Gas mileage better with a fuller tank?

[Nomad]

I have been somewhat obsessive about tracking my gas mileage on fueleconomy.org and on my Garmin iQue 3600, and I have been noticing an interesting trend. When I refill my tank, the more fuel left in my tank, the better MPG I am getting. Refilling with a quarter of a tank of gas left tends to get me towards 31 and 32 MPG. Refilling with the yellow "get some gas NOW!" light on tends to get me towards 28 and 29 MPG. Is this a phenomenon that other Vibe owners are seeing?I have a 2006 Vibe Base Model.

[Katscan]

How exactly are you tracking this MPG change? To be quite honest this would not make sence as your car is heavier with a full tank than it would be with one that was close to empty. So having less mpg with a lighter vehicle does not make sence to me. It would be like saying that loading your car with bags of cement would improve your gas milage.I constantly get around 30 mpg going from full tank to just before empty. I would get better mpg if I didnt have a lead foot.

[joatmon]

I haven't noticed that, but I have noticed a lot of variation in when pumps shut off. Maybe having more gas in the tank causes the pump to shut off earlier, You could try getting 5 1/4 tank fillups in a row, and averaging the numbers, then get 5 "get gas now" fillups in a row and average those numbers. That would help minimize the pump shut off effect and let kyou kow if ou were really getting overall better gas mileage if you filled up at 1/4 tank

[kcorona]

I've been tracking mileage on my vehicles for the past 3 years and have noticed the same effect on partial fills. It makes no sense to me either. I think joatmon may be on to something with the pump shutoff effect. I believe to get the most accurate mileage tracking you should just try to fill your tank at the same relative point (i.e. 1/4 tank, when the light goes on, 60 miles after the light goes on, etc.), whichever one you prefer, and stick with it. In other words, try to lessen the variation between the amount of gas you put in at each fill as much as possible, but always track from a full tank. For instance, my last 3 fills in the Vibe, in gals.:11.811.4211.61

[ZubenElGenubi]

There are other factors to consider as well.For example, what time of the day do you usually fill up? Cooler temperatures translate to higher density for the gasoline.Coeficient of expansion for gasoline is about 0.0006 per degree Farenheit. All other things being equal, the volume difference between 2 gallons and 5 gallons of gas across 30 degrees of temperature is less than a tenth of a gallon. The difference of 3 MPG across a tank is anywhere from .75-1.25 gallons; a lot more than the coefficient of expansion can account for.You also have to remember that fillups are inconsistent; there is no absolute stopping point from tank to tank. You have to measure over numerous fillups before you can see a trend.

[Katscan]

Quote, originally posted by kcorona »I've been tracking mileage on my vehicles for the past 3 years and have noticed the same effect on partial fills. It makes no sense to me either. I think joatmon may be on to something with the pump shutoff effect. I believe to get the most accurate mileage tracking you should just try to fill your tank at the same relative point (i.e. 1/4 tank, when the light goes on, 60 miles after the light goes on, etc.), whichever one you prefer, and stick with it. In other words, try to lessen the variation between the amount of gas you put in at each fill as much as possible, but always track from a full tank. For instance, my last 3 fills in the Vibe, in gals.:11.811.4211.61Thats odd, when I fill my tank at empty I usually see it stop at 10.5 gals. Might be that pump stoppage thing.

[Mavrik]

when the pump stops, I can usually squeeze in another gal or two so from practicly empty I would have pumped in 11-12gal. But I hardly let my tank get that low.

[jake75]

Last time I filled up the pump stop failed to work. Luckily it was pumping slowly and I was right there so only a small amount of gasoline spilled over, but the "fill-up" seemed to be about a 1.5 gallon more than usual given the gas guage showed 3/8 full when I started, and in calculating milage from the last fill-up it was low, i.e. would have expected about 26 mpg and calculated 21 mpg. 1.5 gal seems like a lot - maybe the gas pump was out if wack in measuring as well.

[Fincloser]

remember with such a small tank and a vehicle that gets so many miles on a gallon 1/2 gallon difference on 10gallons of gas would equate to a 2.5mpg diff between tanks. That is on half a gallon not to mention expansion of the fuel. For that reason toyota started usong fuel bladders on the prius stating fuel capacity is 10.6 to 11.4gal.

[ZubenElGenubi]

(Post 1 of 2)For grins, I created a plot of my calculated MPG per tank versus the total amount of gallons per tank. I threw out some atypical events (ex., long driving trips and multiple-fills) to try and see a sample of my regular driving.This first plot shows 97 points (fillups) and an exponential trendline (note formula and R-squared value). It does seem to indicate a downward trend in MPG as more gallons are filled in each tank.This seems counterintuitive and I was quite surprised with this result.

