Engine Management with Physical Computers

[Brozoar]

Has anyone else heard of physical computers? Check out the Arduino.

Basically, it's an inexpensive, miniature computer that you can program using a language derived from Java. It can interface with a plethora of official and third-party sensors and devices (LCD screens, GPS, temperature sensors - basically anything) and use that incoming data to perform specifically-coded tasks.

The possibilities are endless, but so far as I know, no one has yet managed to successfully create a standalone engine management system from one. If, however, someone could... well I dare say, that would be game-changing.

Let's do it! Who's in?

[_tank]

It would certainly be interesting.

I'd wonder what kind of speeds this would have compared to the 944 DME. Signal read > Process > Signal write. The technology of the Arduino is certainly newer, but I'm guessing the DME is most likely made with very task-oriented electronics, very low level programming (or probably on-chip operations vs software operations)
» Pin I/O performance JeeLabs

A quick search shows someone has at least thought of this for a small engine.


I am a web developer and can write code, but I know next to nothing about the code that goes into a DME/ECU/etc, so the probability of me destroying something while trying to figure it out would be quite high.

[krystar]

the arduino is very small and capable. but it still isn't up to the task of doing realtime computations and it's not a very mature platform. you won't find arduino in any industrial applications yet.

that said, imagine what would happen if your arduino dme crashes from bad code.

[cockerpunk]

this is/was essentially the goal of megasquirt: an open source grass roots customer controlled engine management system.

the best question is why don't we have a plug and play megasquirt option for 944/951 yet?

[Brozoar]

Originally Posted by krystar

the arduino is very small and capable. but it still isn't up to the task of doing realtime computations and it's not a very mature platform. you won't find arduino in any industrial applications yet.

that said, imagine what would happen if your arduino dme crashes from bad code.

Yeah, I mean I'm guessing it runs instructions off of un-compiled byte-code. All the current "official" Arduinos are 8-bit, but I think there is a 32-bit computer in the works. And since the hardware design is all open source, MAYBE one could play with memory/bus speed and swap out some bits to make it faster...

I did not know that's what Megasquirt's aim was! That's fascinating. Incidentally, looks like there's at least been an effort to interface an Arduino with a Megasquirt for the purposes of monitoring/data-logging? Link

[DannoXYZ]

Originally Posted by cockerpunk

the best question is why don't we have a plug and play megasquirt option for 944/951 yet?

The major problem is re-using the factory speed/reference sensors. Due to the large number of signals per revolution, and the wide-range of output-voltages, it prevents a lot of aftermarket systems from working out of the box.

1. One problem is the larger number of teeth per revolution compared to a more standard 60-2 wheel. This requires faster processing to count all those teeth.

2. The other issue is having a separate TDC-reference signal that's not part of the timing-wheel. This requires a 2nd AD-converter and extra circuitry and programming.

3. Yet another issue is the voltage-ranges of the speed/ref sensor is wider than other systems. Above 5000rpms, the voltage-output is higher than the range accepted by AD-converters used in most EFI systems. They go blind and sees that the engine has stopped turning.

4. And getting the AMP male plug for the EFI box to plug into the harness is difficult or very expensive.

[Brozoar]

^Maybe you should learn English if you're gonna live in this country, boy!

J/k...

EDIT: In addition to the Arduino, there is also the Raspberry Pi, which may have more processing power. It is worth investigating, but not right now, as I am at work. =P