Self Desgined ECU

[Luis Martinez]

Hello guys.

We're planning to build our ECU this year. Have some of you built your own ECU. Can you give me furher information about this.

Thanks a lot.

Luis Martinez.
Vicepresident FSAE Universidad Simon Bolivar.
Electronics Division.
luism.fsaeusb@gmail.com

[Luis Martinez]

Hello guys.

We're planning to build our ECU this year. Have some of you built your own ECU. Can you give me furher information about this.

Thanks a lot.

Luis Martinez.
Vicepresident FSAE Universidad Simon Bolivar.
Electronics Division.
luism.fsaeusb@gmail.com

[Igor]

I have thought a lot about building my own, but decided against it with complete ECU's being so cheap (and having more features than you would be able to implement in the given time). If you're doing it only for the experience, go ahead, but make sure your planning allows for problems.

Google for megasquirt, you'll find plenty of discussion and schematics to help you design your own.

Igor

[drivetrainUW-Platt]

Cornell builds there own...

[VFR750R]

I hope someone from our EE team reads this and gives you a better explanation, but our ecm is the result of YEARS of hard work by multiple electrical engineers each year, usually at least one masters student a year as well. We always have problems up until the end, and the dyno is a much an ecm tuner as an engine tuner. It has given us the flexibility of designing the whole system to fit our needs and give us exactly the channels we need, and we know exactly what goes on when we change a value or entire algorithm. We do it at Cornell for the experience, performance, and cost, but it costs us lots of man hours and headaches in the process. I've suggested to the team that the ECM have the same channel outs and connectors as a readily available MoTec or other similar ECM so that in a pinch we can use a system that is already been troubleshooted to death. This may hurt the feelings of our EE's, and I much rather have the home ECM, but when it limits in car testing time (which is does) a line has to be drawn.

[Mechanicaldan]

MegaSquirt Do-It-Yourself ECU

http://www.msefi.com/index.php?c=2

[Johnny B]

I built 2 Megasquirts in 3 hours. Only real trouble I've had is 'troubleshooting' Gotta make sure the flux is removed, you have the right DB9 cable. Just do a comparison of features and cost over what's available. All comes down to time and cost.

[Kurt Bilinski]

What they ^ said. What are you building, a car or a computer? Said another way, if you're "producing" a car for the weekend racer, do you think they want an ECU from you or Motec (or Megasquirt, or...)? I'd feel far better as a "customer" if I knew the ECU was made by a racecar ECU maker, not some guys in a single bay of an industrial park. http://fsae.com/groupee_common/emoticons/icon_wink.gif

[Jonathan R.]

kb58, this is a weard thinking... an ECU is a big part of a car. This is what makes a big difference on the performances. There's not that long ago, because the engine managements weren't develop as it is right now, we could not get that much power with a similar engine. Today, a car can easily be considered as a computer with all the controlling electronics. So the car's computer is a big part of it.

With the "cost report", an ECU from Motec is not really the best since, it gets up the cost a lot. I would ratter like to propose a car to a weekend racer with 4K $ less and with a homemade ECU. Many affordable ECU for FSAE don't offer all the features we need. Why would I like to pay for a name if the homemade ECU is reliable.

This is why a couple of teams decides to make their own ECU. And there is one thing we have to not forget, yes the goal fixed by SAE is to make a car for a daily racer, but the real objective is to learn how to make a race car. If we would always say, "hey I'm better buying something already done, it's more reliable", then nobody would learn how to do it. I think this what I like from the FSAE, because of the small teams' budget, a lot of things has to be done by the students.

Jonathan
Laval University Formula SAE

[DaveC]

Jonathan, I totally agree. I'm not sure many thousands worth of electronics are suitable for the car we are supposed to be building. If you have people that can build an ECU, thats great, and I'd prefer it if our team could do it. If not, the really inexpensive ECUs are limiting, I was thinking the AEM unit seems to be a good value, and Megasquirt 'n' spark looks promising if you have the time and ability to be more a bit more hands-on. I may build a ms for my rally car if the custom eprom programming ability of the stock ECU becomes too limiting, and I ever have the time.