Attached files

[ZubenElGenubi]

(Post 2 of 2)So, I decided to narrow the scope of the results to fillups of 9 gallons or more (again, only single-fillup tanks). This reduced the sample size to 90 points (fillups).Again, there is a noticable downward trend - about 2 MPG from 9 gallons to 13 gallons.I have no guess as to why this is the case.

Attached files

[Mrizzle05]

too much math for me.... y= ??? and R squared to the 3rd power, what? Go Zuben... you math wiz.

[ZubenElGenubi]

Quote, originally posted by Mrizzle05 »too much math for me.... y= ??? and R squared to the 3rd power, what? Go Zuben... you math wiz. Don't sweat the math; I'm using a built-in feature of Excel.Basically, the lower the R-squared number, the better the fit of the trendline to the data. This particular type of trendline (exponential) had the best fit of all types I tried.(EDIT: This is WRONG...the close R-squared is to 1, the better the fit (see RIT post below. GR)

[RIT]

Wrong - you want to have a high R-squared. It is a measure of the amount of variation in the data that you are removing with the trendline. The r-squared that you are getting is saying that your trendline is pretty much useless. You need to include other variables in your equation to explain the variation.

[joatmon]

obviously, the more gas you put in for a certain distance, the lower MPG Here is my Zubenstyle chart, about 250 data points, the red trendline is the exponential, the green one is polynomial zuben, I see a data point where you put in over 14 gallons. No wonder you got bad mileage for that fill up (You might want to throw out those 14 gallon data points as invalid)

Attached files

[ZubenElGenubi]

Quote, originally posted by joatmon »zuben, I see a data point where you put in over 14 gallons. No wonder you got bad mileage for that fill up (You might want to throw out those 14 gallon data points as invalid)Yep..that was a legitimate fillup...even have a receipt for it:

[prathman]

I think there are too many extranious variables to draw any conclusions.For example, you might change how low you let the gauge get on a long highway trip versus just driving around town. So the results might reflect the difference between highway and city mileage rather than being a result of how empty the tank was.

[joatmon]

Quote, originally posted by ZubenElGenubi »Yep..that was a legitimate fillup...even have a receipt for itI wasn't questioning the truth of the data point, just that putting that much gas in a 13.2 gallon tank is a fluke that will affect the accuracy of your trendline. That fillup you got 25.4 mpg. What mpg did you get on the tank after that one?

[engineertwin2]

Remember, the density of gasoline is greatly affected by temperature, so if you filled up your tank on a hot afternoon the pump may short change you as much as 1/3 of a gallon, since fuel stations don't correct the volume to a standard 60F volume.Also, fuel pumps can be inaccurate (often times giving you more gas than what you pay for). This begs to question the quantity (and accuracy) you are receiving and using for calculations, unless you were to always use the same pump at the same station at the same time of day (or the time which correlates to the same temperature - perhaps filling up in a garage?) then you will always have error in your data.And Glenn, I'm not too sure about you using a linear-regression correlation to validate an exponential fit. It does tell us that there is only a 14.4% chance that there is a linear (or one-to-one) relationship between the data points and this exponential curve. With the scatter in your data, I believe that 14.4 percent is as close as you are gonna get, but I'm curious what the r-squared value was for a linear regression (just as a reference)?Very intriguing indeed. At any rate, if any given pump is required to be within 0.001 gal of accuracy and 1 out of every 20 pumps is out of spec, can we assume it all averages out in the end? I'd say so...but only if we look at blocks of data that you could logically breakdown.Finally (sorry for such a long post), I'm curious to hear people's thoughts on what would cause a reduction in mileage?We already acknowledge it is counter-intuitive since the car is lighter with an empty tank...thoughts?

[shanemade]

ok so I am no math wizard but here it goes anyway, if i wanted to get a realistic mileage history here is what i would do.... Go to the pump when fuel light comes on, say try to hit a station within 3 miles of it coming on. Put, say 7 gallons in it drive it till the light comes on again and repeat say 5 times consistently and exactly +/- on the three miles but try to keep it as close as possible. and see what happens. I know there will be the variables as discussed already, but for ****z and giggles why not try. These cars get great fuel economy and 2 mpg is not going to break the bank. I drive way to inconsistently to even consider this anyways, I just love driving.

[ZubenElGenubi]

Quote, originally posted by RIT »Wrong - you want to have a high R-squared. It is a measure of the amount of variation in the data that you are removing with the trendline. The r-squared that you are getting is saying that your trendline is pretty much useless. You need to include other variables in your equation to explain the variation.Yep, I was absolutely wrong!! R-Squared will fall between 0 and 1 and the closer to 1, the more accurate your trendline is (right RIT?).Now, I do take exception that this trendline is completely useless. You seem to imply that there is such a disparity of data that no trendline has any value.

