11 replies to this topic

[Mellon]

Found this on www.stewart-ford.com (After the intro select car / tech spec / electronics):

On each lap the car will download approximately 40 megabytes of information to the pits. The equivalent of over 27 floppy discs of information.

Is it actually that much?

What is monitored? At what intervals?

What technology is used to transfer that much information? Is it packed/scrambled?

[Christiaan]

I can't tell you about F1 but I will tell you about my own experience.

I spent 5mths setting up an experiment to develop an engine map. I had a 55O PIII processor with 256Mb Ram and 2x20Gig HD just to monitor the engine successfully under 5500rpm. Over that the PC was just too underpowered.

I So why do I need so much power, the PC was connected to an Audi TT engine. I had to design a system that collects data from the engine while it runs. I needed that much RAM because the engine speed is about 6000rpm which is 100 revolutions per second. Now for every 30° that the crankshaft makes I had to record 11 readings occupying 89bytes. So every second the PC collects 89 x 360/30 x 100 x 11 = 106.8 kB/s (not too much). Then the PC has to manipulate the data in all sorts of feedback loops and send it back to the engine management system. The system stores everything, including the calculations and the data sent out. The net result is that the demand on the RAM just for transfering data is about 60Mb per second and the storage space of all the data over 1hr is up to 18Gb. The 256Mb of Ram are just to make sure the PC doesn't slow down because the system right now might crash if as little as 100° of crankshaft rotation were not measured. The reason is beyond the scope of this text :)Still, 256Mb proved to be insufficient. The project was supoosed to work towards a system whereby the whole system will be compacted into a couple of microchips and probably a 600Mhz processor and just gets bolted onto an engine. This will enable the engine to create its own engine maps when required so that the performance will always be at its peak.

In F1 obviously two way telemetry is not allowed. My system had only 11 sensors, but I would imagine that an F1 car would have more than tenfold that number. And at 18000rpm that would add up to thirty times the amount of data per second. If I assume that they use the same techniques then thats about 3Mb per second. Over a 90second lap thats 270Mb. My number is high because it is very obvious that my data aquisition process was very inefficient, otherwise F1 cars would need Pentium8 processors. The guys who took over the experiment from me said that they managed to do a trick with 2PCs and 3Processors. They also added a couple more sensors.

My 11 sensors were (lets see if I can remember )
1.)Intake pressure
2.)Intake temperature
3.)exhaust pressure
4.)exhaust temperature
5.)Crank angle
6.)cylinder vapor conductivity
7.)intake air torque (vorticity)
8.)inteke flow rate
9.)outlet flow rate
10.)
11.)

10 and 11 were acoustic sensors, I do not exactly remember where they were located because my project involve them. I think they were used to measure engine vibration or something.

I was introduced to the basics of the engine dynamics and I ws surprised to find how critical seemingly small things were in an engine. For example the relationship between the vorticity and the inlet Pressure/temperature directly affected how much air you could squeeze into the engine, and after that, how well that air would burn. The relationship was shown on severaya 3D diagrams which should actually be 4D if our human mind could comprehend that.

I would expect all teams to scramble their telemetry, even though with 100+ variables it would still be hard to tell whats what if you did get the data.

[DangerMouse]

The Lotus Esprit S4 onwards has a "self learning" engine management system, even though the engine was rated at 264BHP it would (as long as the conditions allowed it) provide over 300BHP, it builds an engine map based on ambient temperature and at which point the engine pinks (depending on temerature and fuel octane fuel octane) and gear selected - is/was this an industry first?

[Jonathan]

DangerMouse

My old SAAB 9000 Turbo had a knock sensor that helped influence the amount of available turbo boost. As long as there was no knock you could have all the boost you wanted up to a pre-set maximum. Thus you could actually have slightly more power if you used a higher grade (Octane) fuel.

While this engine didnt produce anywhere near as much power as your Lotus, I think the SAAB "APC" (Automatic Performance Control) system predated the Lotus setup by a few years.

Concidering that this engine was/is not much different from a mid-1960ies Triumph Herald / Mid 1970ies Triumph TR-7, I guess you could say that SAAB did their homework with this engine.

[Jonathan]

Christiaan

What did you write you Engine Management control logic in ? I suspect you did this in a fairly high level langage. For the purposes of "prototyping" an engine management system, your setup was probably okay. My understanding is that as soon as a viable control system is developed and optomized It gets burnt into either an e-prom or some form of ROM. While I suspect most production systems have as many (or more) sampling parameters as yours, their sampling rates are probably much lower, and they typically use LESS than 40mb of memory (?)

