the game could speak to the player in a more personal manner. [ Watch out, Rick, there is a torpedo behind us.] It could also announce the name of the person with the high score. [ "The highest score achieved on this game was by John Smith, who scored 107,740 points, and who is a fine human being and an example to all men."] The actual phrasing would have to have many variations so the operators won't get sick of it and turn it off. TI is bringing out an LPC development system for $20K - $25K. This system, which they are calling "PASS" should be available in a few weeks for a demo at the Regional Technology Center. Mike Walack of TI said that they are planning on making only 40 of them and that they will probably be all gone by the end of June. (It sounds like TI recruits used car salesmen for its staff.) TI has a total vocabulary of about 1500 words which they would be willing to sell to us for about $10K. I will be getting an alphabetised list. GI gave me a document which was supposed to contain the information on the algorithms used to expand the data in the vocabulary ROMs. Several crucial pages had been omitted from the copy I received. GI claims to be preparing a proposal for setting us up to do our own speech encoding. the end _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: A.H.Systems Portable Motorized Antenna Mast Dt: 2/12/82 I called A.H.Systems and talked to Art Colin about their Portable Motorized Antenna Mast. The 14 foot version (AMS-4) costs $4000 and the 22 foot version costs $4200. Delivery is about 3 weeks. The system is completely portable but requires guying in other than no wind. (But the guys can be very light.) The mast is 4 inch PVC reinforced with wood. They are working on a computer controlled mast as well as a computer controlled turntable. They would controlled through an HP-IB interface as well as a manual remote control. The antenna mast would have a resolution of 1 cm and would have an antenna polarity changer. The turntable would have a resolution of 1 degree. Art expects the package to be available in 3 - 4 months and cost $15K. People buying the antenna mast now will be able to retrofit the computer control. He is sending literature. With this equipment, a radiated test could be completely automated except for changing antennas. the end _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Status of Speech Synthesis Dt: 2/22/82 General Instrument has been unwilling or unable to supply the support we would need to successfully use their speech synthesis system. This situation is unfortunate since their system is in some ways superior to TI's. 1). During the past eight months I have repeatedly requested the vocabulary packing algorithms used with the SP0250. Finally, two weeks ago I received an incomplete document that was for a different chip (the SP0256). 2). During the past eight months I have repeatedly requested a comprehensive proposal from GI that would enable Atari to develop speech vocabulary in-house. I have not received any serious proposals. 3). I have discovered that the speech data for the SP0250 can be manipulated on-the-fly to produced some useful effects. Words can be whispered, or have their pitches offset, or have their pitches made monotone, or have their speaking rate modified. I did this with a Mouse Brain (and with no help from GI). In October, GI promised that they would modify the program in the PIC-1650A that is used to control the SP0250 so that it, too, could produce these special effects. The last information on the subject came from Bill Nowlin ( the local GI representative) who confirmed, as per my conversation with Phil Mclaughlin (GI Group Leader) that GI would not consider modifying the program unless Atari pays for the work. I strongely recommend that we do NOT. The most promising developments have been with Texas Instruments. 1). They expect to demonstrate a speech development system next month. The system would cost $20K - $25K. 2). They have an existing vocabulary of 1300 words. 3). They expect to bring out prepackaged vocabularies in 128Kbit serial ROMS. I have vocabulary lists for them (called Time/Weather, Military, and Avionic). They already have an Industrial vocabulary which I have demonstrated. The latest quote, received February 22, is as follows: Quantity 10K 25K 100K TMS-5220 Speech Synthesizer 9.75 8.45 6.70 TMS-61001 128Kbit serial ROM (Industrial) 5.10 4.60 TMS-61002 128Kbit serial ROM (Time/Weather) " " TMS-61003 128Kbit serial ROM (Military) " " TMS-61004 128Kbit serial ROM (Avionics) " " TMS -6100 128Kbit serial ROM 8.50 7.40 5.60 TMS-6100 with words chosen from TI's existing vocabulary: $25/word masking charge (The TI rep thinks it can be negotiated.) I recommend that copies of the vocabulary list be distributed and people choose 150 of the words they think would be most useful. CC: Dave Stubben the end _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Prices For Speech Vocabulary Dt: 3/3/82 The latest prices from Texas Instruments for custom vocabulary, using their speaker, are as follows: 1 - 25 words $140. 