[ZubenElGenubi]

Quote, originally posted by joatmon »I wasn't questioning the truth of the data point, just that putting that much gas in a 13.2 gallon tank is a fluke that will affect the accuracy of your trendline. That fillup you got 25.4 mpg. What mpg did you get on the tank after that one?The next tank was 13.713 gal and 400.4 miles (29.2mpg). Probably hot both of those fillups, ya think?

[joatmon]

Quote, originally posted by ZubenElGenubi »The next tank was 13.713 gal and 400.4 miles (29.2mpg). Probably hot both of those fillups, ya think? Your car is really thirsty then Here's another view of my data, this time US miles per gallon compared to US dollars per gallon. THis graph's trendlines would indicate I get better gas mileage out of more expensive gas. It also indicates that when I first bought the Vibe in 12/02, I was paying US $1.339/gallon

Attached files

[engineertwin2]

Quote, originally posted by joatmon »THis graph's trendlines would indicate I get better gas mileage out of more expensive gas.Do you think that this data would be skewed? I mean, when I see gas prices at $3.30 per gallon I drive significantly more conservative than I did at $2.40 a gallon. Hence you intentionally get better gas mileage our of more expensive gas...And since I haven't fereted any responses just yet (I know, it's still early) regarding what people think may cause the reduction from getting better mpg's when the tank is more full, I'll try one...The design of vapor capture devices on modern pumps recirculates fuel vapors as well as fuel back to the storage tank. Now, when liquid gas flows through this capture device, it triggers the automatic shutoff valve (ASV). Older ASVs tend to stick open when operated infrequently or when the previous customer topped off. So test it out by using pumps that were just used by someone else who didn't top off...this should result in higher mpg calcs...but buyer beware the point calculations and attempting to extrapolate data from a single point calculation..

[ZubenElGenubi]

Quote, originally posted by engineertwin2 »And Glenn, I'm not too sure about you using a linear-regression correlation to validate an exponential fit. It does tell us that there is only a 14.4% chance that there is a linear (or one-to-one) relationship between the data points and this exponential curve. With the scatter in your data, I believe that 14.4 percent is as close as you are gonna get, but I'm curious what the r-squared value was for a linear regression (just as a reference)?Great post, engineertwin2. Regarding the trendline fit, I threw out polynomial trends intuitively. Just didn't see how this kind of physical phenomenon could be described with a polynomial formula. And, when I tried 2nd-6th order polys, they had horrible fits.Here is the last bit of data showing linear, logarithmic, and exponential fits to the data. Note that they are very similar, with the exponential curve having the best R-squared value. Yeah, it's not ideal, but still interesting that the three curves are so similar in this range of data, no?

Attached files

[engineertwin2]

Well Glenn, based on what you just showed me, I'd have to say that RIT is right (sort of). From this data you can prove that there is no direct correlation between fuel economy (mpg) and gallons consumed.In fact, if you study combustion engines and fuel economy, you'll see that fuel economy is often mapped on top of a gear plot. This fuel economy plot very closely resembles a topo-map, with varying concentric shapes that indicate peaks (high mpg) and valleys (low mpg). The gear plot shows what speeds are capable in which gears and subsequently show the approximate mpg you should acheive.I'll see if I can't find an example plot from one of the ol' text books...

[engineertwin2]

Well, I don't have the textbooks (they are in storage) and couldn't find it through a google search, so here is a rough idea.Typically, the fuel economy (or 'elevations') occupy the whole plot underneath the blue curve, but I don't have the time to devote to making a nice fake one - this will have to do.Basically, this plot will always tell you what you already know - that you get the best efficiency by operating the car in the highest gear possible at any speed (to reduce engine revs).You (not you per se, but people in general) should remember that the most significant factors affecting mpgs are mass and speed. When the fed reduced federal highway speeds to 55 mph max, that was a direct result of the fuel crisis of the 70's. Not so coincidentally, most drivetrains have a peak efficiency somewhere near 55 mph. That maximized fuel economy.I can hear it now - "Why not just build a drivetrain that is more efficient at 75 mph?" Well, the average daily driver in the US doesn't operate a car above 55-60 mph. The trade-off of designing a drivetrain for peak efficiency at 75 mph is the poor efficiency of the drive train that would result at lower, everyday speeds. The drivetrain designed to travel at 55 mph serves the country as a whole, despite some commuters who do indeed average 70-75 mph every day.

