Ultimate Tennis is an arcade game running on a custom board developped and published by Art & Magic (our team), manufactured by Themes (Deltatec) and distributed by NOVA in Europe/US and Banpresto in Japan.
How it all started
In the arcades
In 1992 we were just finished with our latest Amiga game 'Agony'. Yet the Amiga was a slowly dying platform and we had been fascinated for years by the power of arcade machines that featured custom hardware specifically tailored for games. These arcade machines easily outperformed any micro-computer or console of the time, they had more memory for smoother animations, more colors for better presentation, more power for richer environments.
We had been playing with the idea of developping an arcade game back when we were at the Ubisoft castle, years before, but this had just been something out of reach for us as we basically were just a bunch of kids.
This time we felt it was the right time to move out of the micro-computer world and enter the arcades. We went to Konami european headquarters in Germany and show them our Amiga games, explaining them what we wanted to do next. In turn they showed us their hardware, including a prototype of their latest board. We were disapointed. Although it was a nice board, it was a far from being powerfull enough to fulfill our vision for the next generation of games.
Then, a crazy idea started to build in our minds. What if we designed a board of our own? We would have complete control over the design and power of the system! This had to be that way. We were not concerned about the economics, we were just being creative !
Yves Grolet and Fabrice Devaux (a programmer who also had been at the Ubisoft Castle) started thinking about using multiple Amiga boards connected in parallel to provide enough graphics power to develop an arcade game. The idea was to have each Amiga board creating 2 playfields of parallax scrolling and combine several of them to provide a large number of simultaneous playfields. The multiple blitters would also combine to provide some serious power for sprites generation.
At the time we were travelling a lot between Belgium and France, and looking out through the window of the train, Yves told me once that in order to better simulate the complexity of nature, we would need a lot of parallax planes, ideally an infinity of them. That was our ultimate goal, and from this idea we borrowed the name of what became project 'Infinity'. Of course we didn't know anything about 3d at the time, and no machine was capable of smooth realtime 3d just yet anyways, so we were sticking to 2d planes.
Back to Belgium, Yves started looking for an electronic company capable of building our hardware based on these high-level specifications featuring multiple Amiga boards. We soon met with the guys from Deltatec, a micro-electronics company based in Liege. They were very impressed by our games and software knowledge, and after a few in-depth technical questions, they quickly understood this was some serious business.
We talked about board design, and quite early on it appeared it would not be feasible, both legally and econnomically, to use Amiga boards in our system. We would have to do it from scratch. Fabrice was not very keen about designing a board from scratch and decided to leave the project and continue his studies in Paris. Marc Albinet also left the team to pursue other opportunities in the consumer space. We were three left, Yann Yves and I, although Yves was still doing his military duties and was not always available.
We talked business too. At some point we all agreed we needed to be together in the same legal structure. Art & Magic was born as a legal business (before that, it had been a collective of freelance) and we started working on the project.
We had several meetings to define how the board would be designed. We would use the Motorola 68000 as the central processor, the same found in the Amiga and other micro-computers of the time. It was cheap and we knew how to program it. Then it was decided to use a Texas Instruments TMS34010 as the graphics chip, because it was top-of-the-line but also cheap enough. However we were not sure about its raw graphics power, so it needed to be evaluated.
Yann did the evaluation and it came out that although it had some great features (such as pixel blending and line drawing abilities) it was not powerfull enough to handle the number of sprites and playfields needed for 'Infinity'.
Hopefully, the hardware guys were specialized in developping FPGA (Field Programmable Gate Array) and producing ASIC (Application-Specific Integrated Cirtuits). In the end, we decided we would create a custom Blitter chip that, much like the Amiga's blitter, would handle all the low level sprites manipulation. The TMS34010 would still be there to serve as a master chip for all other graphics operation. We finally completely droped the idea of having specific hardware to create playfields as with enough sprites we could create all the playfields we wanted.
This thing was going to be expensive to develop, very expensive. Although there had been some money put in the company to start working, this was never going to be enough. We needed to create everything from scratch: The hardware, the drivers, the OS, the devkits, the tools, the test kits, and of course the game. We started looking for funding and it came to us the Walloon government was funding technology projects.