25 - 50 words $120. 50 - 100 words $100. 100 - up words $ 90. In addition there would be a $3000 mask charge for the 128 Kbit serial ROM. Lead time for custom serial ROMs is 12 - 16 weeks. This compares to $25 a word, selecting from TI's existing vocabulary. Since the last memo, I have learned that there would be a $3000 mask charge in ADDITION to the $25 per word charge. Lead times for serial ROMs would still be 12 - 16 weeks. The Speech Synthesizer can also be used with external speech data delivered by the host processor. It would require a new byte of data every 3 ms (worst case). A 2532 would be good for 25 seconds of speech (about 50 words). In quantities of 25K it would cost $7.52 for a custom programmed 6100 128Kbit serial ROM (not counting vocabulary costs). Assuming that 2532s are $5 ea. it would cost $20 for an equivalent amount of speech. However, since each game could have its own vocabulary ROM(s) we might not need 128Kbits for each game. It depends how much vocabulary a game needs. TI has offered to sell us their existing 1300 word vocabulary for $10K. Since only about 600 appear useful, it works out to $16/word. Of course, if we were to put the data in 2532s anyone could copy it and get the vocabulary without paying $10K. TI expects to have samples of the TIME/WEATHER ROM in 2 weeks with the MILITARY and AVIONICS ROMS becoming available in five weeks. They expect to be able to demonstrate the PASS speech development system around the middle of March. cc: Dave Stubben Mike Halley _____________________________________________________________________________ To: Distribution Fr: Jed Margolin Re: Speech Development System Demonstration Dt: 3/12/82 On Tuesday, March 16, Texas Instruments will demonstrate their "PASS" speech development system which takes audio in and produces LPC data out (for the TMS-5220). The demonstration will be held at 2:00 pm in the Downstairs Conference Room. Distribution: Rick Moncrief Dave Stubben Gary Baily Morgan Hoff John Ray Dave Sherman Ed Logg Scott Fuller Joe Coddington Rich Moore Greg Rivera Gary Stempler Owen Reubin(sp?) Steve Calfee Tom Hogg Mike Albaugh _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Speech Status Dt: 3/19/82 Summary: The method of using the Texas Instruments speech synthesizer that is best suited for the way we develop and produce games is where the speech data is stored in standard byte wide ROMs. Also, the most effective way for game designers to experiment with the use of speech in games is for Atari to purchase a PASS Speech Development System from Texas Instruments for $25K. If we arrange to have the PASS system here (for two days) we can, I hope, demonstrate to management some potential uses of speech in games. This assumes we want to start using speech now. If we wait 10 months TI will probably have a less expensive speech development system available and/or other companies might have hardware and software available. ------------------------------------------------------------------------------ Methods of Controlling the Synthesizer: Method 1 - Vocabulary from off-the-shelf serial ROMs (150 - 205 words in each one) 25K 100K TMS-5220 Speech Synthesizer 8.45 6.70 6.70 TMS-61002 Industrial Vocabulary 4.60 TMS-61003 Time/Weather 4.60 TMS-61004 Military 4.60 TMS-61005 Avionics 4.60 Note: These serial ROMs are very easy to use from a game program with 4ms interrupts. Method 2 - Speech Synthesizer with custom serial ROM TMS-5220 Speech Synthesizer 8.45 6.70 TMS-6100 128Kbit serial ROM 7.40 5.60 The TMS-6100 would be programmed with: 1. words from TI's existing vocabulary at a cost of $16-$25 per word 2. custom words from TI at $90 per word 3. vocabulary developed in-house by Atari Note: 1. The ROMs incurr a $3K mask charge and production quantities take 12-16 weeks. 2. very easy to use from a game program with 4ms interrupts Method 3 - TMS-5220 Speech Synthesizer with Speech data in our own byte wide ROMs. A 2532 would hold 40-50 words. Vocabulary would come from TI or be developed in-house by Atari. Note: 1. The ROMs would be our standard ROMs in the program address space. 