Atari/Atari Games Memos and Status Reports 1983 Jed Margolin ___________________________________________________________________________ GAME HARDWARE IDEA - NEW 3D COLOR XY Jed Margolin 1/25/83 The following is a description of the hardware architecture I am plannning on designing for the Tomcat research project. I am submitting it here because the hardware may be applicable to other games. There are 3 basic sections: The Host Processor; The Vector Generator; and the Aux Processor. 1. The Host Processor will be a 16032 or 68000 and will be supercharged with the Warp Speed Matrix Processor. (Perspective division will be done by the Host.) The Host Processor will do all the 3D math and deliver to the Vector Generator an updated point table for each object. 2. The Vector Generator will probably be similar to the analog type currently being used; however, it will contain the algorithms to connect the points for each object so that the Host Processor will not have to spend time formatting vectors. 3. The Aux Processor will be a 6502 or a 6809 and will do: A. Sounds B. Speech C. Coins D. EAROM stuff E. Pot reading with A/D Converter F. Communication between games _____________________________________________________________________________ DATE: 1/25/83 SUBMITTED BY: Jed Margolin NAME OF GAME: YourFaceOnthePlayer BASEBALL OR MAYBE FOOTBALL PERSPECTIVE: 3rd Person HARDWARE: Color Raster, Macho 44, Speech Synthesizer, CCD Image Sensor OBJECT OF THE GAME: Baseball, with two improvements: 1. At the beginning of the game, a CCD image sensor will take the player's picture and put his face on all of his team members. If the player makes the high score table there will be no initials, just players' faces. 2. A speech synthesizer (or two) will provide play-by-play and color commentary of the game taking into account how well the player is doing. ------------------------------------------------------------------------------ DATE: 1/25/83 SUBMITTED BY: Jed Margolin NAME OF THE GAME: BETTY BOOP PERSPECTIVE: 3rd Person HARDWARE: Color Raster, Macho 44, Speech Synthesizer or Video Disc OBJECT OF THE GAME: The player leads Betty through a series of adventures, with her talking and exclaiming Boop-Boop-A-Doop at appropriate times. ------------------------------------------------------------------------------ DATE: 1/26/83 SUBMITTED BY: Jed Margolin NAME OF THE GAME: COCKROACH WARS PERSPECTIVE: 3rd Person HARDWARE: Color XY, or High Resolution Raster OBJECT OF THE GAME: This is a two player competitive game where each player controls an animated, articulated cockroach. The action takes place on several scrolling playfields: UNDER THE KITCHEN SINK, UNDER THE REFRIGERATOR, IN THE PANTRY, and IN THE BATHROOM (don't get trapped in the bathtub). As the two cockroaches fight they have to avoid periodic natural disasters such as a human exclaiming, "Look Fred, we have cockroaches, again." The players would then have a limited amount of time to scurry to safety. Perhaps we could use a CCD imaging device to put the player's face on the cockroach. If cockroaches are too distasteful, use mice. _____________________________________________________________________________ RE:LPB0 I left the listings in the basket because I felt it was the one place where they would NOT mess things up. But if you want them ON the printer then that is where they will go. And while we're on the subject, last week I tried to help a SCI-CARDS person cancel a duplicate listing.(It seems that when his listing did not appear in WESTPRINT he figured something had gone wrong and ordered another listing.) Because the directions for cancelling a printout were so thoughtfully provided on the printer, he and I tried them. They didn't work. JM _____________________________________________________________________________ To: Ken Renda Fr: Jed Margolin Re: ECR 03727 Dt: 2/14/83 1. Per your request that "Resistors R13 and R14 should be moved so as not to short against ground trace (rear of board)." On the board I was shown the resistor leads were excessively long, causing them to short out when they were bent back. But If Manufacturing reallly thinks the traces should be moved, then ok. 2. Per your comment that the "Trace from collector of Q3 to cathode of CR4 is being blown off board resulting in severe damage to PCB." On one board that I examined, CR1 and CR4 were both shorted, causing a short from 10.3 VDC to ground. On another board CR1 was shorted and CR4 apppeared to have been replaced, the original CR4 presumably having been shorted also. CR1 and CR4 are both protection diodes and do not normally conduct any current unless they are installed backwards, which was NOT the case in the boards I examined. I recommend we try to trace these diodes and find out if we had a shipment of bad or incorrectly marked 1N4002's. The only other explanation for these failures would be if the boards had somehow been connected to the supply backwards. Jed _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Dt: 3/2/83 Re: Reasons for Buying more CIT-101 Terminals The department currently has five CIT-101 terminals: Carol has one at her desk for word processing; Greg and Norman each have one in their offices; There are two in the Hanger Lab, each associated with a development system (both are for Star Wars). There are currently no terminals available for general use. Those of us who need to use the VAX have to squeeze in time on one of the above systems. I use the VAX frequently for programming and for generating and maintaining documentation. Earl also uses the VAX extensively and in fact should have one allocated for his exclusive use as have Greg and Norman. Rick Meyette uses the VAX for word processing. Tom Rice has expressed an interest in computerizing parts lists, inventories, etc. I am trying to talk him into using the VAX rather than buying another Atari 800 system. I propose that we purchase three CIT-101 terminals. One will go to Earl and two will be installed in the labs for general use. Alternately, one would go to Earl, one would go to me (in my office) and one would go to the lab for general use. Jed _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: "Video-game makers rush to add 3-D chip" Dt: 3/7/83 Attached is a copy of an article that appeared in the San Jose Mercury News, March 2, 