Every hardware platform needs its flagship title, ours was no exception. We had been thinking for months about what our next project would be. Although we didn't know exactly what to expect from the hardware, we were so confident we could create something amazing, we just went on and started developing ideas.
As soon as the game project was roughly defined, I started working on graphics research and prototype assets. Spellsinger never came out, for many reasons. You can see the full Spellsinger story on its own page, here.
After the early design meetings, Yann went back to France to develop the low level code to drive the hardware, and write the programming tools. Many detailed specifications were written on both the hardware (by Michel Caubo) and software side and the hardware guys started building a first prototype called 'Prelude'.
Prelude was a dual board prototype fitted into a 19-inch rack, all wired by hand with tiny red wires forming an intricate and complex connections network. And yet it worked. It could be connected to an Amiga monitor through a RGB cable and you could plug any 9-pin ATARI-compatible joystick to it.
Right: A top view of Prelude with its RGBSync and joystick connectors.
Bottom: Prelude seen from below showing red wires between boards and joystick port. A flat ribbon cable goes to the PC for connecting with the tools.
I became involved in the developpement of the prototype as soon as the first pieces of software and hardware started working together. Yann would travel to Belgium and we would work together at Deltatec's lab and see the very first pixels output by our hardware. This was an incredible experience, and I learned a lot there. Soon we were able to get some sense of the performance the blitter was capable of, and it was just amazing, pumping dozens of megapixels of sprites every second, a feat for the time.
I had recently left my parents home and bought a house in Namur and I was splitting my time between working home on assets and going to Liege for prototype testing. When the prototype became stable enough, I took it to my house so I was able to work there on the first levels of our flagship game, 'Spellsinger'.
The process involved picking objects from a library running on the PC side, switching it over to the target monitor and placing the object into one of the 64 playfield available in real time. Then using the joystick/joypad (here a Megadrive pad) I could move through the level and see the 3d effect of the many parallax planes.
Right: Me working with Prelude on the first Spellsinger prototype in 1992. The left monitor displays the level I'm editing in real-time, the right monitor shows the tool written by Yann (see specifications here) on the PC with the library of objects I created for level building.
After filling many documents and forms, we finally got our money from the government. It was a re-fundable advance (in case of successful launch) of roughly 600 thousand euros, enough to cover R&D of our team and production costs for a couple of years. See some local press articles here. We setup the studio in the same building where the hardware guys were located, in Liege. Yann moved from France and settled nearby and from now on we would all be working at the same place.
Left: The A&M development studio in Liege in 1993, Yves is seen on the right of the picture.
We had slowly been moving to the PC as our main development platform. As Yann was working on the game code, we realized there was no tools for creating the 2d animations (Deluxe Paint was poorly adapted to the PC and only used for static images) we needed for the game on the PC. Yves wrote a complete painting program with fast flipbook animation capabilities called E-Motion.
Although I was good at environment design, we needed character animations of a quality that I would have been unable to deliver, we wanted Disney quality characters for Spellsinger. We hired our first employee, Iwan Scheer, who was (is) an extremely talented character designer and animator.
Right: As Yves eventually came back from his military duties, he developped the library of routines to drive the soundchip daughter board.
Aurore & Oasis
The hardware was now moving to another stage, the development board called 'Aurore'. Instead of a hand-wired prototype, the Aurore was using a large PCB fully assembled in a factory and could be fit into a standard PC case. We made a few of them, so that both the hardware guys and the software guys had enough boards for testing various parts and changing micro-code of programmable chips.
Right: Aurore board with daughter sound board and external power supply.
Bottom: Gerard (one of the hardware guys) changing some EPROM on Aurore. Iwan in the background
Oasis was the full kit assembled in a PC case and included all the operationnal software. Yann had written tools for communication and remote compiling and debuging code from the PC.
Below: OASIS remote debugger, Yann programming game code.
The company was spending a lot of money and we were nowhere near the end of the development of Spellsinger. At some point, Christian our marketing director came to us and basically told us we needed a game within 8 month or we were screwed.
There was no way we could have finished Spellsinger within that timeframe so we put the game to a stop and started thinking about a project that could be done in 8 month (including testing and manufacturing) with as few risks as possible.
We visited arcades to see if there was any niche in the market we could exploit, and there was one. Sports games were common and easy to develop as the rules for the game were already established, and there was no tennis game in the arcades except a cute anime-style that no one seemed to be interested in.
