Measure fuel consumption from injector signal?

Assuming the fuel pressure regulator is working PERFECTLY. Which in the real world it doesn't; fuel pressure varies substantially between idle and highway driving.

Under similar driving conditions, yes.

Right, you can still tell where the thing opens and closes.

If you have a carb, you can just stick a flow sensor in the fuel line and be done with it. That's even easier... one sensor and a meter movement.

--scott

I don't think we're communicating. It's NOT my car. The fuel fittings on that Honda were non-standard. Obvious way to install a flow meter is to cut the hose. And it's not my car. Looked like hooking a wire to the injector would solve the problem I have...If I get the algorithm approximately right.

Having said that, do you have a recommendation for a cheap flow meter that works over the range of flows we have here... and is gasoline and alcohol resistant? Maybe I can talk the owner to installing one so he can calculate actual MPG. Thanks, mike

What is a flow meter going to tell you? If you take your foot off the gas the flow will decrease and if you step down on the gas the flow will increase. What conclusions are you going to draw from that information?

Accurately measuring the amount of fuel consumed in 80 miles of typical driving would be a more meaningful way to determine mpg.

-jim

-jim

Flow meter and injector signal will tell you nothing useful unless the vehicle is in a closed environment like a dyno room and you can duplicate the same drive cycles every time. Then you could actually measure the devices impact. Then you use an accurate measurement of the weight of fuel used over time.

However I also doubt the results your pals saw are real. Just not possible for some add-on item to give that result on a modern computer controlled vehicle. UNLESS they drove downhill with the engine off.

Take a look at the EPAs testing of some of these "gas savers" and you will discover that none of them make a measurable increase in mileage on a modern vehicle.

additives:

Drive down the road at 55mph. Read the fuel flow. Switch on the magic device. Read the fuel flow. The MPG number is irrelevant. The question to be answered is, "Did the magic device reduce fuel consumption?" Sounds pretty simple to me. And a lot more reliable. Instant debunker!!! What am I missing???

There's some question whether the injector pulse width is a repeatable measure of fuel rate over a wide range of operating conditions. If it is, you can just add up the pulse widths to get the volume and divide into the miles which can be determined from the pulse count if you stay in the same gear, or the odometer if you don't.

Yes, calibration would be required if you cared. I don't. It's quick debunking that I'm after.

If you take your foot off the

OK, how do you do that without modifying the vehicle?

People try to do that all the time by reading the pump when they fill their tank. Averaged over several tanks with the initial and final fills at the same pump may be useful...just not quick. Doing it for two gallons on an

80 mile trip with fills at different stations is not useful.

And what do you consider typical? Changes in driving habits can easily mask changes in efficiency. I've had some smart people tell me how some gizmo improved their gas mileage 10%.

I asked again six months later. The improvement disappeared when they stopped thinking about it and subconsciously driving less aggressively.

Whodathunkit? Putting a holographic sympathetic resonance enhancer sticker on your gas tank doesn't improve gas mileage.

Well the easy way would be to hook a decent scan tool up and see what percentage the fuel tank reads. Fill up the tank so the reading is say

90%. Zero the trip meter. Drive around till you hit 80-100 miles. Read the percentage left and the mileage. Refill back to the original level. Zero out and drive some more. See what the difference is.

The problem is that unless you drive the exact same way, under the same conditions using the same gas you won't get accurate results.

Why? Different additives and base stocks between supplies/stations/ If the air temps are different the readings will be different due to the air density difference. Unless you do the driving over the same route every time and using the same throttle levels the results will be wrong.

Correct. But that is what you seem to think the flow meter will allow. When it will not. The first method is shown to work and be reliable.

Which is why you don't test with a human driver. You test using a dyno and a computer controlling the vehicle.

Interesting. What's the accuracy of the fuel readout? Do I need to obtain a different scan tool for every vehicle I want to check?