[Chris Boyden]

kb58,

Just because somebody started in a shack or garage or a small bay in an industrial park, doesn't mean that they don't know what they're doing. Why don't you go tell that to Bill Hewlett and David Packard? Ummm....that's where we build these cars....in some small bay in an industrial park or University. Maybe we should just abandon the whole project and build minis.

[Kurt Bilinski]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Maybe we should just abandon the whole project and build minis. </div></BLOCKQUOTE>


I'm not using a homebrew ECU because I want it to "just work."

That's a rather childish remark you make, I thought you guys were in college, not high school. Building an ECU isn't trivial, especially when not worked on full time. If your school has an ECU program in place for the time it takes to make it bulletproof, great, but don't expect to get it right in one year, start to finish.

[Chris Boyden]

Hey where's the sense of humor?

You could apply that same logic to every project ..."I just want it to work, so just buy it off the shelf".....

Taken directly from your website....
"Why bother, just buy a car.", which you list all your reasons as to why not, which is cool.
But, you're kind of one sided with that approach.

I see that it's OK for you to be selective with your diy projects, but when it comes to other people taking on projects, you'd rather stand back and tell them, "why don't you just buy it"...that arguement doesn't make any sense relative to the way you present yourself and your site. If somebody wants to take this on then by all means.

Future ECU designers have to come from somewhere, why not FSAE. Look at Performance Electronics, they have a background in FSAE plus all of their collective experience and they make a nice product.

The idea of making it pin and connector compatible with an off the self unit is a good one, in terms of project management and keeping things on schedule.

Developing good engineers with some experience is what this project really comes down to. Not "producing" cars for weekend autocrossers.....although they are really good for weekend autocrossing.

[awhittle]

I courently own 5 Megasquirts. The Extra stuff is plenty powerfull to control most any FSAE car. With Alpha-N, Speed Density and MAF support, Full ignition, auto tuning with wideband, launch control ...

Look at MS2 with CAN support for even more options.

see http://www.ncs-stl.com/mike and http://www.ncs-stl.com/xp

Andy

[Matador Motorsports]

CSU Northridge Has a custom built ECU you guys can check it out at performancefi.com

Ricardo Valdez
Matador Motorsports

[Ben Beacock]

I attempted to build an ECU on the last 2 cars. Both times we reverted to a stock ECU at the last minute. The problem was mostly time, since I was the only one building it and had to take on designing a large portion of the remaining car. It should be feasible with a couple EE's working on it continuously.

My setup uses an Axiom PB555, which is a simple MPC555 board with oscillator, serial port, power supply and headers for all the pins. I made 2 PCB daughterboards-- one for input filtering and the other to drive outputs.
The problem I ran into was trying to use the stock 8 tooth wheel (none missing) and syncing with the cam sensor. The code on the TPU is designed for crank wheels with missing teeth and isn't really adaptable. I was contemplating using a 12 minus 1 wheel, but it would then lose compatibility with a stock ECU (which works just fine above 6 or 7K rpm). &lt;rant&gt; Metrowerks has custom TPU microcode for engine control. They want ~$15K. I talked about sponsorship, but they only came down to $5K. It's not like its a commercial application or anything.. &lt;end rant&gt;
So I tried writing TPU code and was reasonably successful(considering the lack of simulation and debugging software) but there are no noise handling routines. This ended up being the major problem getting it going.
Other than the noise issues, the most time consuming parts are the communication protocols and the PC programming applications. Having proper programming commands and access to the ECU data is the key to quick development. In my case I use buffer RAM for all PC accessable variables/maps, and have PC commands to write updates to the ram buffer. if everything is good, a command will copy the buffer to flash to commit. If it's bad or I mess up, I can reload from flash.