2. The system is moderately easy to use from a game program - however, the sample program I have written for it requires attention every 1.5 ms or so (although half of the time it is called, the routine is guaranteed to take 24 us or less). Method 3A - Super Sound TMS-5220 Speech Synthesizer Standard ROMs Dedicated 6502 system to do speech, Pokey sounds, and all coin routines Note: 1. This is the most expensive method but would allow the data from the synthesizer to be modified in real time for special effects (I have done a sample program to show these capabilities). It would also allow more processing time to be devoted to Pokey sounds, and would allow these sounds to be fully developed, separate from the game. Discussion: Method 1 is feasible only if we can work the existing ROM vocabulary into a game. Method 2 is feasible only if we can afford 16 weeks from release to production. Methods 3 and 3A are the ones best suited for the way we develop and produce games. ----------------------------------------------------------------------------- The next question is, "How can we get vocabulary?" 1. We can buy TI's vocabulary of 1300 words for $10K. 2. We can pay TI $90/word for custom words, using either their speaker or ours. 3. We can buy the PASS Speech Development System from TI for $25k and do our own vocabulary. A caveat is in order concerning number 3. 1. Not everyone's voice is suitable for LPC encoding 2. The data rate without editing is about 1700 bits/sec (versus 1200 for edited). We may be able to learn to do our own editing. 3. The PASS system can process about 10 secs of speech at a time. The system can produce EPROMs from this or can upload it through an RS-232 port. A programmer would have to create the software to enable the VAX to handle the speech files so a game programmer could put together an entire vocabulary. 4. The PASS system does not have the capabiliy of downloading and and playing speech, however, I already have the hardware and software for doing this. ----------------------------------------------------------------------------- If we want we can have the PASS system here for two days. It would cost us $500 for two days ($250/day), which we could deduct from the cost of the system if we buy one. We would also get Lisl (their resident linguist) for a day. In addition we would have to agree to be responsible for the unit while we have it. Iraj was somewhat embarrassed about the money but explained that the Regional Technology Center has a separate profit and loss responsibility and I assume that means they want at least a small piece of the action since if we buy the PASS system we probably won't use their services. Before bringing in the system (if we do), a memo should be sent to all project leaders, engineers, programmers, technicians, and anyone else who might be interested, asking them to think about what a game should say, and also to audition prospective speakers. Two weeks' notice should be sufficient. Then we should have a meeting. The purpose of bringing in the PASS system would be to see if we would be able to use it; the purpose would not be to have it here for two days and develop all the vocabulary we might need for a particular game. ------------------------------------------------------------ We need to start thinking about what the policy is going to be concerning the speaker. First off, we need to find one (or two). Will the project teams have to find their own? Will the speaker used during development be the same one used for the final version? Will the speaker receive game bonus or will the reward be purely ego? Do we want to use a professional speaker? Do we want to conduct a company wide talent search? We will eventually need to set up a quiet, acoustically dead, room. -------------------------------------------------------------- cc: Lyle Rains Dave Stubben _____________________________________________________________________________ Preliminary Draft To: Distribution Fr: Jed Margolin Re: Speech Synthesis Meeting Dt: 2 April There will be a meeting on Tuesday, April 13 at 2 pm in the downstairs conference room for all of those who are interested in using speech synthesis in a game. Please come prepared to discuss how you propose to use it. Of the speech synthesis systems available, the system by Texas Instruments is the one best suited for the way we develop and produce games. They have a speech development system available (which we are thinking of buying) that takes audio in and generates LPC data for the synthesizer. We will have the speech development system here at Atari on April 20 and 21. TI's linguist will be here on the 20th to show us how to use the system. Please see me if you will be coming to the meeting or if you want technical details on the speech synthesis system. Distribution: Programmers Engineers Technicians Project Leaders _____________________________________________________________________________ XY Graphics Terminal 1. Target a Market . Home or Commercial determines FCC class . What types of host computers will it be used with? . What types of interfaces do they have? . What is the target cost? 2. Define The Product (Or Product Line) . what are its proposed uses? . Design Philosophy - Do what we know how to do . The choice of interface is intimately connected with the definition of the product and the targeted market . The terminal might also be useful as a general purpose XY graphics terminal for in-house use. 3. How Will It Be Marketed? . Coin-Op can design it and license the design to some other company . Coin-Op can build it and provide documentation and have another company market it, support it, and provide warranty service. . Can Coin-Op market it and support it itself? . should it have the Atari name on it? . Will the operators object to our selling it? . Can the design be protected? Will we need to use custom LSI for that purpose? . How will it be test marketed? 