Fill up the tank so the reading is say

Please elaborate on why the flow meter won't measure FLOW. While I'm waiting for that answer, I'll jump ahead...

Assuming I really do want to measure flow, I'm asking if measuring the injector pulse width and frequency and applying some math won't give me RELATIVE data that's almost as good as an inline mechanical flow meter. All I gotta do is clamp a current transformer on an injector wire.

IFF the injector injects some number of microliters per microsquirt-millisecond, (And IFF is the question) why can't we just add up all those microsquirt-milliseconds and learn the total microliters in real time, big time, 4-4 time or any other time we want??? For my purposes, what I want to know is the rate of fuel consumption right now, in real-time.

When I did the math, most of the terms canceled out. For a fixed gear ratio, with the clutch out, Gallons/Mile is directly proportional to the width of the injector squirt. If you drive twice as fast, you get twice as many squirts, but travel twice as far. Only variable remaining is the width of the injector pulse (assuming the squirt volume is directly related to pulse width). Yes? NO? Sounds WAY too simple, but that's what my math sez.

We're getting way off track. All I wanted was a reasonably repeatable method to measure fuel flow in real time without spending any money or modifying the vehicle or buying any specialized equipment.

Turn on your gizmo, maintain same speed, does the fuel consumption go down? Simple! Doesn't cover all driving conditions, but with a few tests under different repeatable conditions, it should be relatively easy to convince an enthusiast that his gizmo ain't workin'.

Not if the vehicle is OBD II or CAN or newer. Fuel level is as accurate as the sender unit in the tank. As long as the vehicle orientation is the same it will read the same.

The flow reading is WORTHLESS, Very few systems run dead headed, They run with an open loop through the system.

Fuel comes from the pump up the line, into the fuel rail and out through the fuel pressure regulator and back to the tank through the return line. The injectors use a portion of the fuel coming through the system.

That is the same basic idea that most of the instant mileage readouts use. However the ECM that does the calculations uses all the sensors on the vehicle as data inputs to compare against the map programming and load of the vehicle. The problem is that those maps are not going to be easy to find, and you cannot just guess at them.

Then grab a vehicle that has that built in and test using it. Adding it to a vehicle that doesn't have it can be done IF the ECM supports it. BUT even then the instant readouts are a joke. The only true number is an average over time. That is the only real way to tell if there is a difference.

Take a look at the testing that was done on the electronic gimmick on the EPA site. Note how they were told that they didn't test it correctly and that they didn't follow the directions because they used it on a newer vehicle. So they went out, found older vehicles and followed everything that the maker said they needed to do. The thing still didn't work.

And it is way to simple. When you test an injector for flow you use a pulsed current to open the injector, then you use the known pressure on the rail, viscosity index of the test fluid (usually something like stoddard solvent) and a very accurate timer. In your case you will only know the pulse width (which changes EVERY time the injector fires) You won't know the fuel pressure unless you install a gauge.

Then you go with the scan tool. Fill up to a set amount, drive around and back to the start, record the fuel used, repeat a few times so you know whet the vehicle gets for an average. Then repeat the process with the gadget in place. There is NO quick and repeatable method that is also free or devoid of modifications.

You are asking for the same thing that you derided your pals of "something for nothing"

Not really. Most of these people will come at you and tell you that YOU'RE just not testing it correctly. Or that it takes more than one tank to work, Or that you are not using it correctly.

Why? Because they KNOW that it works for them...

That would work great if getting better fuel economy was something that can be "switched on"

It's not like you can flip a switch and your BMW becomes a Prius

-jim

Good point, this one isn't tho...

Fuel level is as accurate

Got a NUMBER for that accuracy spec?

As long as the vehicle orientation is

That could be useful if you can watch the readout while you fill the tank.

Excellent point that I'd missed. All the more reason to use the pulse width method.

Now, you're confusing me. If I want to measure fuel flow. And I DO measure fuel flow. How does this map data help, or hinder me me?

