Feeler: Open Source OBC Firmware

[benemorius]

I've never used those but I'm sure I'll have something to say about them in a day or three when I've had a chance to look at them.

[m2pc]

FYI, I added screen shots of all the fuel levels to the photo gallery on the wiki. Anyone care to take a crack at it?

[benemorius]

FYI, I added screen shots of all the fuel levels to the photo gallery on the wiki. Anyone care to take a crack at it?

Gladly, but I'm not seeing any indication of the analog voltage given to the cluster (or indicated fuel level I guess) for each image. Have I overlooked something?

[m2pc]

If you mouseover the image or click the "info" icon it should show the tank resistance and analog reading.

I even edited in fixed width bit markers to make it easier. They line up with the middle of each bit plot.

[benemorius]

I wondered if that might be the case. I should have kept my mouth shut until I saw it on a real computer.

[johnf]

What about the new AVR32 series? From my research it seems those are well supported, cheap, and powerful. I think I saw one with like 5 USARTs onboard!

Have you looked at some of the ARM Cortex-M3 offerings?

[benemorius]

The original theory appears to still be useful. The parity bit is definitely the last one before the stop bit, leaving 11 possible data bits between the start and parity bits. It would appear that 8 data bits are now out of the question, but I can't quite tell whether 9, 10, or 11 bits total are used. It's probably 9 or 10 as 11 would be far too much precision.

Is there a chance of getting the ???'s filled in for all the images? Everything seems to fit correctly until the last two images, which both show lesser values than the preceding images instead of growing progressively greater like the rest of them. Most specifically, I believe that 3/4 and full readings would enable us to finally arrive at a conclusion. It may also be useful to get a capture of the "largest" value you're able to get it to output, whatever that may be.

S                 P  S
0  0000 0000 000  0  1
0  1000 0000 000  1  1
0  1010 0000 000  0  1
0  1110 1000 000  0  1
0  0010 0010 000  0  1
0  1101 0110 000  1  1 no light
0  1100 0011 000  0  1 quarter
0  0000 1110 100  0  1 half
0  0010 1110 010  1  1
0  1011 0110 010  0  1
0  1001 0111 100  0  1

$half = 368
$quarter = 195
$no_light = 107

half * 2 = 736

half / 736 = .50
quarter / 736 = .26
no_light / 736 = .14

[m2pc]

The "???" entries confused me; I had the resistance set to what should have been 3/4 or full tank, but the gauge was pegged at zero with the fuel warning lamp on. I'll re-test these and see if I had a loose wire.

There's also the odd behavior of the gauge, presumably to keep the needle from moving all over the place when the tank sloshes. I need to basically power the cluster off, set the resistance, then power on and take my measurements.

[328iJunkie]

You guys are flipping champions

[m2pc]

Ok, I re-captured images 9, 10, and 11 and updated the Wiki.
Apparently the test cluster I have it slightly broken (I had to smack it to get the needle to move to the correct spot).

I also uploaded a video of the digital output with a potentiometer sweeping from Empty -> Full:

[ame]http://www.youtube.com/watch?v=hjUk1zI4dM4[/ame]

From the video I can tell:

1) The last bit is definitely a parity bit since it's changing even at very low tank levels
2) The data is formatted LSB-first.

[DStein]

Dude, this is ridiculous. I wish I was genius

[m2pc]

Ok a nice little exciting update for you guys:

I cracked the fuel level signal code!

P   S  MSB ---> LSB  S  P  
l   t  10            t  a  
o   a                o  r  
t   r     DATA       p  i  
    t                   t  
#              BITS     y  
1   0  0000 0000 00  0  0 = 0x000 (0  ) - empty (infinite Ohms [open circuit])
2   0  1000 0000 00  0  1 = 0x001 (1  ) - empty (31 Ohms)
3   0  1010 0000 00  0  0 = 0x006 (6  ) - empty (32 Ohms) 
4   0  1110 1000 00  0  0 = 0x017 (23 ) - empty (45 Ohms)
5   0  0010 0010 00  0  0 = 0x024 (36 ) - bottom of empty (75 Ohms)
6   0  1101 0110 00  0  1 = 0x06b (107) - no light (100 Ohms)
7   0  1100 0011 00  0  0 = 0x0c3 (195) - quarter (150 Ohms)
8   0  0000 1110 10  0  0 = 0x170 (368) - half (240 Ohms)
9   0  0011 0111 10  0  0 = 0x1ec (492) - 3/4 (300 Ohms)
10  0  0000 1100 01  0  1 = 0x230 (560) - 7/8 (325 Ohms)
11  0  0010 1110 01  0  1 = 0x274 (628) - full (350 Ohms)

The data is 10 bits. The last bit in all the plots before the parity bit was always ZERO, so it looks like that's the stop bit, not the return to logic HIGH at the end of the bit sequence.

The highest number is 628 decimal, giving us accuracy of 0.02 gallons per bit.
Of curious note to me is this:

Standard E36M3 fuel tank capacity = 16.4 gal
16.4 gal = 62.08 liters, or 620.8 deciliters

It appears the binary value is directly equal to the fuel tank capacity in deciliters!