The current team is purchasing an Electromotive Tec3r. Megasquirt is a cheaper option and I've used an original v1.01 board with the 'n'spark
firmware for over 3years without ECU reliability issues. (for the curious: on a VW 2L 16V with suzuki GSXR750 throttle bodies in a 1981 Scirocco).

Has anyone looked into VEMS? Its an open-source ECU and I beleive it uses an Atmega128. You can purchase one for EUR$398 http://www.vems-group.org/

[Luis Martinez]

Thanks a lot ofr the information.

Best Regards,

[RSam]

If you still want to build your own ECU start by designing noiseless inputs circuits, specially for RPM. Pay attention to power input stability and allow program space for software filtering. Try to avoid computer connectors, minimise the use of connectors and if it can't be avoided use automotive ones. KEEP TIME FOR TEST AND DEBUG, this is what can kill you.
I worked on many electronics application for FSAE car and I think a good idea would be to start by designing a good data acquisition for your engine, you will see all the problems only the measuring part will bring. Good luck and keep us informed.

Sam

[Jonathan R.]

Interesting Ben, I'm also using the MPC555 to make an ECU but on a phytec board. I have evaluated couple of options throughout the summer and I came to the conclusion that the MPC555 was one of the best. But, now that I have started the programming, I'm facing the exact same thing, there are no pre-done functions for a crank wheel without a missing teeth. So now I've to make my custom TPU functions or convince the engine's guys to change the wheel (which would't be easy). I'm lucky, I have the ASH WARE simulator (thanks to them for sponsorship), but I'm not sure if I want to get into microcode yet.

I saw the Metrowerks functions too. By the way, I was wondering how much they were asking for that, thanks for the price info. I tried the Demo version, everything compile, but because I don't have the Axiom board, I have to change some initilisation setting in the code which is not that simple for me. So I didn't really spend some efforts to make it work.

We also use external RAM for debugging.

I agree that if you want to spend lots of hours doing an ECU with complex functions, the MPC555 is the right one, but for beginners and people that want some easy and fast ECU design, it's not the best of what I've seen so far.

Jonathan
Laval University Formula SAE

[drivetrainUW-Platt]

how cheap you guys costing these dyi efi systems? last time I checked that for 200 bucks you could get a Megasquirt WITH PROGRAMMING that works all the time, with no bugs, and it has to be infinitaly easier to tune/ data log with.

[Jonathan R.]

No, it's not that cheap for a MPC555.

While we are in an ECU thread, has anybody done a signal conversion, like convert the signal from a flywheel with 12 teeth without a missing tooth to a signal 24-1 (with a missing tooth) using the cam signal to do it? I think it is possible with some logic gates, but I haven't figured it out yet.

Jonathan
Laval University Formula SAE

[Dan G]

Jonathan,

I will be experimenting with a Divide by N counter for our Megasquirt crank position sensing. We have a 100 tooth flywheel, which is more than the input can handle. We're going to try dividing the teeth by 5 to yield a converted 20 teeth. The chip has a pin dedicated to input of a sync pulse, which can create the "-1" portion of the signal. So running just a plain old VR sensor on the flywheel teeth, a single sync tooth, and about $3 in electronics to do the math between the two, we'll (hopefully) end up with a simulated 20-1 trigger input.

All this is untested, and the nitty gritty side of it is beyond my level of understanding, so I might not be able to describe it in more detail until I test it out and see what happens.

Here is a link to the chip's datasheet...

http://focus.ti.com/lit/ds/symlink/cd4018b.pdf

[Igor]

Dan,

Make sure you pay extraordinary care to filter for noise and spikes. Logic ICs will easily count pulses you won't even see on most oscilloscopes.

More robust and much more pain in the ass would be programming a small micro to do a divide-by-N.

Igor

[Dan G]

Igor, I was planning on running the VR signal(s) through the standard LM1815 based conditioner circuit first...

http://megasquirt.sourceforge.net/extra/pix/vr-input.jpg

http://megasquirt.sourceforge.net/extra/setup-wheel.html

So hopefully the Divide by N can do its thing with just the clean output from that conditioner circuit.