I can't just grab a vehicle. It's not my vehicle. It's not my gizmo. I gotta take what I get and make it work.

Adding it

OK, let's simplify. It is a simplification to focus the thread to get at the answer to the question I asked. Begin Thought Experiment.

Gallons = Constant X Ton

Assume I have an infinitely accurate and precise way to measure gallons of fuel consumed. I make a test run adding up all the pulse widths of all the injection events and call that Ton. Back at the lab, I read the infinitely accurate and precise volume gauge. I get out my calculator and solve the above equation for Constant and write it down as ConstantRun1.

I repeat the experiment multiple times with different, but normal driving parameters recording each value of Constant.

There is no experimental error in this thought experiment. Temperature, humidity, altitude, fuel composition, phase of the moon, time of the month are all fixed in this thought experiment.

Can I expect all those calculated values of Constant to be the same? If not, what factors affect the outcome?

Remember, I'm not looking for ways to make the experiment fail. I'm talking about normal, typical driving conditions. I'm looking for ways to identify and reduce the impact of things that would cause the values of constant to differ.

One such thing is suggested by the literature: subtract a millisecond off the pulse width of each injection event to account for delays in opening/closing. Does that sound rational?

End thought experiment

End of Simplification.

Since I don't have an injected car to test, I have no idea how long a typical injection pulse is. I assumed it couldn't be longer than half a revolution, but it seems that having a valve open is not a necessary, or even desirable condition for an injection event.

I intend to measure the pulse width every time the injector fires with whatever accuracy and precision is required. In a previous life, I designed frequency counters for a living. It's not rocket science...it's an afternoon of fun.

I am making the assumption that all the injectors track, so I can measure only one and infer the total.

You won't know the fuel pressure unless you

The fuel pressure is regulated by the fuel pressure regulator. Any reason that it should change from one run to the next? See thought experiment above.

Ditto for the way the injectors work. I'm looking for RELATIVE data. If relationships are constant, I can deal with the correction factors.

Again, not my vehicle. Don't know if it's new enough to have a scan port. Even if it does, I ain't buying a scanner for it.

Well, I've proposed one. I have yet to hear why measuring the pulse width won't get me to a reasonably accurate, real-time, relative measure of fuel consumption. Or better yet, suggestions on how to tweak it for better accuracy.

You're correct. I mis-spoke. It should be relatively easy to convince my friend that the other guy's gizmo doesn't work.

Thanks for all the inputs on the psychology of efficiency improvements.

What I really want is input on how to improve the accuracy of my model for the relationship between injector pulse width and fuel flow:

Gallons = Constant X Ton.

mike

mike wrote: Snipped-

If it is older than OBDII then it's not possible because they don't show the data on the buss.

I would bet that it's within .25% or better. The sensor is nothing more that an open potentiometer. The wiper arm is connected to the float. I wouldn't trust that 100% = a full tank or that 0% = empty. Mainly because of the tank design. However I would bet that if you took a number like 60% and measured the amount of fuel from 60% to 80% it would work out to the same amount every time.

Yep, but you don't have that option.

Now you know why the flow numbers don't work in regards to overall fuel use.

Because without it you won't stand a chance of calculating the actual fuel flow through each injector.

If you have this then you just use a simple math problem.

Amount of fuel consumed divided by miles driven - MPG

If all the variables stay the same, the result will still vary some.

Actual injector flow. A worn injector will not flow the same as a new one, a weak spring or a build up in the injector as well as the actual fuel used all make a difference. Also the injectors do not fire the same all the time.

The ECM looks at all the sensors and actually follows a closely sinusoidal wave to control the engine. It cycles from lean/rich constantly to attempt to keep the mixture correct.

The pulse train varies based on sensor inputs. That is the problem with the simple approach. It isn't valid.

Then consider that the injector pulses are a high frequency square wave. The pulse opens/closes each injector using the square wave pulse. BUT that pulse itself varies as well.

