Something I’ve noticed in multiple makes and models of car and have long wondered about - when I do the crudest, most high-level of miles-per-gallon (MPG) calculation, which is take the odometer miles since the last fill-up and divide by the gallons you’ve just put in at this fill-up, I get a certain number.
In construction terms, this method is basically equivalent to a crude hut made of mud and sticks.
But the car’s ECU has literal continuous millisecond-granular information across hundreds of sensors on fuel pumped and consumed in the engine, and even how rich or lean each of the detonation cycles are - in addition to that, it has literal computer chips capable of thousands of operations per second that are then able to use that data to calculate to the N-th decimal place the actual fuel consumption of the car.
This methodology in construction terms is basically an ornate towering cathedral to the crude mud hut of miles driven / gallons pumped.
And yet the number you get from the buttresses and stained-glass car computer method is consistently 10%-20% higher than the mud-hut actual MPG (aka it overestimates MPG by a non-trivial factor), despite having vastly better information and computational power to throw at it.
I’ve noticed this across many makes and models with manufacturing years spanning at least 3 decades, so it’s not a case of just one automaker doing it, or older computers and sensors vs new ones actually being accurate.
Has anyone else noticed this, and does anyone have any informed opinion on why this is so?
I have a ScanGauge plugged into my ODB, and it seems fairly close. When I fill up, I usually end up putting more gas into than it says I will, but not that much, and sometimes it is a little less. I am going to guess that the fuel flow sensor may have a bit of play in it. Just a tiny tiny bit of slop would end up making a big difference over a tank.
Yes. I’ve kept a spreadsheet since I got my car, and kept one on the one before that. The number on the MFD is an average of 3% to 4% higher than the number I calculate by dividing miles by gallons. In the real world, it’s not possible to control all of the variables. But I try to fill up at the same pump, at the same gas station, at the same time of day each time. I lock the nozzle to fill, and remove it when it clicks off.
I have a 2014 Mazda 3, and so far as I can tell, the average MPG is bang-on with my real-life numbers. That said, I don’t spreadsheet it, and I don’t know if I would notice a 3% difference. All I know is since I’ve had the car, it reports about 32 MPG average over all my trips, and somewhere in the range of 31-34 mpg is what I average between fill-ups, so if it’s off, it’s not by a noticeable amount.
Are you a driver who tends to either accelerate quickly, brake late? That can add to the problem as those peaks will be averaged down but it is a complex issue.
If we tracked consumption (gallons per mile) vs. economy (miles per gallon) the calculation would be slightly more accurate.
Tire diameter changes over time, and is not perfect so the speed displayed on the speedometer is off by a small percentage.
The speedometer calibration is typically set to be a few MPH higher than actual due to tolerances etc. They tend to read by an amount equal to the average error as to never indicate a lower speed than the actual speed of the vehicle.
ODBC has a ‘fuel consumption correction’ factor that is set to be conservative and would need to be calibrated to be exact.
If they overestimate on MPG people would run out of gas doing the math, if they underestimated on MPH they would have customers complain about getting ticket.
Increasing precision costs a lot of money and most consumers don’t care
The volume of fuel being injected in most systems is highly accurate for the most part these days. But there are several factors besides what I listed above which may impact absolute accuracy and it doesn’t take too much for a drift in the true value to be off.
For the main use case most of these systems like fuel injection need Precision or closeness of two or more measurements to each other more than Accuracy or the closeness of a measured value to a standard or known value. To state this differently FI system cares that it is operating within emissions and performance bands far more than it cares about the absolute accuracy of the MPG calculation. In this application the Precision of fuel flow measurements is more important than the Accuracy of the MPG calculation.
I have a '13 and have noted the same problem. My calculations were always 3-5 mpg under the computer’s calculation. I did the refueling as consistently as is possible every time I filled up over several dozen tanks. It always came out lower than stated on the dashboard reading.
I was swiftly corrected by another poster who told me in no uncertain terms that I was wrong and the car was right. Sorry, guy, I just don’t believe you…
I believe the reason the computer reads high is to make the driver feel better about the car, of course. Just like how many cars have speedometers calibrated to read the maximum legal error to the high side- early Miata speedos were done that way.
I haven’t checked the MPG against the computer in my car in a few years. It was recalled for a brake switch repair recently, and they also said they did an update “to make it run more efficiently.” Perhaps I should check again.
These are some interesting ideas, I absolutely agree on your points 1 and 6, and I can certainly buy that in a precision / accuracy tradeoff, the car manufacturers would err towards precision – but I’m not quite clear why you think these are inexorably interrelated factors.