1983. The chip being touted as a magic chip to do 3-D is the 2516 16*16 multiplier. I talked to Dave Barnet of Weitek last year when it first came out and again this afternoon to update my information. The 2516 is an NMOS version of a TRW Bipolar Multiplier. Because it is NMOS it dissipates less power and so can be packaged in plastic. It also runs slower than the TRW part. Last year the 2516 was being offered in a 1 us version for $68. This year it is a 200 ns part and goes for about $40 in quantities of 5,000. (The TRW Bipolar Multiplier does a multiply in about 40 ns and costs twice as much.) The 2516 was not cost effective last year at $68 and it is not cost effective this year at $40. The multiplier/accumulator that I use in Star Wars is made out of MSI and costs about $12 (and does more than just multiply). The basic operation in doing 3-D transformations is to accumulate products such as in ACC'= ACC + A*B. The 2516 does not accumulate products, it is only a multiplier. (Weitek has expressed an interest in working with Atari to produce a low power multiplier/accumulator.) My Multiplier/Accumulator also does a presubtract to make calculations more efficient [ ACC'= (A-B)*C + ACC ]. On the other hand, my multiplier is slower (2.7 us vs 200 ns for the Weitek). Raw multiplier speed is only part of story in doing 3-D transformations. In performing the 3-D math, lots of data has to be fed into the multiplier, and results retrieved, and additions performed (if the multiplier does not also accumulate products). If a device such as the 2516 were put on an 8 bit microcomputer bus, even if the multiplier took zero time it would be several times slower than the current Star Wars hardware. (A 2516 put on a 10 MHz 68000 or 6 MHz 16032 would probably be nice and fast, but a multiplier/accumulator would be better). The Star Wars hardware deals with the throughput problem by having a simple microprogrammed device keep the multiplier/accumulator fed. The points that define an object are downloaded by a 6809 host processor at the beginning of the game (or wave) into a RAM it shares with the microprogammed machine (the Matrix Processor). When it comes time to display an object, the host processor supplies a pointer which tells the matrix processor where the object's points are stored. The host processor tells the matrix processor which algorithm to use, and each of the object's points is transformed, one at a time. I resent having to generate this memo, having to defend hardware that I designed THREE YEARS AGO. When the project was Warp Speed it received no support from management. Now that it has become Star Wars and Atari has paid George Lucas one or two million dollars there is more interest in the project. However, the game that is being developed does not take advantage of the real capabilities of either hardware or algorithms. This is very discouraging and makes it very difficult to maintain any enthusiasm in a subject area for which management obviously has so little regard. And I don't want to hear any more crap about Weitek's great 3-D chip. Jed _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Votrax SC-02 Dt: 3/8/83 Summary: The SC-02 doesn't sound as good as the TI system we are now using. We would lose the ability to have the speech recognizable as belonging to a particular person (like Luke Skywalker). The development tools available for the SC-02 are much cruder than what we already have for the TI system. The information I have on the Votrax SC-02A comes from the following sources: A Concept Definition of the SC-02 in a memo from Steve Bristow dated 5/4/81; A memo from Steve Bristow summarizing the meeting with Votrax on 6/5/81; Preliminary Specification of SC-02 in a letter from Al Lubienski (Votrax General Manager) to Michel Ebertin, dated 9/1/81; A sample tape from Votrax demonstrating the SC-02; A conversation with Randy Hall (of VCS fame) on 5/4/82; This is what I know: My notes indicate that I thought the sample tape made with the SC-02 sounded better than the SC-01 but not as good as the TI synthesizer. It does not appear that the SC-02 can be programmed to be recognizable as belonging to a particular person. (The sound effects weren't bad, though). The SC-02 is an improved version of the SC-01. The phonemes have supposedly been improved and the user can control more of the atttributes of the speech. The tape mentions that the atttributes could be extracted from recorded speech. If all of the attributes are used, the bit rate is 240 bits/sec (versus 1200-1800 bits/sec for TI). If the attributes are not used the bit rate is closer to 70 bits/sec and it ends up sounding just like the SC-01. The schedule Randy gave me was as follows - samples - Christmas production - 1st Q 83 Development Station - 1st Q 83 ($15K) (does not extract features from recorded speech) Estimate 100K - $10. 1M - $5. In order to fully utilize the additional capabilities of the SC-02 we would have to hire a linguist to develop vocabulary and/or we would have to design a unit that analyzes recorded speech and extracts the required attributes from it. I do not think a person who is not a linguist could sit down at a VDT and create acceptable vocabulary for the SC-02. If Votrax has made really significant changes since last May, my information may be out of date. The last I heard from Dave Remson (VCS) was that General Instrument was developing for them a two channel speech synthesizer based on the GI SP0250. The SP0250 would have been my first choice for Coin-Op to use if it hadn't become apparent that GI was unable to provide the necessary support for us to use it successfully. If they are successful in developing it and providing decent development tools we may want to switch over to it at some future time (and take advantage of VCS quantity discounts). If this deal with GI is still in the works why this sudden (and misplaced) interest in Votrax? Jed _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Dt: 4/16/83 I amm writing this memo on Saturday. I came in at about 1:45p to work on Star Wars. Facilities chose today to wax and polish the floors in our labs. Equipment has been disconnected, moved around this way and that, Erik's radio has been turned on full blast, and the Star Wars game is no longer working. I called Security and talked