I keep hearing the term "fuzzy-logic" as it applies to real time systems, But I am not sure I understand all the principles involved. How they get this behaviour with so little memory is beyond me...

Heck if NASA could get us to the moon using crude computers that had like 8k of memory, I guess that assembly languages still have a purpose in life...

[Christiaan]

I did use a very high level programme for the data manipulation, but the aquisition itself was low level binary format, were each parameter had a designation to identify it, and its value. What is supposed to happen is that the guys who took over the project are supposed to use highly complex software to optimise the combustion process when the engine is subjected to every imaginable circumstance. They are to record the setups and make them into a complex matrix in a very low level language and programme it onto a ROM chip with about 20K. The system on the engine will have a processor and two RAM chips, one for active data aquisition and manipulation. Another to store new findings and adjustments to the ROM matrix. The principle sounds simple but the implementation is something else. I do not know about the ROM matrix, but I imagine it will be multidimensional. Whats interesting about this is that the project managers know that the technology currently available on normal Intel processors and RAM is inadequate, but they are working on the assumption that computer technology will work faster than they can, and so by the time everything is ready to go to machine codes and final product, we will have 1000Mhz processors available to the ordinary man. I'd say they are not far off, because when we went shopping for PC's at the start of the project, 550Mhz was the fastest commercially available thing. But now I think its 700Mhz in a space of less than 6mths.

[Martin]

Christiaan - how on earth did you keep the chip cool enough to function properly?

[Christiaan]

Oh, I guess I didn't give you all the details of my harware. The tower was uncovered because of the number of components on the PC. There was a very expensive card from Advantech on board which took care of the imput and output signals. This card was amazing, you can manipulate that card to become a soundcard, video card, graphics card, almost anything you want. I guess thats why it costs DM6000! Anyway the card and the CPU were cooled by two gigantic fans and a small water pump(from a fishtank)with an intricate circulation. This cooling system was quite neat if I say so myself. I would like to develop it further if I get the resources, because I think it would enable us to make laptops with the same motherboards and CPUs as desktops.

[Phil Brett]

40MB over a radio channel in a short time.... hhhmmm... not impossible but it's going to be tuff with all the noise ( radio ) around the cars. And how long is it going to be in range?

It must be spread spectrum ( nice and secure ) up in the GHz.

Impressive stuff.

[Christiaan]

Mellon I thought you might want to read this article.

******************************************
Networks in pole position

Motor racing requires data to be networked from the design stage right through to the chequered flag.Network News discovers what technology is at play while cars are burning rubber
JULIAN GOLDSMITH

DARYL WILCOX

Formula One Grand Prix racing is a glamorous sport where a fraction of a second can mean the difference between bursting open the bubbly and struggling to get sponsors for the next season's competition. To gain those extra milliseconds, all the top racing teams have turned to increasingly sophisticated network technology.

The onboard computer system in an F1 car is vital as it relays data in real-time back to the pits for monitoring and "tweaking", despite new rules last year which restricted the amount of control it could have.

Away from the race, computing is also used in the design of the car. Many racing companies use CAD for both the design and for distributing information across the network so that a team of engineers can collaborate closely. The first 100 per cent CAD/CAM produced F1 car came from the McLaren stable last year. The Marlboro McLaren Mercedes MP4/10 was designed on a network of Sun Sparcstations and servers. McLaren used its network to redesign and manufacture its car in three months after regulation amendments required that the car had to be dramatically altered from the previous year. The generally accepted figure for developing a car is £30m and the technology also produced significant savings to development costs.

The McLaren team, like many of its rivals, has also extended its network to its test track. The technology used is radio Ethernet from the pit wall to the garage. From the car to the pit, Mclaren deploys a proprietary telemetry system using a microwave link developed by its sister company TAG Electronics to relay data to the pit at the end of each lap.

Williams gets on track

The Rothmans Williams Renault team has also been using networking technology for collaborative design work by storing data centrally using Raid arrays and conducting predictive stress and aerodynamic tests with modelling systems before resorting to the wind tunnel. The company's use of networking began in 1987 with 10 workstations and 15 PCs linked using Digital Equipment hubs.

The network was extended in the early 1990s by installing NetWare Lan segments in the garages. These were not initially linked to the fibre backbone at head office because it was thought that this sort of distributed network would tax the IT support resources. Instead, Tandem PCs with removable data packs were used. These packs were filled with test data at the garages and test sites. They were then removed and slotted into PCs which were wired into the network at the central site.