Yep, they are controlled through engine vacuum to vary the actual flow volume and pressure based on engine load. That is why when you test them you start with a static measurement window. IE a GM high pressure pump may put out 100 psi. The regulator on the engine may be set up to provide 55 pounds nominal. So for a static test you turn on the key and measure the pressure. the window may say that you're OK if the fuel pressure KOEO is between 53 and 58 pounds. Then you start the engine and test the pressures at idle and at WOT. You will also have a window for that.

When I hear a statement like that, I normally cry "BS!!" and ask to have some of what you're smokin'. While it's certainly possible, the odds are stacked against it.

1) There's no reason to know the fuel level to that accuracy. 2) You'd need four digits of readout to make use of it. 3) The cost of a .25% pot is ASTRONOMICAL compared to your garden variety, crank-em-out-by-the-million-and-don't-even-bother-to-test-em pots you're likely to use in a fuel gauge. 4) The bean counters care a LOT about cost.

I'd wager there's more than .25% backlash in the system due to the buoyancy of the float against the friction of the pot...in an older car.

The sensor is nothing more

The purpose of a thought experiment is to examine a hypothesis. You're welcome to have any thought experiments you wish. This one is MY thought experiment.

Methinks you're just trying to be contrary. That's OK, cause you're disclosing little tidbits of additional useful info along the way.

I've stated over and over and over that I'm looking for a RELATIVE measurement on a SPECIFIC instance of an engine/car/device. If it has a weak spring for experiment one, it'll have that same weak spring on experiment two. Can we just assume that the car is mostly working?

Ok, riddle me this... If I have an engine with a fixed number of cylinders, say 4 just to pick a number. Will not ALL the injectors fire some number of times per two-revolutions? Won't there be a constant factor that I can use to multiply the injection events of one injector to calculate the total number of injection events per two-revolutions?

I'm assuming that the ECU is not controlling on a per-cylinder basis and that any differences in the flow from each injector will be systematic errors and irrelevant in a RELATIVE measurement.

That's exactly what I'm looking for. Switching on the fuel saving device should have some effect that the ECM corrects by reducing the gas consumption, which is being measured by monitoring the injector pulse width. If pulse width doesn't change, the device is NOT saving any fuel. I can't see how it can be any simpler than that.

The nice thing about zero change is that precision and accuracy are irrelevant. Zero is still zero no matter what. If it's NOT zero, it's time to do more careful and time consuming tests. I don't expect to ever need to do that. I'm interested in debunking bogus claims by supplying scientific evidence through simple, easily replicated experiments.

Maybe I should market it...the TURBO-DEBUNKER 1000.

Yes, but I'm MEASURING the pulse train. How is measuring the thing you want to know not valid?

The item in question is the exact relationship between the injector pulse width and the amount of fuel injected.

Do you really mean what you said? My expectations are that the pulses aren't anywhere near square and are very low frequency. 6000RPM is only

100Hz. Even if you had 10 injections per injector per revolution, that's still only 1KHz. Aren't pulse widths in the few milliseconds range? Useful reference info on typical pulse widths would be helpful.

I can't seem to get across the fact that I'M MEASURING that pulse. Yes, I know it varies. That's what I'm MEASURING.

Now, we're getting somewhere.

I started with zero knowledge. I interpreted one reference thusly:

At Idle, manifold vacuum increases. At load, it decreases. The purpose of the fuel pressure regulator is to maintain a constant differential pressure across the injector. (fuel pressure - manifold pressure == constant) That way, the injection volume is an invariant function of pulse width and the ECU has fewer variables to deal with when calculating the injector pulse width.

If that's the case, all I have to determine is the transfer function between injector pulse width and fuel volume/squirt. This should be mostly delays in ramping up the injector current, moving the mass of the valve and the fluid dynamics of starting and stopping the liquid mass flow.

And that's where the magic number of 1ms to be subtracted from the injector voltage pulse width comes in. Yes, it's an estimate, but it's consistent with the electrical and mechanical "inertia" involved.