to Barbara (whose supervisor's name is Dennis Johnson) who said it was a matter for Facilities, not Security. She said she would try to have someone from Facilities paged but was apparently unsucessful. The guys doing the work here said they worked for Gavaer Ramerez. When I left here at 3:30 am last night everything was intact, and working. This is not the first time that Facilities has shown a callous disregard for projects that I have been working on. On several occasions they have moved prototype games (with circuit boards and cables trailing behind) to replace stained ceiling tiles. Presumably they do this to other projects as well. And just when I was starting to work on the game I was asked to leave so they could put on the wax on the floor. I request that you (and Dan, and John, if that is what it takes) TELL Facilities that under NO circumstances are they to move ANY of our equipment without PRIOR approval. Getting Star Wars into production is already tough enough without this SABOTAGE from the Facilities Department. CC: John Farrand Mike Halley Dan Van Elderen _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Microprocessor Test Connector and Darl Davidson Dt: 3/29/83 Manufacturing does not use the Microprocessor Test Connector. Repair Technicians in the field do use it because that is how the CAT Box connects to the game board. -------------------------------------------------------------------------- What the CAT Box Does: The CAT Box gives repair technicians (in the field) access to the processor's bus and enables him to test for open or shorted address and data lines, read switch inputs, turn outputs on and off, test RAM, and perform Signature Analysis. When people use the CAT Box to perform signature analysis on game ROMs, the CAT Box connects a counter to the address lines and simply runs through the addresses. By using a ROM Select to start and stop the Signature Analyzer, a signature is taken for the bit pattern in that ROM, thus establishing whether the ROM is good. This is usually done just for the ROM containing the reset and interrupt vectors and test routines; if this ROM is good the game will probably be able to run Self-Test. Because the address decoders are not strobed but are free to glitch until the address is valid, it has been necessary, in order to use the CAT Box Signature Analyzer, for the technician to attach capacitors to a few address lines to remove these glitches. ---------------------------------------------------------------------------- All the pins that appear on the Test Connector also appear on the microprocessor socket, so my idea was to do an interface board so that people would remove the processor (which they must do anyway) and plug in this adapter (which would connect to the CAT Box). Also on this interface board would be a strobed latch so that Signature Analysis could be performed without using kludge capacitors. ---------------------------------------------------------------------------- I talked to Darl Davidson and Frank Becker on 3/28 about this and they were adament against doing this. Their reasons were: 1. Taking off the Test Connector would be perceived as "Atari cheaping out". 2. It would require twice as many insertions and removals in the processor socket. We use cheap sockets which are supposedly good for 10 insertions. 3. They were concerned that the people who needed the adapters wouldn't buy them and would become very angry when they couldn't use their CAT Boxes with the game. 4. Their experience with adapters was that adapters don't work. They referred to a Z80 adapter that someone is alleged to have perpetrated. I was not aware that our sockets were that poor. On the other hand, I learned that the kludge capacitors are usually stuck into the processor's empty socket. That can't be very good for the socket, considering the diameter of most capacitor's leads. The other reasons might be valid but probably are not. --------------------------------------------------------------------------- I have an even better solution. When the 6809E HALT line is brought low, the processor finishes the current instruction and halts with Address, Data, and R/W lines TRI-STATE. Therefore, with the HALT line low, the CAT Box could be connected to the game without removing the processor (unless there was some reason to believe that the processor was bad and wouldn't go Tri-State). This could be accomplished in two ways: 1. HALT would be automatically grounded with the CAT Box plugged into the current Test Connector. 2. An IC clip (similar to the ones we use on the HP Logic Analyzers) would clip onto the processor, grounding the HALT line. This cable would have to go to a small interface board which would accept the CAT Box connector. If this is done, there will be no place to stick the kludge capacitors. If we use method 1, we could put in the traces for a latch and leave it unstuffed. If someone wants to do signature analysis they will have to solder one in (talk about cheap). We could, instead, stuff a socket, or probably for the same cost, stuff the IC (a 74LS74). I do not have any unused latches on the Star Wars Main Board. If we use method 2, a strobed latch can be included on the interface board. I prefer method 2. ---------------------------------------------------------------------------- I told Darl that I was willing to keep the Test Connector on the Main Board but that I would not put one on the Sound Board. He seemed satisfied with the arrangement. In any event, I strongly disapprove of using kludge capacitors, especially when they are shoved into a microprocessor socket. If we go with method 1 I want to put the extra LS74 on the Main Board. If we go with method 2 it would go on the Interface Board. ---------------------------------------------------------------------------- My next step is to have Erik put together a CAT Box Interface Board so we can demonstrate its advantages. I wanted you to know what I am doing in this matter because Darl has strong feelings about the Test Connector and has threatened to go to Dan and John over it. Jed ___________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Speakers Dt: 5/10/83 The following applies to the TDA-2002s as they are used in the Regulator/Audio I and Regulator/Audio II boards. All power ratings assume an