4. Decide who is on the project and what the responsibilities will be. 5. What about UL and FCC? . FCC Class B (home computer) would probably increase development time because of certification. ------------------------------------------------------------ Possible Product Line 1. XY Graphics Controller with Color XY Display 2. XY Graphics Controller with B/W XY Display 3. Remote Color XY Display (self-powered) 4. Remote B/W XY Monitor (self-powered) ------------------------------------------------------------ Possible Interfaces RS-232: 9600 Baud max which is about 1000 bits/sec which is 100 bytes/100 ms update (minus housekeeping) Atari 800: 19,200 Baud which is about 2000 bits/sec which is 200 bytes/100 ms update (minus housekeeping) IEEE 488: typically 250 KBytes/sec which is 25 KBytes/100 ms update Custom 16 bit parallel with full handshaking PDP-11 Bus LSI-11 Bus *** The graphics controller will have to be shielded *** and each I/O line will have to be decoupled. Packaging considerations make it undesirable to have multiple interfaces. _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Dt: 4/26/82 Re: TI Speech Development System The Texas Instruments PASS Speech Development System does work pretty much as advertized and could be used to generate vocabulary for their speech synthesizer IC. Several people have expressed an interest in using the system if we buy one. They are Milt Loper (4x4), Bruce Merrit, Rich Moore (Adventure), and Greg Rivera (Warp Speed). As things stand, TI is the only company that offers a speech development system for LPC Speech Synthesis.(The other systems that are available are for Allophone Synthesis systems like Votrax.) Nevertheless, I do have some reservations about buying the TI PASS system: 1. As things stand TI will not give us schematics for it even though all the boards except one are standard boards. This would make it difficult to fix if it broke down. 2. The user documentation is poor. For example, it does not explicitly state how to read in and execute speech data from the PROM burner. 3. More importantly, the user's manual I was shown did not document the RS-232 uplink (either hardware or software). This information is vital if we are to be able to upload the vocabulary data to the VAX. 4. The unit does not always work after it is powered-up. The pushbutton which is alleged to be a reset button does not perform a full system reset. 5. It would be useful if there were a written document on the rudiments of editing the LPC data. According to Lisl (TI's linguist) this lack of documentation is the result of TI Corporate policy. If this is the case perhaps we can get one of our big cheeses to talk to one of their big cheeses. Despite my reservation, if we want to do our own speech development this is the system to get. the end _____________________________________________________________________________ POWER SUPPLIES - The following power supplies are currently in use, and are UL Recognized for use in video games. PS ASSY A037671-XX A037672-XX A037396-XX A037875-XX NAME Power Supply Shielded Color XY Shielded Color XY Power Supply TRANSFORMER 142003-001 142003-003 142005-001 142005-002 142003-002 142003-004 TERMIN.TRANS A035888-01 A035888-03 A037395-01 A037395-02 A035888-02 A035888-04 WIRING DIAGRAM 037669-01 037669-01 037394-01 037394-01 USE B/W XY B/W XY Color XY Color XY Color Raster Color Raster PS ASSY NAME TRANSFORMER TERMINATED WIRING USE FOR TRANSFORMER DIAGRAM A037671-XX Power Supply 142003-001 A035888-01 037669-01 B/W XY,Color Raster 142003-002 A035888-01 A037672-XX Shielded 142003-002 A035888-03 037669-01 B/W XY,Color Raster Power Supply 142003-004 A035888-04 A037396-XX Color XY 142005-001 A037395-01 037394-01 Color XY A037875-XX Shielded 142005-002 A037395-02 037394-01 Color XY XX = 01 Domestic (Voltage Block - 120) = 03 International (Voltage Blocks - 220, 240) = 03 European (Voltage Blocks - 220, 240) ------------------------------------------------------------ The Following Regulator Audio Boards are Available: REGULATOR/AUDIO II Assembly Wiring Audio +5v @ 6A +10v +22v @ 1A -22v @ 1A +12v @ 1A -5v @ 0.5A Diagram Reg. Reg. Reg. A035435-01 035435-01 X X X A035435-02 035435-02 X X X X X X X A035435-04 035435-04 X X X X X A035435-05 035435-05 X X A035435-06 035435-06 X X X X X X ____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Texas Instruments Dt: 5/3/82 I have attached to this memo a copy of the purchase agreement that Texas Instruments proposes that Atari sign if we buy a PASS system. Iraj's reassuring cover letter is completely worthless according to the terms of the purchase agreement. See Section 4 LIMITED WARRANTY. The main points are: 1. After the 30 day warranty period, TI is under no obligation whatsoever to support the system. If it breaks down we are S-O-L. I should add that it would probably be at least 2 months before it could be fully usable with the VAX system (longer, if we move to Milpitas during this period). 2. TI will not warranty that the PASS system is even fit for our purpose, that of developing LPC vocabulary to be used with TI speech IC's which Atari is considering using. 3. Section 6 PATENT RIGHTS seems to indicate that buying TI's speech IC's does not give us the right to use them without negotiating a License. ALSO, IT IS NOT EVEN CLEAR THAT WE WOULD HAVE THE RIGHTS TO THE VOCABULARY WE DEVELOP ON THE PASS. We have several alternatives open to us: 1. Buy the PASS system and hope for the best. 2. Try to have Jim McNally negotiate a new agreement. 3. Inform TI that their proposed purchase agreement is unacceptable and therefore Atari has decided against using TI speech IC's in its products at this time. What should we do? _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: 4x4 Schematics Dt: 5/25/82 As per your request I have reviewed Bob Martin's schematics for 4x4. I have the following comments. 1. Depending on how far away the 11v is, the LM323 may need a capacitor on its input. 2. The 100 ohm resistor that is connected between the +11v and the TDA2002 amplifiers will limit the power available to the speakers to 4.5 milliwatts sine wave RMS. (For two 8 ohm speakers) 3. The TDA2002 has very poor power supply rejection. Normally, we operate them in a bridge mode and the power supply hum cancels. In this circuit they are single-ended and the power supply hum will appear in the speaker output. 4. The LM324 has much better power supply rejection. Whether it will give satisfactory results with the unregulated 11v will depend on how much ripple there is on the supply and on how well the AUDIO+5V is decoupled. (I don't see this decoupling circuit on the schematic). 5. Most people use at least some kind of filter after their Pokeys. The unfiltered square waves from the Pokeys will be particulary objectionable in the headphones. 6. Have you checked with the Safety Group on any safety requirements relating to the use of headphones? cc: Bob Martin _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Applied Research Project Dt: 5/27/82 As per your request this is a proposal to set up a research project for the purpose of investigating several areas of game design which are currently being neglected. The goal would be to produce a game. The areas to be explored will include: Game cabinet design with an emphasis on acoustics Control panel design Graphics techniques for Color XY Use of speech synthesis (maybe) By using Color XY and Warp Speed hardware we would be able to do the project almost entirely within the department. The project team should consist of the following: Project Leader - Rick Moncrief Engineer (and algorithms) - myself Programmer - Carl von Loewenfeldt Technician - Erik Durfey The other members of the department will be encouraged to contribute ideas. Scheduling - In order to know how long it will take to do something you have to know just what it is you are going to do. If you already know just what it is that you are going to do then you are not doing research. You do not wake up in the morning, look at the schedule, and declare, "Today I will be brilliant". My proposal is that things will be presented as they are developed. ------------------------------------------------------------- Other Problems - The programmers have reserved to themselves the right to design games. There is no chance for anyone else to design a game (except for Lyle). The reason for my request last week to set up a Warp Speed bench was so I could work out some graphics techniques that I am interested in investigating. In particular I want to do the game that I have already described as TOMCAT. It would be a terrific demonstration of some of the things that are possible with Color XY and MY hardware. I also believe it would be highly successful as a game. If this game is selected for the project and if it becomes a repeat of my Warp Speed experience I would be very disappointed. the end _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: TI PASS Purchase Agreement Dt: 5/28/82 I have attached a copy of TI's new proposed purchase agreement for the PASS system. Item 1 in my memo of 5/3/82 has been addressed. Item 2 has not. I'm not sure about item 3. It depends whether Iraj's cover letter constitues a binding modification as per Section 9 ENTIRE AGREEMENT. Since my last memo I have learned the following: The new Votrax chip won't be available until next year. National might be coming out with an LPC Speech chip with a development system but will not return my calls. Milt Loper has dropped plans for using speech in 4x4. Greg Rivera and Rich Moore appear to remain interested in speech. What do you want me to do? _____________________________________________________________________________ TOMCAT - WARP SPEED Conversion Plan of Attack - Security Clearance required: Red Level B Jed Margolin 7/7/82 Erik: Get WARP SPEED boards (probably 7/14) Harness/Control Panel (Use pot for thrust) Become familiar with the hardware Carl/Greg: Test Program RAM Test (Main and Math) Divider - use selected values 4000/4000=4000 5555/4000=5555 2AAA/4000=2AAA 5555/5555=4000 2AAA/2AAA=4000 Matrix Processor Initialise Math RAM Check for busy (reading or writing Math RAM generates a Halt) Run ROLL, wait 71 uS, check for done,compare results to known values Test DOWNLOAD Program Carl: Get copies of WARP SPEED program Bring up WARP SPEED Disable enemies Change YAW to ROLL Make stars flat (temporary terrain) Make screen horizontal Implement flight equations (I will add microcode as required) Possibly implement horizon line (I will add microcode as required) Add on-screen readout of airspeed, altitude, rate of climb, heading(?) Implement runway with control tower (and rotating radar antenna from Battlezone) Decide on Terrain generation Bring out other independent objects: Tanks from ABZ Doug's aircraft Saucer from Warp Speed Sounds(?) Speech(?) Begin game play _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Speech Synthesis Status Dt: 9/20/82 The T.I. Speech Development System is in-house and appears to work as advertised. Therefore, my part is done and I am ready to turn it over to whomever is designated to receive it. The Speech development system can analyze up to 12 seconds of speech at a time and when it is connected to a terminal the user can edit the speech parameters and make EPROM's. There is also an RS-232 port for uploading speech data (I have documentation on the protocol used). I recommend that this be connected to the VAX and that a VAX programmer implement a system for creating and handling Speech files. My suggestions are as follows: 1. A typical speech ROM will contain address pointers at the beginning of the ROM. Each word or phrase will have associated with it a starting address of data and an ending address of data. a). For 6502 it might be LO,HI. b). For 6809 it should be HI,LO. c). The addresses might be local to the ROM or be absolute in the address space of the game. d). After the look-up table comes the actual speech data. 2. When vocabulary is added to a file it should have associated with it a user supplied header consisting of 'Word', speaker, date recorded, recorded by (optional). 3. When a user assembles a vocabulary file for a game, he/she should be required only to supply a list of words. The program should automatically create the address look-up table and move the vocabulary data around appropriately. Then it should print a list of the words and their index numbers. (Either alphabetically or by index number or both.) 4. The Speech Development System requires an RS-232 Terminal (which are currently in short supply) and a room to put it in (for which I have some acoustic absorbing foam). It should be a room free of ventillation noise, or one where the fan can be turned off during recording. Acoustic noise can significantly degrade the speech quality that comes out of the system. 5. There are ways to manipulate the speech data to produce special effects. In a game with enough processing power it can be done in real time. Otherwise the data can be manipulated before being put into the ROM. I am willing to explain it to whoever does the VAX programming. 6. I propose sending this memo to Steve Calfee. - the end - _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: XY Monitor Array Dt: 9/22/82 To maintain the aspect ratio there must be a square number of monitors. The central part of the screen tends to have special significance so the minimum is 3x3=9 monitors. If you consider the middle row of monitors, the horizontal signal for the middle monitor is simply the XY signal x 3. The signals for the other monitors are derived from offsets and multiplies of the original XY signal. Note: The monitors must all be of the type with integral pincushion correction, or each monitor needs its own pincushion correction circuitry. Also, If the array were used with a game board with pincushion circuitry, the circuitry would either have to be defeated or the monitor array input would need an anti-pincushion section. Difficulties: 1. The vector drawing speed is effectively 3 x original speed, so the monitors must be able to handle the increased speed. 