By 1995, Williams Grand Prix Engineering required the speed offered by an extended campus network. The company moved its site and was networked in discrete Lan segments. Williams, like all Grand Prix competitors is very security conscious and guards its network closely. The company has Internet links so that its engineers can browse for information published on it, but access to and from the network is strictly controlled by a firewall. Williams also has an ISDN link so that it can contact independent suppliers and the disparate elements of the company.

During the race season, the ISDN link also connects the race venue to the company's head office. Though useful, this cannot be relied upon, says Andrew Hope, IT manager at Williams Grand Prix Engineering: "You never know what the availability of ISDN is going to be like from country to country. Some of the most advanced countries seem to have more trouble with ISDN delivery than the more developing ones."

Linking the car to the pits

The Williams team has recently extended its network to the car on the track using a spread spectrum radio link, developed by wireless networking company Telxon. The link allows data to be captured from the car's onboard system when racing.

Jim Wright, senior acquisitions manager at Williams, explains: "We wanted to create a real-time telemetry system which would give us data from the car straight away. The quicker we have data the quicker we can use it to make the car go faster."

Telxon was selected to provide that link because it is a specialist in spread spectrum, a technology that would allow the transmission of more data than any other form of radio link. Wright adds: "The system has an advantage over the previous card system used in that it has more than one channel. We are currently using about 34 channels to monitor separate aspects of the car at once."

While developing the link Telxon had to overcome the rigorous Formula One time constraints. Telxon had three months between the end of the 1995 season and the start of testing the car to design and build a system that could do the job.

"Since November, we have designed, built and tested a brand new car in 16 weeks. Data capture from testing is so important to us. Telxon is already ahead of our planned schedules." says Wright.

The transmitter in the car is little bigger than an audio cassette and is housed in one of the car's wings. It is connected to a blade antenna situated on the front of the car. The transmitter relays information from the onboard computer system to a variety of antennae on the pit wall and directional antennae on the roof of the pit. The engineers are able to track the car almost the entire length of the circuit whilst other systems only have the eight seconds or so when the car is in the pit straight.

The system transmits data in repeated bursts. So if the car goes through a dead spot, the system will retransmit once the car is in the clear and the data will still get through. The data can also be transmitted back to the factory where it can be further analysed. As the season develops, engineers use the data to learn how the car is performing and try to solve any problems in the week and a half between races.

So far this season the Williams team has come first and second in the first race in Melbourne. Williams admits that the win is in part due to the Telxon system.

Ironically, because the TV companies believed that the radio signal from the car would interfere with their own signals, Williams declined to use the system in the actual race. Telxon is currently modifying the system and are confident that it will be in place sometime in the middle of this season.

World championship team Benetton Renault, which won both the driver's and constructor's titles last year, achieved its success using arguably the most sophisticated IT infrastructure in motor racing.

Benetton relies heavily on advanced design, manufacturing and trackside analysis systems to keep it ahead of the pack. It prides itself on extensive sharing of data between designers, car builders and mechanics - all of which are dependent on a powerful and reliable network infrastructure.

The 10Mbps Ethernet network, running Novell's NetWare, makes it possible for the team's 125 IT users to share data during the design and production cycle of the car - a process known as "concurrent engineering". Designers working on CAD software can share design information with those using CAM systems through a central Oracle database, all running on Hewlett-Packard Unix workstations and servers. The data ends up at the Computer Numerical Control machines that construct the parts - the final link in a totally electronic cycle.

Multimedia on the network

Compaq PCs and Hewlett-Packard workstations have access to multimedia email - including video clips, shared whiteboards and most recently videoconferencing, enabling staff to share information and speed up the design and production stages. The systems, based on HP's MPower software, allow engineers at the factory to share data and images with the wind-tunnel, 65 miles away. During races and testing, such information is also

[DangerMouse]

Jonathan, I think having a knock sensor to effect the reduction of turbo boost against a system that actually dynamically builds a new engine map according to conditions are rather different beasts!

The Saab system also has a boost cut out where if despite it's crude boost control it over boosted, it cut the engine out completely until the boost dropped below a predetermined figure - clumsy and not exactly rocket science!

Not knocking Saab (I like Saabs) but you cannot compare the two systems!

[Yelnats]

Yes, The turbo overboost cutout is universal along with the ignition ******/knock sensor.

If you wan't to have a bit of fun with your turbo, you can disconect the waste gate actuator pipe as I have done on my 89 Dodge 2.5 lite turbo. This allows boost to rise to the maximum as the wastegate hase been disabled. What a blast! It will cutout above 3500 rpm if the boost gage is not monitored and throttle applied accordingly. As I can't keep the throotle down for more than a few seconds without attaining suicidal velocities I don't expect too much damage but Hey, It's only an 89 junker and replacments are cheap!