unclipped sine wave. SUMMARY: When used in a game that produces a single audio channel the power amplifiers are usually connected in push-pull and will deliver 12.5 Watts into a 4 ohm load. This can be a single 4 ohm speaker, two 8 ohm speakers in parallel, or an array of four 4 ohm speakers. Games with a single audio channel tend to use two 8 ohm speakers. In a game that produces two audio channels, each amplifier must have a 2 ohm load if the total output is to remain 12.5 Watts. Star Wars uses two audio channels (because of the Image Expander) but only uses two 4 ohm speakers and therefore its total output is 6.25 Watts (3.12 Watts per channel). Ideally it should have had four 4 ohm speakers or two 2 ohm speakers. If each amplifier were driving an 8 ohm speaker, the power would drop to 1.5 Watts per channel (for a total of 3.12 Watts) and it would probably not be loud enough in an arcade environment. If only 8 ohm speakers were available, the Star Wars Image Expander could not be used. If only 4 ohm speakers were available, a single channel game would be limited to 6.25 Watts instead of the 12.5 Watts currently available with 8 ohm speakers. Standardizing on speakers is a bad idea. ----------------------------------------------------------------------------- BACKGROUND: When operated from a 10 Volt supply, a TDA-2002 will deliver 6.25 Watts (sine wave) into a 2 ohm load or 3.12 Watts into a 4 ohm load. With two TDA-2002s, the amount of power that can be delivered to a load is 2x6.25 = 12.5 Watts (sine wave). This is with both amplifiers driving 2 ohm loads. At this point you might object, "But we don't have any two ohm speakers." This is the answer: When a game provides only one audio channel, the two TDA-2002s are usually operated in push-pull with the speaker connected across the amplifiers outputs. This doubles the output voltage across the speaker, which multiplies the power by four. (P=V*V/R). If we take the example of a single TDA-2002 driving a 4 ohm speaker we get 3.12 Watts. If we run two TDA-2002s in push-pull we will get 3.12x4=12.5 Watts. The quadrupling of power is not free. The current required from each amplifier is doubled. The main effect is to "split" the load between each amplifier so that each amplifier sees R/2. Thus in a push-pull system a 4 ohm speaker will be seen by each amplifier as a 2 ohm load, and the TDA-2002s will deliver 12.5 Watts to the speaker. If the designer wanted to use two speakers and put two 4 ohm speakers in parallel, the result would be a two ohm system which each TDA-2002 would see as a 1 ohm load which it couldn't drive effectively. If the 4 ohm speakers are put in series, the result is an 8 ohm system which each TDA-2002 would see as a 4 ohm system. That's fine except we now have each amplifier driving a 4 ohm load and delivering 3.12 Wattts for a grand total of 6.24 Watts. This decrease in power would be offset somewhat by the increase in speaker efficiency produced by having a greater cone area, but there is a better solution. If we take two 8 ohm speakers and put them in parallel (making sure they are phased properly) we get an effective system load of 4 ohms, which each TDA-2002 will see as 2 ohms. The amplifiers will therefore deliver 12.5 Watts to the system. If we wanted to use four speakers (like Space Duel) they would be 4 ohm speakers in both series and parallel to again produce a 4 ohm system. Which is why Space Duel is so loud. So, it seems we can standardize on either four 4 ohm speakers or two 8 ohm speakers in every game. Wrong. Games with a single audio channel have tended to use two 8 ohm speakers. Star Wars (with fake stereo) is using two 4 ohm speakers. Ideally it should have had four 4 ohm speakers or two 2 ohm speakers. Star Wars uses a stereo synthesizer so that each TDA-2002 must must used separately; they cannot be connected in push-pull. The upright game originally used four 4 ohm speakers connected to produce two arrays of 2 ohms each. This would have produced 12.5 Watts of audio. Unfortunately, in a futile attempt to hold the cost down, two speakers were eliminated. Thus, Star Wars is limited to 6.25 Watts of audio. With 8 ohm speakers the system would deliver 1.5 Watts/channel, for a total of 3.12 Watts. The sitdown game currently uses four 4 ohm speakers but it may be necessary to replace the upper channels with 8 ohm speakers to give a better balance between the upper and lower speakers. It is too soon to tell if fake (or real) stereo is the wave of the future. But standardizing on either 4 or 8 ohm speakers would be a mistake and would send a clear message to those who have been working in this area that there is no support for audio in this company. __________________________________________________________________________ To: Chris Downend Fr: Jed Margolin Re: Power Supply and Audio/Regulator Specs Dt: 6/3/83 Chris, Assuming this is for Star Wars, there is only (1) Audio/Regulator Board per game. These are the maximum ratings for the power supply. Star Wars does not draw the maximum current ratings. Jed ------------------------------------------------------------------------ POWER SUPPLY Transformer-based design capable of delivering: +10.6 VDC @ 8 Amps 6.1 VAC @ 2 Amps 50 VAC CT @ 2.5 Amps 36 VAC CT, specified so that after rectification it will deliver: +21 VDC @ 1.5 Amps -21 VDC @ 1.0 Amp AUDIO/REGULATOR PCB Audio: (2) TDA2002 audio amps, each capable of delivering 6 Watts into 2 ohms Regulators: +5 VDC @ 6 Amps +12 VDC @ 1 Amp -5 VDC @ 1 Amp ____________________________________________________________________________ To: Alan Nusbaum Fr: Jed Margolin Re: PMI DAC-312 Dt: 6/29/83 According to the data sheet, PMI's DAC-312 should perform as well as AMD's AM6012 in all the parameters that are relevent to the Analog Vector Generator. Erik and I have tried the nine DAC-312 samples in a Star Wars AVG Board and they all appeared to work ok. This kind of test should not take the place of a full scale component investigation, but it appears we don't have the equipment to fully test DACs. cc: Dean Charrier Bruce Hartman Rick Moncrief ---------------------------------------------------------------------------- To: Dean Charrier Fr: Jed Margolin Re: Motorola Dt: 6/29/83 According to the data sheet, Motorola's 74LS385 should be equivalent to the TI part. Erik and I have tried a tube-and-a-half of Motorola 74LS385s in a Star Wars Main Board and they all worked ok. cc: Rick Moncrief _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Cinematronics Video Disc Game Dt: 7/13/83 I had the opportunity to examine Cinematronics' new Video Disc Game at Central Park and noticed the following: 1. The game did not have an FCC sticker. 