CarTalk.com Blogs Car Info Our Show Deals Mechanics Files Vehicle Donation

An observation regarding onboard MPG readouts

Several times, in various threads, people have been told that they should not rely on the onboard MPG readouts on their dashboard, due to a lack of accuracy from these devices. I am currently driving my first car that is equipped with one of these devices, and after almost 2 years with this car, I finally decided to verify the accuracy of that readout.

The result, over several tanks, was that the onboard readout IS optimistic…to the tune of 1/2 of 1%.
Ergo, accurate enough for most people’s uses.

For example, on my last tank of gas, the dashboard MPG readout was 26.1 mpg.
When I calculated my actual mileage with paper and pencil (and with a calculator, just to verify things) the “real” figure was 25.85 mpg–a difference that is, IMHO, statistically insignificant.

So, while I can’t speak for the accuracy of these devices in general, I can tell you that I have found mine to be…acceptably accurate.
What type of experience have other forum members had with these devices on their own cars?

I’ve had a couple of cars with in dash mpg readouts and I consider them accurate. Current car with a readout is an '01 Sequoia which can provide instant mpg and overall mpg since the last “zero out”. The DTE distance to empty is somewhat suspect IMO but the mpg readouts are good. I had an '00 Saab and a '98 Dodge Intrepid that had trip computers and were also accurate.

Not sure about all cars, but in general I think these readouts are accurate if the driver zero’s out the readout and starts fresh every tankful or so. I observe these readouts vary quite a bit based on conditions, such as interstate trips my Sequoia gets 20-21. But as soon as I mix in any stop and go around town driving it drops quickly. All in town driving gets at best 14. Wind is another factor, temps in the low '20’s hurts mpg, towing really drops the mpg down to 11-12 mpg. Saw similar fluctuations with the other cars.

I agree with Uncle Turbo. While I find the mpg readout to be accurate, the DTE readout is overly pessimistic. However, I think that this pessimism is probably part of the system’s algorhythm, in order to keep as many drivers as possible from running out of gas on the highway.

When I have run the gas tank very low in order to hold out until I can get to the gas station that always has the lowest prices, the DTE readout might be telling me that I only have ~40 miles until the tank is empty, but when I fill the tank it is obvious that there were still about 3-3.5 gallons left in the tank. When I have been averaging 24-28 mpg, those 3+ gallons would actually give me ~70 miles–or more–until the tank was empty, but this type of little lie should help to keep people out of trouble.

My readout is also accurate to a few %, plenty good enough for me. Others I know of are about 5% optimistic.

Years ago I actually designed a system to do exactly this for Naval application and it is actually fairly complicated system to do well.

To measure MPG you’d have to measure over a long time and sort out speeds because you get worse mileage at lower than higher speeds. You can’t just average over time because the results will be totally bogus when you speed up and slow down.
They’d basically have to calculate the mpg for - say - every 5 mph speed band (depending on whatever error you accept between each band) and store it, storing mpg and time it took to calculate it.
Then, while you’re driving at constant speed, the algorithm would need to go in that history bucket and average that data, giving the data that took the longest the most weight. That makes sense because if you’ve measured something longer, the data is more accurate than data that’s only been recorded briefly.
For the scientific types here, maybe it would need to use some sort of means square averaging calculation or something like that so you still get a decent read out response.

I don’t think the cars we own do that because the mpg is all over the place, certainly when you slow down or speed up. Even at constant speed, it takes a while to integrate out the errors and outliers so it comes up with bogus results until it has been running for a bit.

There’s that^ to contend with, and also the error of each component of the system because it is accumulative. Speed is probably pretty good within 1 percent but can one really measure the current level of the gas tank within 1 or 2 percent? I doubt it. Then there’s variability in parts, temperature, etc.
Even if the car has the ultimate accuracy in instrumentation, you’d still have to make up for slosh and non-linearity of the tank shape.

It is an approximation at very best - as in ‘half an axe handle’.