We thought if we could develop a realistic-looking tennis game it might be attractive for players. We had all the tools ready, the hardware was running, it was just a matter of making the game.
One of the key aspect of the game was to use digitized images for the characters instead of hand-painted sprites like many games of the era. Actually, Ultimate Tennis is one of the very firsts arcade games to use digitized animation at all. To capture the images, we hired a professional tennis player and managed to have acces to a gymnasium in a nearby school for our shooting sessions.
The process of capturing digital animation was however very complicated at the time. The bandwidth and memory required to capture uncompressed video was uncommon as it would fill up an entire hard drive in a matter of seconds. Remember there was no such a thing as MPEG video compressionback then. To reduce memory, we decided to capture in grayscale and colorize later-on using palette techniques. This would also enable us to use the same sprite library in the game for all characters and change the color palette according to the selected player.
Even with the grayscale trick to reduce data, we had to limit the capture sessions to 30 seconds. After which we had to backup the entire hard drive onto a tape ! A process that typically took 10 to 15 minutes for each take. Needless to say the tennis player had to keep moving all along in order to keep warm and be ready for the next take.
Then came the time to process all the data. We had to manually rotoscope each frame to extract the player from the background. We also had to mask the different parts of the player (skin, short and shirt) so that we could assign a different palette to each part. I asked Iwan for help on this pretty boring job.
Above: Yves and Yann testing the game on OASIS devkit during development.
The development was quite short and effective, Yann programmed most of the gameplay and a lot of time was reserved in the end to finetune it, so I think it is a very well balanced game.
Besides developping the actual game, a lot of secondary code had to be developped, such as the cabinet operating mode and attract/replay mode.
I did pretty much all the art for the game except some of the rotoscoping done by Iwan, as mentionned earlier.
Right: The terrain image as seen on the PC tool before being applied to the Infinity parallax system.
I also did all the sound effects (recorded from a live TV tennis match) and the attract music. I had done music as a hobby for years and this was my first contribution to a commercial game. The sound hardware was very limited, only capable of playing back a few seconds of low-fi samples without any pitch or filtering change.
I first did the music with several synthesizers and sample playback boxes (mostly Roland) sequenced with Cubase on an Atari 520 ST computer. Then I recorded chuncks of the sequence into a sampling box. Those samples were fit into the sound chip rom, and by re-sequencing those samples in the game code, we could reconstruct the music on our target hardware. The sampling frequency was very low (8 KHz), hence the low-fi sound produced.
The Art & Magic logo music theme was composed by Christian, and I did the arrangements and orchestration. The wonderful A&M logo character animation was created by Iwan.
Time to ship
In the end we managed to finish the game on time, 6 month development and 2 month testing and manufacturing. It soon became our favorite game during our lunch gaming sessions.
The hardware was finalized (including sprite encryption system to avoir hardware piracy) and the first batch of boards was manufactured. The boards for the arcade cabinets were much smaller than the Aurore development boards, and were conform to the JAMMA (Japan Amusement Machinery Manufacturers Association) standard.
The distributor tested some demo boards (such as the one on the picture above) in arcades. It was an immediate success and we received order for manufacturing many more boards than was initially expected. So many in fact that we decided to produce a true ROM chip instead of relying to EPROM for the game content.
Over its life, the game grossed more than 3 millions euros.
Game images gallery
Reviews and articles
Funny french video review: http://www.dailymotion.com/video/xj26ps_ultimate-tennis_videogames
Around 5000 boards worldwide. 500 for Japan alone.
Programming: Yann Robert, Yves Grolet
Art: Franck Sauer
Aditionnal art: Iwan Scheer
Marketing and quality control: Christian Dutillieux
Game music and sound: Franck Sauer
A&M Music: Christian Dutillieux, Franck Sauer
Movements performed by: Eric Dardenne
Hardware design: Deltatec
How to play the game today
You can get a real board quite easily on ebay for between 25 and 50 euros, but youll need a JAMMA cabinet. Alternatively you can build a RGB video cable based on JAMMA specs to connect to a monitor but you'll need a separate power supply (a PC power supply has all the required voltages).
You can play the game on PC (and other platforms) using the excellent software emulator MAME. You can download the ROMs below