Fuel consumption readings

[JoeHorner]

I've got a bit of a puzzle with the fuel consumption readings on our W203 C270CDI.

Basically, they're wildly optimistic. On a run to London and back last weekend the dash claimed 55mpg but a full-to-full check over the 562 miles travelled worked out at 44.5.

I've previously adjusted the gauge to its maximum 30% correction factor using xentry because it was claiming an absurd 70+ mpg but there's no further adjustment available.

Now, I know that the dash reading is never going to be perfect but a 25% error at maximum correction seems excessively optimistic!

Does anyone have any idea what might be happening? The only two possibilities I can think of are:

(a) It's leaking. But it would be a huge and obvious leak to lose 2 1/2 gallons over a weekend, which there isn't!

(b) The injectors are leaking / injecting more than the ECU thinks. But, again, I'd expect the quantities involved (25% extra average) to show up in pretty serious performance / smoke issues. It runs beautifully and the exhaust is clean. All injectors are also properly coded.

So, am I missing something or is the dash estimate REALLY that bad on these?

 

[supernoodle]

The ECU doesn't really know the actual injected quantity.
In the software there is a map that has axes of desired injection quantity and measured rail pressure. The output of the map is an injector energising duration. That map is calibrated using nominal parts during development.
If the injector behaviour were to change, that can mean for the energising time that is expected to deliver Xmg/st now gives Ymg/st. Quite often as the injector cokes that means Y is less than X. It's entirely possible that there has been some nozzle erosion or seat wear so that Y is greater than X.
You the driver are the feedback loop in the system. So if the injectors are over delivering you don't press the accelerator pedal as much(lower demanded fuel) to reach your desired speed. Conversely, if under delivering you press the pedal a bit more (higher demanded fuel).

Yes, there are correction mechanisms like zero fuel corrections, but they only change the zero point. Lambda.correction may also be used, but not on the whole operating range.

In reality the the injected quantity is dependent on number of injections, separation and cylinder pressure. More recent software will have many maps to improve accuracy. When your car was built it was a lot less complex which may magnify any drift

I do agree though, the error you report is substantial. Just to throw something else out there, rail pressure is one of the inputs to the energising map. If that were wrong, that will throw the demand to actual out.

 

[Rotorhead500]

Have you got standard wheel sizes on?

 

[JoeHorner]

Have you got standard wheel sizes on?

Yes, the car's bog standard. Would almost win a concours event if it wasn't for the rust, couple of dents and general wear

 

[JoeHorner]

The ECU doesn't really know the actual injected quantity.
In the software there is a map that has axes of desired injection quantity and measured rail pressure. The output of the map is an injector energising duration. That map is calibrated using nominal parts during development.
If the injector behaviour were to change, that can mean for the energising time that is expected to deliver Xmg/st now gives Ymg/st. Quite often as the injector cokes that means Y is less than X. It's entirely possible that there has been some nozzle erosion or seat wear so that Y is greater than X.
You the driver are the feedback loop in the system. So if the injectors are over delivering you don't press the accelerator pedal as much(lower demanded fuel) to reach your desired speed. Conversely, if under delivering you press the pedal a bit more (higher demanded fuel).

Yes, there are correction mechanisms like zero fuel corrections, but they only change the zero point. Lambda.correction may also be used, but not on the whole operating range.

In reality the the injected quantity is dependent on number of injections, separation and cylinder pressure. More recent software will have many maps to improve accuracy. When your car was built it was a lot less complex which may magnify any drift

I do agree though, the error you report is substantial. Just to throw something else out there, rail pressure is one of the inputs to the energising map. If that were wrong, that will throw the demand to actual out.

Good point about the "driver feedback" and I guess nozzle wear wouldn't necessarily create other symptoms if they were still shutting off ok / not leaking.

As for rail pressure, guess that'd be a trip to our local diesel specialist given the price of gauges I'd probably only use once!

 

[rayhennig]

a full-to-full check over the 562 miles travelled worked out at 44.5.

Surely that's the only definitive measure of consumption. Yes?

RayH

 

[JoeHorner]

Surely that's the only definitive measure of consumption. Yes?

RayH

Yes, but it'd be nice if the dash was giving something within a couple of MPG of that in between times!

 

[malcolm E53 AMG]

Occasionally if I’m interested to know an accurate guide to fuel consumption I brim the fuel tank and check the amount when refilled against mileage. The onboard computer can be way out at times especially with short runs

 

[rayhennig]

Yes, but it'd be nice if the dash was giving something within a couple of MPG of that in between times!

Haha. Maybe it's those dodgy MB electronics at play!!!

R

 

[JoeHorner]

So, quick update on this and a teensy bit of embarrassment.

How on earth did I own this car for over 12 months and not notice this was fitted???



Someone's been playing with the rail sensor signal

Only noticed it yesterday because there's been a couple of EDC warnings with "pressure sensor signal low" fault codes over the past week. The broken off lead (bottom right) might well explain that!

Back to stock now, throttle response seems a little slower but still a decently rapid car, and trip computer adjustment reset to standard ready for tweaking on the next full tank.

I must admit to being curious about the box itself. It's microprocessor-based so presumably isn't simply applying a fixed "correction" to the rail pressure. May have to see if I can get the firmware out of it and see what it's actually doing - it's a LONG time since I tried to reverse engineer assembly code but can't imagine this thing being too complex if I can read it in the first place

 

[supernoodle]

Good find, so it was a dodgy rail pressure.

It would be interesting to find what exactly it does. I would think it modified signal based on the rate of change to give the improved response whilst not modifying the signal as much during steady state.
You could try and characterise by feeding in a signal and measuring the output.

 

[JoeHorner]

Good find, so it was a dodgy rail pressure.

It would be interesting to find what exactly it does. I would think it modified signal based on the rate of change to give the improved response whilst not modifying the signal as much during steady state.
You could try and characterise by feeding in a signal and measuring the output.

Yes, I can't see it being a simple proportional reduction in the signal because a simple voltage divider would have been enough for that.

Good shout on characterising its response like that. Should give a reasonable idea of what it was doing and whether it's interesting enough to try and explore in more detail.

I'm also a little intrigued by the missing components and those extra external contacts. If they were options for a super-duper version that manipulated something else as well then the firmware may well be the same and give some clues to what!