2. There was NO video overlay. The picture on the screen came solely from the Video Disc. Scoring, credits, and lives were done on a separate LED panel. I didn't see any high score table. 3. The game logic was on a small board and had several ROMs, a processor (I couldn't see whose it was), a GI Sound IC, maybe a separate processor for sounds, two audio amplifiers, option switches, and the usual assortment of other stuff. 4. Even though it had the sound IC, I cannot recall having heard it in the game. It sounded like all the audio came from the disc player. 5. The Video Disc player was a standard unit, sitting on the bottom of the cabinet. I couldn't see the manufacturer's name on it. The digital interface just plugged in; it probably came with the unit rather than have been kludged on. 6. When the game was turned on, there was no color and there was a noticeable jitter in part of the picture. The attendant said that the game had always had that problem. When we turned the game off and then on again, when the disc came up to speed the color worked and the jitter was gone. When we moved the game the color went away again but came back after a disc player restart. 7. The screen blanks out for varying periods of time during disc track seeks, but I did not find it objectionable for the amount of time I was with the game (about 40 minutes). 8. In normal game play, if you get killed you go on to the next scene. In the unlimited-lives-mode, if you get killed you start that same scene over, presumably so you can figure it out. interface 9. The production values of the animation are high quality, but I do not care for the death sequences, especially the ones involving death by fire. 10. The game used an Electrohome monitor with a composite video adapter. cc: Dan Van Elderen _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Game Serial Communication Dt: 7/13/83 If we can run the serial communications at 19.2 KBaud, 19.2 KBaud/11 bits= 1750 bytes/sec, not including housekeeping. Assume 1000 bytes/sec. If the update rate is 50 mS, 1000 bytes/sec * 50 mS = 50 bytes of communications/ update. Using a straightforward approach, nothing fancy, the player's unit vectors = 9 words * 2 bytes = 18 bytes the player's position = 3 words * 2 bytes = 6 bytes fire button, collision detects,etc = 10 bytes, max ______________________ = 34 bytes The serial I/O will be done by one of the Pokeys on each game's Sound Board. Each game can receive and send simultaneously. _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Star Wars Description Dt: 7/20/83 Specifically: 1. The game can produce 7 colours with 255 levels of brightness. 2. The game uses: One 68B09E Game CPU. One 68B09E Sound CPU. One Custom MSI 3D Math Processor. One Custom MSI Display List Processor. 3. There are 12K bytes of Vector Display RAM (and 4K bytes of Vector Display ROM). 4. I do not know what the highest score achieved in the marketplace or in-house is. Technically: The hardware is on three boards: the AVG Board, the Main Board, and the Sound Board. The AVG Board contains the Analog Vector Generator which draws the lines on the screen. It has 12K bytes of Vector RAM and 4K bytes of Vector ROM. The Main Board contains the 68B09E Game Processor, 48K bytes of Program ROM, 2K bytes of Program RAM, option switches, means for reading control panel switches and pots, a pseudo-random number generator, a digital Divider, and a Matrix Processor which performs the math for the 3D graphics. The Sound Board contains a 68B09E, 16K bytes ROM (for the program, sound tables, and speech vocabulary), a 6532A (with 128 bytes of RAM, two bidirectionmal I/O ports, and an interrupt timer), 2K RAM, a speech synthesizer, four Atari Custom Sound ICs, a stereo image synthesizer and an interface to the Main Board. Self-Test (selected by the Self-Test switch on the operator panel) tests as much of the hardware as possible. There is also a hardware diagnostic mode which permits the repair technician to select specific routines to aid in troubleshooting. Additionally, kindly tell Lyle: 1. I do not use Bit Slices in Star Wars. 2. I DO NOT use Bit Slices in Star Wars. 3. I DO NOT USE BIT SLICES IN STAR WARS. ____________________________________________________________________________ To: Mike Hally Fr: Jed Margolin Re: Star Wars Power Supplies Dt: 7/26/83 Two Star Wars production power supplies were brought to me to find out why they did not work; even with no load they got really hot. In the unit I have tagged as #1, the transformer harness going to J3 (the voltage selection connector) has the Red/Black and Brown/Black wires reversed so that J3 pin 1 and pin 2 are reversed. This was verified by an ohmmeter check. With the 120VAC voltage selection block in place the result is effectively that part of the primary winding is shorted. In unit #2, the wire colors are correct, but an ohmmeter check reveals that J3 pin 1 and pin 2 are reversed. Thus, the problem is a wiring error inside the transformer. An additional problem with unit #2 is that the mounting bracket for C1 (the 27,000 uF filter capacitor) has been tightened so much that a section of the capacitor has been caught in the clamp and bent. This is not nice. Both of these units use Ravenswood tranformers. I am concerned that because the transformer stack height is so short these units may not meet the specifications for Color X-Y transformers. I would like a production Color XY Power Supply in order to test the transformer temperature rise in a Star Wars Game, and I would also like Components to perform a full test on one. cc: John Aubuchon Dean Charrier Rich Cocciolone Chris Downend Mike Laschinger Pat McSweeney Rick Moncrief Alan Nusbaum _____________________________________________________________________________ To: Dean Charrier Fr: Jed Margolin Re: RAMs for Star Wars Dt: 7/26/83 I have tried the RAMs you gave me by substituting them in the Star Wars Math RAM positions. The RAMs