2. When the vector passes from one display to the next, the beam must be there, waiting for it. If the signals are just hard-clipped at the edge of each monitor, the power dissipation will be very high. This can be ameliorated somewhat by additional circuitry but the dissipation will still be pretty high, especially when the vector is at the intersection of four monitors. A memo on the raster monitor array is in the works. Another possibility is a box with a 13" color raster, a 13" color XY, a half silvered mirror, and a TV camera (preferably the Hitachi with the CCD imaging element). The signal would then be shown on a projection TV (which could still be used for other purposes). jed _____________________________________________________________________________ TO: Mike Fr: Jed Re: National 16000 family Dt: 11/10/82 It looks like something I could use (instead of the 68000). I am concerned with price and availability (especially of the 10 MHz part) and second sources. The software they demonstrated did appear to actually work. The symbolic debugger looks like it would be very useful. They should have made more of an effort to find out what business we are in and what processors we have used in the past. I don't particularly care that they have a version that executes 8080 code. I can use either the 16032 or the 68000. It really comes down to which one you guys want to support. By the way, how well suited is the 16032 to doing Forth? jed _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: New XY System Architecture Dt: 11/29/82 SUMMARY: I would like to pursue a new architecture for 3D XY games. It would use two 16 bit microprocessors to replace most of the MSI now used in Tomcat and would have a 6809 Auxiliary Processor for Sounds, Speech, I/O, etc., with the capability of linking two games together. The 16 bit processors would be either 68000's or 16032's (National) depending upon which one the Software People will support. I will need an appropriate development system for this project. I would like to start the circuit design in January, have a wirewrap in February, and have a demo ready during the middle of July ('83). I will do the programming myself. Potential uses include in a regular production game, a demo for the Marriot Arcade, or as a low cost Graphics Terminal. -------------------------------------------------------------------------- ARCHITECTURE Attached is a block diagram of a new architecture for an XY system. It uses two 16 bit microprocessors to replace most of the MSI now used in Tomcat. The first processor does the math calculations (rotations, translations, etc.) and the game play. The second processor is used as an intelligent vector generator, optimized for 3D projected objects. What that basically means is that the math processor hands to it an updated list of object points; the vector generator will know how they are to be connected. The third processor, the Auxiliary Processor, will be a 6809 and will handle Sounds, Speech, Switches, Pots, EAROM, Coin Counters, and Serial I/O. Because the Aux processor will be operating at a higher interrupt rate than is feasible in existing single processor games, two games will be able to talk to each other and exchange enough information to do a two person Tomcat game. ---------------------------------------------------------------------------- DEVELOPMENT SYSTEMS The 16 bit processors can be either 68000's or 16032's (from National). The selection will be determined by which one the Software People decide to support. In processing power, a 6 MHz 16032 seems to be about equivalent to a 10 MHz 68000. However, the 16032 has a clear advantage in that its architecture makes it possible to actually know which instructions have been executed rather than just which ones have been prefetched into the pipeline. The Graphics Terminal Project's 68000 Development system (the HP 64000) is not available to me under conditions that I can accept: I could use a development station if they have one available and I would have to relinquish it when they needed it. The in-house 68000 game interface board appears to have gotten exactly nowhere. National makes a development system for the 16032 which operates in a manner similar to what we already do. The VAX is used to edit and assemble programs; the object code is downloaded into a box out of which comes a pod which