All I think the current systems do is add up the fuel injected (they get that directly from the engine control computer) and divide it into the miles covered (they get that from the traction control/abs sensors), both over a given time interval. Mine has an instantaneous readout (that bounces all over the place) and an ‘average since last reset’ value that changes slowly, as you’d expect. I can also display a bar graph with each bar representing the average mpg for a given time interval, like 10 minutes.

Yup, that’s why these systems aren’t accurate at all.
It is actually a fairly complex problem, believe it or not.

I’m very nitpicky at mileage and have monitored the dashboard readouts on my current and past Lincolns religiously during both normal day to day driving and extended road trips.
They’ve been shown to be very accurate and within .1 of a mile per gallon when based against the tank fill/miles driven equation.

I have always used miles per gallon as one indicator of the health of the engine. Even though today’s cars have check engine lights, computer controls, etc. I still pay attention to gas mileage. The mpg readout in both our 2003 4Runner and 2011 Sienna seem to be very accurate when I have compared it with calculating mpg by distance traveled divided by gallons of gasoline consumed. However, I use the dash readout as a relative indicator. If it begins indicating a lower mpg than I had normally been attaining, I start to look for the reason why.
Given a choice, however, I would prefer to have an oil pressure gauge and an ammeter to the mpg readout. I can calculate gasoline mileage, but an instantaneous readout of oil pressure or current flow is much more important.

I record my odometer reading and reset it when I fill it. By dividing however many gallons you fill with, you get your average mpg.
It is within give or take two axle handles because it doesn’t take speed into account but if you see a change, something is wrong or you’ve been changing driving habits.

Like Triedag, I use my mpg figures as sort of a quick & dirty way of determining the health of my engine.
If I see a drop in mpg that can’t be explained by a lot of stop & go driving or mountainous terrain, that is an early signal to start looking at whether my maintenance is up to date and/or if something is amiss with the engine.

On all of my previous cars, I calculated the gas mileage by hand, after each fill-up.
Now that I see how accurate (consistently ~ 1/2 of 1% error) my onboard readout is, I can dispense with the traditional methods, and just rely on the very accurate electronic system for this assessment.

Fuel economy is calculated from fuel injector pulse-width, engine speed and distance.
From the warranty complaints I’ve seen over the last twenty years there have been a few models with software updates to correct the display data. Most complaints were due to a misunderstanding of how to use/reset the display data.

The systems are fairly simple…the computer knows how many miles were traveled and how much fuel was injected to cover those miles…

I have a 2004 Monte Carlo that is never right. My calculation is always 2 to 3 MPG less than the trip computer. I zero out at every fill-up. What I believe is happening with mine is the actual fuel used calculation is always wrong. I can put 14 gallons in but the car “thinks” I have only used 12.5 gallons. That is the measurement that I suspect causes my MPG estimate to be off.

my 300m hasthe same thing its dte & low fuel light our pre mature somebody dosn’t want me letting the tank get that low on fuel.

Sounds like it’s miscalibrated. Most don’t read the amount of gas you put in the tank, just the amount injected into the engine.

I think the biggest benefit of the calculators is serving to show consumption on driving habits. I like to reset it on my wifes van, because as time goes on the average moderates. I love seeing 60 mpg as I am coasting to a stop.

+1 on Barkydog’s comment.
It really is only meant to be a reference. I suspect they don’t want those things to be too close to the actual mpg because they’d get calls from anally retentive people that get 25.4 miles to the gallon as opposed to the specified 26 mpg. It is in their interest to have some obfuscation in the system.

On my CRV, it is a bit optimistic, misses by ~ 2 mpg. It is helpful though for trending and also to see how driving habits changes the number. Angry days and nasty traffic jams now have a numeric attached to them. I have trip A left untouched since I bought the car and reset trip B with every fill-up.

The mpg readout is a good way to monitor the driving habits of a teenage driver. When my son was a teenage driver, we had a 1990 Ford Aerostar with a mpg readout. Without him knowing it, I would reset the mpg indicator before he took the Aerostar and then after he came back, I would check it again. The embarrassing thing was that he would get a higher mpg reading than either my wife or I would get. However, it was great to know that he wasn’t doing jack rabbit starts.