Acoustic method for measuring the injector delay

[Superturnier]

Calibrating the MPGuino is a time consuming task for me. It seems to go on forever, because there is two variables which you can play with. The inj. delay and the uSec/litre.
I tried to find shortcuts to make the calibration easier. I found the flow rate for my injectors from the web. Don't know how reliable that information is, but I will use it just for starters.

But the injector delay was an unknown mystery.

So I tried to measure the injector opening and closing time by "recording" its sound with a pietzo sensor ("microphone")
I attached the pietzo to the injector with tie-straps and connected it to a scopemeter.
I also connected the injector voltage signal to the other channel of the scopemeter, so that I can compare them at the same screen.



I found out that the injector opening seems to take longer time than closing.
I'm assuming that the two "bursts" in the sound waveform comes from the injector hitting the open position and the closed position. The beginning of the "burst" is when it hits and the "burst" itself is just the echo of the sound

Then I carefully measured the tickmarks on the scope screen and I got that the injector voltage pulse is 2,68ms long and time between the open/close sounds is 1,80ms long.
Then the total injector delay would be 0,88ms, Right?
I took some other samples and the result is the same.


At the moment the inj. delay setting is 980 in my MPGuino and it seems to give little bit optimistic results, so I think the 880 sounds reasonable
I just have to drive a little more to get some results, then I know more.

How does this sound to you? Do you think this sound method is sound?

[dcb]

nice I would test that the difference in length between electrical and audio signals is relatively constant, i.e. open the throttle a bit and re-measure and compare, but it looks promising to me.

The other option, and I should probably implement this on the prebuilt, is to display the counter for the number of pulses for the tank trip and then after a couple careful refills it should possible to figure out what the injdelay should be.

[Superturnier]

Originally Posted by dcb

nice
The other option, and I should probably implement this on the prebuilt, is to display the counter for the number of pulses for the tank trip and then after a couple careful refills it should possible to figure out what the injdelay should be.

That's exactly what I have been thinking. I have tried making calculations with Ms Excel (sounds very scientific), but some important information is always missing: "The number of pulses during tank trip"!

[ngrimm]

Originally Posted by dcb

nice
The other option, and I should probably implement this on the prebuilt, is to display the counter for the number of pulses for the tank trip and then after a couple careful refills it should possible to figure out what the injdelay should be.

That would be really useful.

[abently]

Originally Posted by dcb

the other option, and I should probably implement this on the prebuilt, is to display the counter for the number of pulses for the tank trip and then after a couple careful refills it should possible to figure out what the injdelay should be.

This is assuming that the owner knows the exact flow rate of their injectors, otherwise there will always be two variables, i.e. the injector delay and flow rate, correct?

[bobski]

Originally Posted by abently

This is assuming that the owner knows the exact flow rate of their injectors, otherwise there will always be two variables, i.e. the injector delay and flow rate, correct?

There will always be two variables whether you know the exact flow rate or not. The injector delay time is fixed - it doesn't change with engine RPM or load. As such, at low load (short injector pulses) the delay time represents a relatively large portion of the injector electrical signal. Under high load (long injector pulses) the delay time represents a much smaller portion of the electrical signal. If you try to find the flow rate without accounting for the delay time, it will look like the injectors' flow rate changes with load, rather than just the pulse length.
The only way you can get around the delay time issue is to use injectors that open instantaneously. Since that's not possible due to the laws of physics (even if the valve could open that fast, the fuel itself still has inertia), you will always have that delay. You could possibly minimize the delay to the point of it having no measurable effect on accuracy by using some kind of special injectors, however it would still be there.

[dcb]

Originally Posted by bobski

...
The only way you can get around the delay time issue is to use injectors that open instantaneously. ..

Sorry to be pedantic, but it is the difference between opening and closing time that we are accounting for. If you had injectors that took the same amount of time to "open" after the current is applied, as to close when the current stops then you would not need to account for the difference. Such injectors are not currently practical however (and we have workarounds).

But I do also like the OP method, gives me some ideas for measuring other things. I would be curious to see how much accuracy improves after changing to 850ish.

[bobski]

Originally Posted by dcb

Sorry to be pedantic, but it is the difference between opening and closing time that we are accounting for. If you had injectors that took the same amount of time to "open" after the current is applied, as to close when the current stops then you would not need to account for the difference.

Not at all. Good point.

[Superturnier]

Now after 6 fuelings I can give you some results.

I have put 300,48 litres to the tank.
And MPGuino says I have used 298,44 Litres.

The total error is then 0,68%, which is not bad at all. I must say I'm very pleased with the results and finally I can trust that MPGuino in my "Plastic spaceship" is accurate.

Well, ofcourse I can do yet a little fine tuning to minimize the error a bit more and.... AAARRGH! why can't I just be happy and let it be?