[m2pc]

UPDATE: I documented this better on the Wiki:
http://www.openobc.org/wiki/tiki-ind...Signal&no_bl=y

[Matt]

Nice find.

[m2pc]

Ok, now on to decoding the fuel consumption signal. I've done a lot of reading and it appears this is a tKVA or +KVA signal, indicating either engine load in 0-100%, or some sort of pulsed signal that, when combined with injector flow rate and number of cylinders, the fuel consumption can be calculated.

I'm going to stick a portable DC-AC inverter in my car tomorrow and connect a scope to the fuel consumption signal X1071/Pin 6 to see what sort of output it gives.

Does anyone have any further insights into this? I'd like to incorporate this into the Wiki.

[jarda]

Fuel economy signal is a signal from a single injector. It is a square wave - "1" means the injector is open and "0" that is is closed.
Every revolution of the engine produces a "1-0" pair. The (bosch) injector has a constant flow rate - volume per open time (assuming constant pressure - 3 bar).

This gives you fuel consumption of a single cylinder per revolution. Multiply by number of cylinders and whatever you need (current speed, distance traveled, time, etc...)

...

Here you can translate OEM part numbers to Bosch part numbers
http://ecat-online.bosch.de/toc/

and here are the specs for different Bosch injectors
http://users.erols.com/srweiss/tableifc.htm

[m2pc]

So taking the OEM M3 3.2 as an example:

Bosch P/N: 0-280-150-792
Flow rate: 244.9 cc/min * 6 cylinders
Flow rate for entire engine @ 1 revolution = 1469.4 cc/min (0.388 gal/min [1.46 L/min] or 0.00647 gal/sec [0.024 L/sec]) assuming the injector is open the entire time, which is false.

So to figure out fuel consumption, we tally up injector timings (how long the signal was at a '1') over a specified time span (say 10 seconds).
Example: 0.1 seconds * 6 cylinders = 0.6 seconds total "on" time during the 10 second window.

0.6 seconds * 0.00647 gal/sec = 0.00388 gallons (0.0144 liters) of fuel used in 10 seconds
0.00388 * 6 = 0.0233 gallons/minute (0.0864 liters/minute)
16.4 gallon tank capacity / 0.0233 gallons/min = 703.86 minutes /24 hours = 29.32 hours to exhaust all fuel in the tank.

Now how do we calculate the MPG or km/L from these values?

Acquire average vehicle speed over same time window (10 seconds). Say it's 70 mile/hr (112.6 km/h)
70 miles/hr / 60 = 1.20 miles/min (1.87 km/min)

=================================================
1.20 miles / 0.0233 gallons = 51.5 miles/gallon
1.87 km / 0.0864 liters = 21.6 km/liter

Can someone verify this?

[Matt]

Your fuel flow number is too low because injectors are rated at 3 bar and they operate at 3.5 bar in BMWs.

[m2pc]

Your fuel flow number is too low because injectors are rated at 3 bar and they operate at 3.5 bar in BMWs.

Hmm, well I got the flow rate from this table:
http://users.erols.com/srweiss/tableifc.htm

0-280-150-792 23.3 244.9 176.1 - - - - 35.8 42.6 - High BMW 6-3.2l, v8-5.0l -

Is it wrong?

[jarda]

Your fuel flow number is too low because injectors are rated at 3 bar and they operate at 3.5 bar in BMWs.

Nope, I don't know about other engines, but my pressure regulator is 3.0 bar.
http://www.realoem.com/bmw/showparts...&fg=15&lang=en

[m2pc]

dam ive always loved that bmw just its parts are unique! look how hard it goes in a race http://www.youtube.com/watch?v=DARjSqs2JBY

BANhammer for the SPAMspreader!

[Mefis]

Let us start this in british imperial units. We'll do the metric later.
Fuel consumption in MPG (distance/volume) equals the speed divided by the flow rate.
The flow rate is the single injector flow rate times 6 cylinders and times the time factor:

V/R = MPG

L | T
------- = MPG
T | V

60 mile | 1 minute | 1000 cc | 3.7854 liters | 1 hour____ |
--------+----------+---------+---------------+------------+ / ( 6 cyl * injector open factor)
1 hour_ | 244.9 cc | 1 liter | 1 gallon_____ | 60 minutes |

=2.576 / f

This means that if the injector would be permanently open on a car at 60mph, consumption would be 2.576 MPG.
If the injector open ratio is 10% (the signal is a 1 10% of the time) then the consumption would be 25.76 MPG. Sounds about right??

[m2pc]

Excellent work mefis! I think a simple timer running at high speed (100uS) will be sufficient to tally up how long this signal stays '1', and update a master counter every x samples. Next is the vehicle speed signal, does anyone have any info on that?

I've read reports of 4 pulses per tire rotation, others of 10 pulses per mph. Any ideas?

[benemorius]

Does anyone have precise measurements for the board dimensions and the cutout positioning?

[m2pc]

Not me, I'm waiting to prove my prototype before doing any board layouts; I think Mefis does however.