If we consider your points 2 and 3, ie cars are designed to indicate higher speed than actual, why would this necessarily impact the MPG calculation? For the car, the indicated speed is an aggregation and reported result of various sensors and calculations with its own intentionally biased error bar, I don’t think it would then take that aggregation and use it in the MPG calculation (with corresponding built-in biased error), rather than simply going bottom up again and using sensors + calculations for calculating MPG itself.
Another implication if that were true (aka all the car’s results are true-ing back to reported MPH) is that actual odometer miles would be consistently biased higher too, due to MPH being consistently overestimated - and I thought odometers were pretty tightly calibrated AND regulated (ie the calibration is tested rigorously). And if odometer miles were consistently higher in the same way, there wouldn’t be that disconnect between methods either, because the mud-hut method relies on odometer miles.
I fully admit I could be wrong, I just don’t know why a manufacturer would make independent sensor chains and calculations dependent on each other’s top-level reported results rather than going bottoms-up on all of them.
On the point 4 fuel consumption correction, do you know more about that? What exactly is it correcting for?
And if you don’t mind my asking, what is the source of your knowledge? Did you work at an OEM or a manufacturer or aftermarket tuning shop, or on ECU control systems or something?
1983 Cadillac: Bang on within 0.5 mpg. So 35 years ago technology existed to get accurate readings.
2006 Chevy: Very close and it was always off just little to the low side, meaning the display showed slightly worse mpg than actual.
2004 BMW: Displayed mpg 10% high. I’m convinced they did it on purpose to make the car look better than it was. Typically showed 17.5 mpg, only got 16.
2012 Ford: Display was about 5% high.
So my anecdotal data shows GM being the most honest.
The first car I had with a trip MPG computer that I tracked consistently for ‘actual’ odometer miles divided by gas pumping reading gallons was a 2005 Lexus GX, and yes that trip computer was consistently high by at least an mpg on highway trips, on a 20-ish MPG highway car so significant error. Based on that experience I’ve never really paid attention to the cumulative MPG calcs of my current 2015 BMW 3. I just measure the gallons the gas station says I put in every fill up. Based on which the car gets significantly more MPG under my foot on highway trips than the EPA rating, which would seem to reduce the incentive to have the trip computer give yet higher results, but again with this car I don’t keep track of what the trip computer says.
Some posts seems to suggest we look at inaccuracies in the ‘simple’ measures also. I guess odometer errors would tend to cancel out between the two methods, assuming the odometer is also how the trip computer measures distance. But I suppose it’s possible gas pumps consistently tell you you’ve bought more gas than you actually have and the fuel flow meter in the car is more accurate. Still seems the simpler explanation is other way around. Whether it’s done on purpose I’ve no idea.
The input that the system cannot measure is tire diameter. It has to be assumed. And if it is correct when the tire is new, then as it wears the car travels less distance for each revolution.
And the computer does not have information on fuel pumped. It has information on injector open time, delivery pressure and Hz. I think it is a case of tolerance stack-up.
So just thinking this through here, if tire diameter variance is throwing off the “miles” calculation in the cathedral method, why isn’t it similarly affecting the odometer reading (used in the mud-hut calculation), thereby canceling out the discrepancy?
I can buy that it’s a case of tolerance stack-ups, although injector open time+delivery pressure * Hz sounds an awful lot like fuel volume pumped into the engine to me.
But certainly, I’ll stipulate there are numerous sources of error in the cathedral method, and rather than averaging out like unbiased errors would, perhaps they are additive because of consistent-directioned knowledge gaps or deliberate bias that the manufacturers want and have introduced for outside reasons.
I don’t have any authoritative information, but manufacturers covering themselves against lawsuits and complaints is always a good bet. Making speedometers read on the high side to avoid speeding tickets, making the gas gauge read a bit low to avoid running out, are things I would certainly consider doing if I made cars.
Yes it’s possible for them to get complaints about MPG readings being optimistic, but all they have to do in this case is mumble something about margins of error and averages and complex systems, and they’re more or less off the hook.
And in truth, from their point of view, being off the hook is what it’s all about. Some take it too far.
And back to the OP, Your mud hut method is not crude. In fact it is perfect. It is the only way of actually measuring MPG instead of calculating it. If you fill one tank up a bit higher then the last fill up and get a slightly lower then correct MPG because of it, it auto corrects the next MPG calculation by starting with a bit of extra gas.