were: AMD AM9128-15DC 8217 AMD AM9128-15PC 8316 Hitachi HM6116P-3 2B1 Q Unfortunately, they all work, so this test is not meaningful. cc: Rick Moncrief Alan Nusbaum _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Star Wars RAMs at 5F, 5H Dt: 8/25/83 Summary: On the boards that I received, the problem with the RAMs at 5F and 5H were all solved by connecting a 180 Ohm 1/4 Watt 5% carbon film resistor from 9D pin 2 to ground. There is a convenient trace on the top of the board to attach it to. ----------------------------------------------------------------- On Tuesday, August 23, Mike Laschinger told me about the problem they have started having with the Star Wars RAMs in locations 5F and 5H. These RAMs are used by the Matrix Processor. My understanding is that in about 10 percent of the boards, Self- Test says that the RAMs at 5F and 5H are bad, but in most cases the game runs ok. Changing the RAMs sometimes fixes it and sometimes does not. This problem coincides with the introduction of the REV C boards. Mike took some RAMs that failed Self-Test in a REV C board and put them into a REV B board and apparently they worked. I asked for a sample bad board and an unstuffed REV C board and received it on Wednesday. On the board I received, 5F and 5H failed during Self-Test and the game didn't work. On Wednesday I did some tests but didn't get anywhere. On Thursday I had Erik put it on a development system and I tried some things. On Thursday afternoon I got Greg and Norman involved. The RAM test worked when the program was running on the development system, but not with the board stand-alone. While running hardware diagnostics I discovered glitches on the WP strobe in the Matrix Processor clock generator, which went away when I put a 'scope probe on the 12 MHz clock input at the LS161 at 9D pin 2. It appears that the noise and glitches on 12 MHz (at the end of the line) were screwing up the Matrix Processor clock and probably causing it to extraneously access memory. On the boards that I received, the problem with the RAMs at 5F and 5H were all solved by connecting a 180 Ohm 1/4 Watt 5% carbon film resistor from 9D pin 2 to ground. There is a convenient trace on the top of the board to attach it to. I do not not know why this problem has suddenly appeared on REV C. We did not change the circuitry connected to 12 MHz. Perhaps we are using a new batch of 74S04 inverters in the 12 MHz clock (at 1N). _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Stanford Instructional Network Dt: 9/8/83 At your request on Tuesday I went with Herb to the meeting concerning the Stanford Instructional Network. The situation is that another department has signed up for the service and paid the $5K or so for a one year trial. The problems are: 1. The equipment costs $14K. 2. We would need to set up several classrooms (for which we have no space). 3. The fees are enormous: Honors Coop Program $419/unit Non-Registered Option $229/unit Audit Option $78/unit The feeling at the meeting was that managers might not realize that the fees had to come from their OWN department's education budgets and that very few managers had an education budget large enough to support the potential demand. 4. The Autumn Quarter starts September 28. Stanford has kindly extended the registration deadline for Atari to Sept 13. The Honors Coop Program is for full credit toward an advanced degree. It requires formal admission to the Stanford graduate school, which requires a transcript and taking the GRE. It is unlikely that anyone starting now could make the deadline. For the Non-Registered-Option students must still submit transcripts, but do not need the GRE. Students take all the exams and receive grades. Up to 18 credits can be transferred to the Honors Coops Program (if the student manages to get admitted to the program). It would be difficult to be admitted to the Non-Registered-Option at this late date unless the enrollee happens to have a transcript lying around. And had a 3.2 (or better) Grade Point Average. The Audit Option requires no qualifications, no homework or exams, and no credt. A group in 1196 is buying the equipment. If we wanted to set up a unit here, we might not be able to get it installed in time. I have a copy of the course descriptions if you want to see it. This could have been a good idea if it had been better organized. Jed _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Another Idea Dt: 9/13/83 I think there might be a large market for a new personal computer designed around the NTSC interface. It would have: 16032 or 68000 processor [with an additional 6502 perhaps it could be made compatible with existing Atari 800 software] 256K byte (or more) RAM Slots for ROM cartridges Disk (memory) interface [maybe for hard disk] Peripheral interface NTSC interface If video disc games are going to be popular as coin-op games, they will also be popular for home units. The NTSC interface will also allow game video to be mixed with the output from a standard consumer camera. This might allow some kind of innovative game as well as serve as an extremely versatile titling generator. It is this last capability that would have the greatest appeal. People would have at their disposal all of the advanced editing and storage (as well as advanced graphics) of the computer. Jed cc: Mike Albaugh _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Latest Star Wars Problem Dt: 9/16/83 On Thursday (9/15/83) Mike Laschinger gave me a board set that was representative of a problem that they have started to notice in Production. The symptom is that sometimes after power-up, the game comes up with no stars. The game plays; there just are no stars. We were not able to make the problem appear on our test setup. We took the board set back to Production and put it on a CTS fixture and the problem did appear. (Unfortunately, Mike says it sometimes happens in assembled games.) The reason there were no stars was that the Pseudo-Random-Number Generator (ICs 4D, 5B, 5C, 5D) had gotten hung up in the invalid state of all ONES. The Pseudo-Random-Number Generator has a CLEAR input which is cleared during RESET (both power-up and Watchdog Reset) and which is also under program control. I don't know why it sometimes fails to clear during power-up, but when I put the circuit in, I told Greg to do a software clear at the beginning of each game. Instead, he does a software check which is supposed to reset the Pseudo-Random-Number Generator if two consecutive numbers are the same. Due to a coding error, this feature does not work. On Friday, Mike's board set exibited the problem in the lab. Once. I talked to Frank Becker and he said it hasn't been a significant problem, yet. There are several options: 1. Find out why the Pseudo-Random-Number Generator sometimes fails to be cleared during a power-on Reset. (A manual Reset does clear it.) 2. Suggest a hardware fix involving cutting a trace and adding a jumper. 