plugs into the game's 16032 socket. A VDT plugs into the development system for talking to the VAX and patching, running, and debugging the program. The cost would be $16K - $20K. ------------------------------------------------------------------------------ SCHEDULE I would like to begin the circuit design in January and be ready to begin a wirewrap board by the beginning of February. (I would need to know which processor will be supported before I can do much circuit design.) I would need the development system by the middle of February. I anticipate doing the preliminary programming myself to demonstrate the usefulness of the system. Once the development system is usable I would expect to take 2 weeks to debug hardware, 6 weeks to program the Intelligent Vector Generator, 6 weeks for the Math Processor, and 6 weeks for the Aux Processor. ------------------------------------------------------------------------------ POTENTIAL USES It could be used in a regular production game, with the capability of linking two games together. This capability has been discussed for the last several years but has not been feasible due to the 4 ms. interrupt rate; the amount of information transferred at a rate of 1 byte every 4 ms is not enough to link games. (The Aux Processor will operate at 1 ms interrupts and that should do the job.) This architecture could be used with the powered moving seat to do a demo for the Marriot Arcade, especially with two games linked together. The serial interface would make it feasible to do a low cost Graphics Terminal the end _____________________________________________________________________________ To: Dan Van Elderen and Rick Moncrief Fr: Jed Margolin Re: Speech Recognition Dt: 12/10/82 There is currently no work being done by Coin-Op in the area of Speech Recognition, although Tom Hogg has expressed an interest in it for the Graphics Terminal Project. If we do become involved in Speech Recognition it probably won't be with Verbex; their systems cost $65,000 ($6K-$10K for just the boards). ------------------------------------------------------------------------------ (Additional material which may be of interest) In order for a Speech Recognition System to be used successfully in a coin- operated video game I would suggest the following requirements: 1. Produce a 99+ % correct response which must be independent of player gender and/or dialect (assuming only American English), with no player training, and with probability of considerable noise from other games as well as the player's own game. 2. Cost less than $100. (and this figure is probably high). There are currently no systems available which even come close to meeting these requirements: 1. The single chip devices from Weitek and Interstate (about $10 in quantity) work poorly, if at all, in the presence of noise. 2. Even the expensive systems require that the user wear a microphone headset like the ones telephone operators wear, which would not be practical in an arcade video game. 3. Matsushita supposedly has a speaker dependent system for $350 which is user trainable; chip sets will probably be under $50 . 4. Texas Instruments has done substantial work in Speech Recognition (as well as synthesis) and may bring something out in '83 using the TMS320 Digital Signal Processing chip. 5. There was a meeting on Thursday 12/9 between Weitek and representatives from various Atari Groups (hosted by the Atari Semiconductor Group). Weitek has decided that it can no longer afford to work on speech recognition and wants to either sell off its expertise (and several projects in varying states of development) or form some kind of joint venture to continue the work. The project with the most promise (and which is 80% complete) was for a speaker dependent system where the user trained the system and which would probably cost less than $20 to produce. 6. General Instrument claims to be committed to Speech, and is apparently working on Speech Recognition. Other products may have different requirements: 1. The Graphics Terminal Project could probably tolerate 97% correct response with user dependence. On user dependent systems there is a learn mode where the user speaks the words which the machine will later recognize. 2. For similar reasons Speech Recognition might be practical as part of a Personal Computer System. 3. If Atari goes into the telephone business, it would be interesting to have a telephone with Speech Recognition instead of buttons. Also, frequently called numbers could be referenced by a name instead of the user having to write down and use an index number. 4. I would like to have a VCR where I could set the Timer schedule with Speech Recognition rather than funny little buttons. _____________________________________________________________________________