How did I misinterpret this? Help me understand what is really going on here.

One more time, I'm interested in relative numbers on the same system. Systematic errors are of less importance. I'm more interested in precision than accuracy.

I don't think anybody said that. In this particular instance, the guy has a switch on the dash. One way says, "OFF" the other says, "ON". Am I gonna argue that his switch isn't valid?

OK you might have explained that from the beginning since that is relevant and all the other crap you did say is irrelevant.

How do you know this switch doesn't do something that gives the car bad mileage when it off and normal mileage when its on?

Measuring instantaneous fuel consumption is likely to be difficult because it will involve serious modifications to the fuel system. The car probably has a return fuel line so you have to account for that also

The easiest way to measure fuel consumption is still going to be accurately measure how much fuel left the gas tank

The manufacturer should have a method by which to demonstrate his claims.

You have it all backasswards trying to prove it is ineffective when you seem to have no -evidence- it is.

In the olden days I filled it up to the tippy top, drove around for a while, then returned to the same pump and filled it again.

That should be sufficiently accurate, but I am unaware if that method is feasible with more modern fuel systems. -----

- gpsman

Are you of a news-group or message board concerning fuel economy? Ben

Not sure how that helps anybody. Not even sure what you're asking.

But, If you're asking what I think you are, the answer is NO!

Thanks for telling me I'm full of crap. You're now a full member of the thread. I won't take it personally; it's the way of the web. I just gotta keep at it until someone who knows how the system works weighs in with some constructive data. If you think I'm full of crap, that's ok. If you don't wanna help, that's ok too. You wanna tell me I'm full of crap, just go ahead. Useful tidbits of relevant info tend to creep in.

Maybe someone with real understanding and a desire to actually help will show up. If not, I'll just build it and make it work.

What's relevant to me is a mathematical model of the relationship between injector pulse with and fuel flow. In the "middle" in steady state it's ~linear. The question is, "What happens at the ends." Idle? Max Load?

Whether you think that's useful information is irrelevant. It IS relevant to me.

Actually, I know a LOT more about the system than I'm allowed to disclose.

I didn't ask any questions about the switch or the test methodology. I asked about the relationship between injector pulse width and fuel flow. THAT'S ALL. Anything else is a diversion from the objective. I regret supplying any back story, it's diverting you. You don't have to believe that measuring fuel flow will help me. I believe that measuring fuel flow will help me. Only I get to vote on that one. The function of the switch, the gizmo, your opinion of my competence or the phase of the moon are IRRELEVANT. I don't want to know what the web thinks about the switch, my competence or anything else EXCEPT THE RELATIONSHIP BETWEEN INJECTOR PULSE WIDTH AND FUEL FLOW. If you don't have that info, it might be helpful NOT to press that "send" key.

Well, I've proposed a simple non-invasive method. The closest thing to help I've been given is, "you're full of crap; it won't work 'cause it's complicated". No actual model of that complication or data has been presented. The responders have given no evidence that they understand how the system works much better than I do. And I don't know squat. Any acquired data can be "fixed in software" if you have an accurate model of the system and have enough parameters to work with.

Yep, that's conventional wisdom. If you want to know MPG, measure miles and measure gallons and divide. I fully support measuring exactly what you want to know. But the devil's in the details. By that method, the MPG (claimed by the user) went from

34 to 57 with the gizmo turned on. Do YOU believe that???

Sometimes, an indirect measurement can tell you what you need to know with less experimental error and in much less time.

What I want is a method to measure fuel flow in real time non-invasively. No more, no less... If what I've proposed won't work, what would it take to fix it up? Telling me I'm full of crap won't help me fix it up. It's much more helpful to put yourself in the frame of mind of an educator and teach.

How about telling us about this "Gizmo" and how it is theoretically supposed to work. I'm sure there are enough people here who will be able to tell you if it is going to work. Some may already have had experience with it.