3. Release a new ROM set with the bug fixed, and send out a technical note to the operators. 4. Do nothing. How do you want me to spend my time? cc: Mike Halley Mike Laschinger Greg Rivera _____________________________________________________________________________ To: John Aubuchon Fr: Jed Margolin Re: TI Speech Synthesizer Problems Dt: 9/27/83 On Friday, Sept 16, I spoke to Keith Barber from TI Marketing and Dave Watkins from TI Qualtity Control about the potential problems with the TMS-5220 Speech Synthesizer IC used in the Star Wars game. Their story is as follows: 1. They were doing some tests on the TMS-5220s and found that after 1000 hours at 85 degrees Celsius and 85% Relative Humidity some of the parts exhibited a certain amount of noisiness and raspiness that was not supposed to be present. 2. They have developed a test to screen for this problem. Tested parts go into BIN #1, BIN #2, or BIN #3. 3. They expect 2% - 4% of the ICs to have this problem. I asked them some questions which they were totally unable (or unwilling) to answer: 1. Are 2% - 4% bad before selection, or afterwards? 2. Will this problem crop up in the 10K we have already used? I asked them to send me some "bad" parts so I could judge for myself the extent of the problem. It seemed that the "real" QC engineer was on jury duty that week but he would call on Monday (6/19) and they would send me some "bad" parts by Federal Express. By the end of that week I hadn't received anything so I called Iraj. Finally, On Monday 9/26, I received some "bad" parts, but no phone call. I compared the "bad" parts with a known good part. Of the two "bad" parts, one was noticeably worse than the other, sounding like a worn out record or perhaps a speaker with a torn cone. However, I do not believe that an average player will notice the difference, especially with the game playing, and especially in a typical arcade. Rick Moncrief and Earl Vickers were present and agreed with this conclusion. I want to emphasize that these were only two parts and were at room temperature. Perhaps it gets worse inside a game. Perhaps these were the "best" of the "bad" parts. If the ICs were any worse than the ones I heard, I would not approve their use in a game, especially since we are paying for good parts. I would like them to answer my original questions. If they continue to pull this kind of crap, perhaps we should look for another speech synthesizer. cc: Mike Hally Rick Moncrief Earl Vickers _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Speech Recognition Demo Dt: 10/3/83 Cynthia Petroka and I went to the Demo at 1272 Borregas to see the Dragon speech recognition system. The company is named Dragon Systems and the gentleman who gave the demo (whose name I did not get) was formerly with Verbex. It seems that he and his associates have developed some new and very efficient speech recognition algorithms tailored to run on an 8 bit microcomputer, specifically a 2 MHz 6502. The demonstration was given on a 1 MHz Apple II which is the reason he gave for why some of the features had been removed. Cynthia thinks that the vocabulary had been chosen to make the system look good, and I agree. Nevertheless the demonstration was very impressive. Even though it was set up in the speaker dependent mode it was very successful in recognizing other speakers. He trained the system to recognize a new vocabulary but by then it was very late and Cynthia and I had to leave. Dragon Systems is set up to do R & D and they have no plans to market the device (algorithms) themselves. I do not know how much they want for it or what kind of business arrangement they are interested in. I expect that Steve Bristow and Ted Hoff will know. (They were both responsible for the meeting.) If we were to use it in a coin-op game it would require a dedicated 6502 (2 MHz), 1K RAM, 24K ROM, a microphone, a preamp, an A/D Converter, and a game interface. This would provide for a 16 word vocabulary with speaker independence. The estimated cost for developing the data base for a speaker independent vocabulary is $10K/word. The major problems in using speech recognition in a coin-operated game remain: 1. Can it be sufficiently speaker independent with high enough accuracy? Will it still work when the player is excited? 2. Can a way be developed to provide sufficient signal-to-noise for the players input? It may require that the cabinet be specifically designed around this requirement, such as with a totally enclosed game. 3. Can speech recognition be used effectively in a game? Cynthia had a good idea. It could be used for controls that are used infrequently, like the shield button on Alpha One (Tollian Web). 4. Does anyone want to use speech recognition in a game? When I asked, last spring, people were basically indifferent to the idea. Jed cc: Steve Bristow Cynthia Petroka _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Manufacturing's Dual Audio Regulator Board Proposal Dt: 10/20/83 The following concerns Manufacturing's memo of 9/29/83 "Dual Audio Regulator Board Proposal". SUMMARY: The Audio/Regulator II series was designed to meet rather rigorous specifications so that it would work in a wide variety of product. What they want to do would probably work, but would work only with Pole Position. This would mean maintaining additional inventory in Manufacturing, Customer Service, and in the Field. Atari has traditionally put a high premium on standard componants. We should not design new, nonstandard Audio/Regulator boards for each project. ----------------------------------------------------------------------------- It is not clear from their proposal exactly what they mean to do and how they mean to do it. 1. They state that "some games require more audio power than our current 'universal layout' can provide. A. This is debatable, as previous studies have demonstrated that the way the cabinet is designed is much more important to the loudness of the game than is the power of the audio amplifiers. In fact, with a poorly designed cabinet (that is, acousatically) even 100 watts would not be enough. B. Are they proposing to increase the power of the existing amplifiers? If so, how? C. Are they adding more amplifiers to the board? 