As long as you calibrate the tire and that is dead simple. I do it on all my cars when I change tires and every so often. Just use the mileage signs on the highway and run 10 miles or whatever and compare your odometer reading. A quick calculation shows you how much to adjust each MPG calculation. Right now I need to add 4% to my xB’s mileage reading to correct it.
I don’t think I’ve ever had a car that reports my mpg. I have noticed, as others have mentioned, that the speedometer reads about five mph faster than I’m actually going. I recall this in every car I’ve owned. I’ve got another item along a similar vein, again in every car I’ve owned: they have a funny way of reporting the amount of gas remaining in the tank.
In my current car (2008 Yaris), I go about 275-300 miles between fill-ups. The breakdown of my gas gauge reading usually goes something like this, roughly:
After 100 miles: 7/8 tank remaining
After 150 miles: 3/4 tank remaining
After 200 miles: 1/2 tank remaining
After 250 miles: 1/4 tank remaining
I’ve always assumed the reason for this is that a floater in a weirdly-shaped gas tank can’t accurately measure the height of the fuel in the tank, and there’s no good way to measure the actual volume. But now I’m wondering, is it done on purpose as a “feel-good” measure, like the mpg calculation? If so, why? I mean, it feels good during the first 200 miles, but feels pretty shitty watching the fuel level drop so quickly during the last 100 miles…
Yes like I said before, odometer errors would result in an absolute discrepancy between the measured mpg by either simple method or the car’s trip computer and the car’s real gas mileage. It wouldn’t explain a relative discrepancy between the two methods. The trip computer must also use the odometer or anyway the same variable the odometer uses, wheel revs. Wheel diameter must be assumed by either method, and you could manually apply a correction to either method by measuring the odometer against a known distance. I don’t bother to do that, just measure MPG according to what the gas station says for gallons, what the odometer says for miles.
On fuel flow I guess on the latest cars it is actually the integration of how much fuel the engine control computer thinks it needs to, and assumes it is, putting into the cylinders on each rev. rather than a flow meter in the line from the tank. Fuel can recirculate to the pump or tank in some designs so it’s not simple to measure flow.
On inaccurate measures more generally, like I said I assume my BMW 3’s trip computer mpg isn’t right and don’t pay attention to the cumulative. I sometimes look at the instantaneous mileage as a general idea. But it’s broadly complained about that BMW speedometers, not the odometer necessary (differences of opinion on that) are not very accurate, tend to overstate how fast you’re going several mph at highway speed. It’s presumed to be company policy to avoid the speedometer ever understating speed with variations in wheels etc. Also BMW understates hp on most models. An outfit called Dinan makes tuners for BMW’s, which change the turbocharger mapping to get more boost, that don’t void the warranty (other makers produce similar products for less but which can affect the warranty). IOW BMW is implicitly agreeing the stock boost is unnecessarily conservative. Then Dinan almost always finds the actual pre tuner hp to already be more than the BMW rating, sometimes quite a bit. It’s one reason BMW’s tend to have strangely low 0-60 times compared to their hp per unit weight, even before tuning. So there are various errors in what manufacturers say about their cars, not always unpleasant surprises.
Yeah, it would not work that way. If you are going 10mph, the speedometer is not going to read 15. I drive past a neighborhood sign that tells you if you are going over 25 and what you are actually going. My speedometer reports about 27. So, in theory, if it says 60, I would probably be doing 55 or so. The error is more like a percentage than a fixed number.
There you have it. Winner, winner, chicken dinner!
Actually, auto tank level gages are pretty fair at measuring fuel height from 0-100% (as related to the tank itself), just not volume, (the necessary factor in calculating MPG) for the reason you point out above… Irregular tank geometry.
Same here. We took a road trip this weekend (well, still are on the road trip) and I was comparing the GPS speed with the cruise control speed readout and the analog speedometer, and, at 80 mph, they all lined up. Maybe a few times the GPS was dip down to 79 and up to 81, but, for the most part, all three readings were 80 mph.
As for gas tank capacity and how much fuel I have left, my Mazda 3 can go up to 100 miles before the first tick comes off the gauge. (Each quarter tank is subdivided divided into three bars, with the final quarter tank subdivided into six.) Despite that, the fuel economy reads out fine. Also, when I run to empty, I still usually have about 60 miles at least I can run on it, judging by how much of a tank I have left when I fill up. It’s a 13.2 gallon tank and when I run it down to 0, when I fill up, I’m usually at 10.5-11.0 gallons to fill.