2. They state that Pole Position, Ice World, and Firefox require two Regulator/Audio II - 01 boards which use 67% of the PC board and 50% of the heatsink. A. If the empty board space on the A/R II-01 bothers people, they can use the A/R I which is electrically equivalent but shorter physically. In the past we have used A/R II in its various configurations because the A/R II -02 can be used as a replacement for all the other dash numbers in the A/R II series, simplifying spare parts inventory for the operators, distributors, and for ourselves. B. Pole Position did use two Audio/Regulator II-01s but Ice World (now called something else) will use two Audio/Regulator I boards. Firefox is scheduled to use the Switcher, besides which it requires +/- 22 volts and would not use the A/R II-01, anyway. C. Their estimate that the A/R II-01 uses only 50% of the heatsink may be correct from a visual standpoint but is thermally incorrect. The major source of heat is the 5 volt regulator which may dissipate as much as (11.5 - 5) Volts * 6 Amps = 39 Watts. (If the 10.3 V is not loaded to a total of 8 Amps it tends to be closer to 11.5 Volts.) The +12 volt regulator dissipates a maximum of about 4 Watts and the -5 volt regulator adds about another 4 Watts maximum. (Actually the -5 volt supply has not been loaded this much since we stopped using the 3 supply dynamic RAMs). The TDA2002 amplifiers probably average 1.5 Watts each when they are both used. (The audio power delivered to the speakers when averaged over times of 5 minutes is small compare to peak power during parts of game play.) Therefore, when everything is fully loaded, the 5 volt regulator contributes 39 watts out of 50 Watts for 78% of the heat. 3. They state that a prototype was developed and tested. Is this what the drawing in the memo is of? (It shows two amplifiers and two 5 Volt regulators). What is the extra heat sink that they are referring to? Does the 2N3055 "surface" mean the "case" ? 4. The test they conducted is inadequate for two reasons: A. They don't say what the game installation temperature was. We spec games to operate at an installation temperature of 38 degrees Celsius. If the game was tested at 24 deg.C (75 deg F.) you must add (38 - 24) = 14 deg C. to all the readings to find what the maximum temperature could be. Still, when this is done, the temperatures they report would be acceptable. B. The A/R II was designed to meet rather rigorous specifications so that it would work in a wide variety of product. What they want to do would probably work, but would work only with Pole Position. This would mean maintaining additional inventory in Manufacturing, Customer Service, and in the Field. Atari has traditionally put a high premium on standard componants. We formerly had a transformer that was designed to operate only on 120 VAC (The old Domestic Transformer). A power base using one of these beauties saved about $4 over one using the International Transformer which had the taps to operate at 100, 120, 220, and 240 VAC. Atari did not want to maintain the additional inventory, with the result that now all of our power transformers are International. We should not design new, nonstandard Audio/Regulator boards for each project. (Aren't these the same people who want to have a standard cabinet?) 5. The Audio/Regulator II is already beyond the point at which linear supplies are cost effective. That is why the Switching Power supply has been developed. _____________________________________________________________________________ To: Rick Moncrief Fr: Jed Margolin Re: Linear Power Supply for Firefox Dt: 11/15/83 The only game currently running stand-alone is the one that went to the Show and is now in the Shield Room. They do not have one upstairs that is running stand-alone. The total current in this unit on the 5 Volt supply was measured to be 7.2 Amps. We were not able to measure the boards separately but I doubt that one board is drawing more than 6 Amps. If this is representative of the game, then we can run it with one power tray (with an extra capacitor as in Pole Position) and two Audio Regulator II boards. Otherwise it will require two power trays. However, the hardware is said to be still changing. They expect to remove some memory and that should decrease the current. If they get below 6 Amps, then only one Audio Regulator II would be required. (I think this is unlikely with their hardware.) Assuming that they will require two Audio Regulator II boards, the following things should be done to ensure that Firefox can be used with the Linear Power Supply: 1. The +5 Volt supply should be brought onto J20 as two separate sections, each with a +5 Volt Sense line. 2. These two sections must be run separately to the Main and Graphics PCBs (and kept separate). 3. Do Not Daisy Chain the Power Lines. 4. The 5 Volt return lines (and their sense lines) should all be connected together on the Video PCB, unless someone can figure out a reason for doing it some other way. When the Switching Power Supply is used, the two 5 Volt sections will be connected together. 5. The Reset circuit must be able to work with either Switching or Linear supplies. Sam Ly said he has a circuit for this. 6. A way must be figured out so the PC boards will work with either +22,-22 unregulated (The Linear Supply) or +15, -31 (the Switcher). We may have to change one or two resistors on Game boards that are used with a Linear Supply. I expect that the sitdown and upright versions with the Linear supply will require separate FCC testing and UL investigation. That means building upright and sitdown prototypes. Will the games need different model numbers to differentiate them? Jed _____________________________________________________________________________