[joatmon]

Quote, originally posted by engineertwin2 »Do you think that this data would be skewed? I mean, when I see gas prices at $3.30 per gallon I drive significantly more conservative than I did at $2.40 a gallon. Hence you intentionally get better gas mileage our of more expensive gas...Actually, I think all the trendline things are inherently inaccurate since there are so many factors affecting MPGs, looking at just one factor is unreliable. However, I think the effect of gas price on efficient driving is not necessarily the absolute cost per gallon, but the cost compared to recently previous costs. For example, right now I have been paying between 2.80 and $3 per gallion for a couple of months. If the price drops to $2.68/gallon, that's still double what I first paid for Vibe gas, but relatively, it would be a drop and I'd probably punch the gas more often. Here's a plot of $/gallon vs. time, the trend line shows a definite increase. Quote, originally posted by engineertwin2 »And since I haven't fereted any responses just yet (I know, it's still early) regarding what people think may cause the reduction from getting better mpg's when the tank is more full, I'll try one...The design of vapor capture devices on modern pumps recirculates fuel vapors as well as fuel back to the storage tank. Now, when liquid gas flows through this capture device, it triggers the automatic shutoff valve (ASV). Older ASVs tend to stick open when operated infrequently or when the previous customer topped off. So test it out by using pumps that were just used by someone else who didn't top off...this should result in higher mpg calcs...but buyer beware the point calculations and attempting to extrapolate data from a single point calculation..Here are a couple of other possible (and unproveable) theories.1. People are psycologically more prone to overfilling the tank when the low fuel light is on. This skews the MPGs for those tanks that take more gas.2. When gas stations are more frequent, people are more willing to let the tank get emptier, but on long highway trips, people tend to get gas before it gets too low. This causes city/around town and lower MPG tanks to be for more gallons than the smaller, highway/better mpg tanks.3. Nomad is in CT, but it's a new car, so summer fillups so far. Zuben is in TX, where it's warmer than average (for genvibers) Gas in the tank is at ambient temp, more often than not, gas in the ground is cooler. Warmer gas in the tank vaporizes faster, and when there is more gas in the gas tank this effect is more pronounced. This increased vaporization of gas causes the pump to shut off faster. When the tank is nearly empty, this effect is minimized. So, gas pumps shutoff quicker when there is already more gas in the tank, and skews the data to look like the previous tank's gas got better mpgs. If this were true, then for people in colder climates, this trend would be inverted in the winter, and the pumps would shut off sooner when there was less gas in the car's gas tank. (sooner being a relative term, meaning pump shutoff would occur at a lower tank fill level).

Attached files

[ZubenElGenubi]

This has been a great thread!!! We've covered physics, statistics, engineering and psychology in less than fifty posts! Pretty amazing stuff.Don't know that I've learned much...except not to spout off about R-squared values!!!

[chemrebel]

Quote, originally posted by engineertwin2 » The drivetrain designed to travel at 55 mph serves the country as a whole, despite some commuters who do indeed average 70-75 mph every day.like me i travel about 33 miles each day in commuting to and from work on the higway, typically going 70-75 mph. plus i spend about 13 miles doing city driving. i figure i'm averaging about 29 mi/gal with that kind of driving. i wish i got better mileage from my vibe but hey i gotta make a living somehow...

[engineertwin2]

Quote, originally posted by joatmon »Gas in the tank is at ambient temp, more often than not, gas in the ground is cooler. Warmer gas in the tank vaporizes faster, and when there is more gas in the gas tank this effect is more pronounced. This increased vaporization of gas causes the pump to shut off faster. When the tank is nearly empty, this effect is minimized. So, gas pumps shutoff quicker when there is already more gas in the tank, and skews the data to look like the previous tank's gas got better mpgs. If this were true, then for people in colder climates, this trend would be inverted in the winter, and the pumps would shut off sooner when there was less gas in the car's gas tank. (sooner being a relative term, meaning pump shutoff would occur at a lower tank fill level).Excellent! I love this logic. It (at first glance at least) appears to be very solid. It could also explain some of the difference in winter gas. I know that (in Arizona) the winter formulation is to reduce pollution (specifically PM10 and PM2.5) problems. But with colder air temperature, the air is more dense, providing more power and also requiring less fuel consumed...several people have reported lower fuel economy in the winter, but mine is lower here in Arizona because I run the AC a lot more in the summer and hardly in the winter...Keep them coming...theories are great...Joatmoan has provided a terrific theory......and I agree with Zuben - this thread has been fantastic! Simply amazing the thoughts...

[Nomad]

Wow, I never figured I would get so much interest when I started this thread, but it appears I am not alone in my experience with differing mileages depending on amount of gas in the tank.I am probably a pretty good test case for now and the future, so I'll keep tracking. I drive pretty much only to work each week, 5 days a week, and try to always fill up at the same (BJ's wholesaler) gas station each week. Being a creature of habit, I almost always even use the same pump. And I tend to fill up as soon as I can once the yellow light comes on. But as someone noted, I am a pretty new owner so my data to date may not be statistically significant... yet. Up to now the only other theory I had for the phenomenon was the weight of gas putting more pressure on the feed when there was more in the tank. But with electronic fuel injection systems, that seemed pretty unlikely.

[coldmm803]

another factor to consider would be how fast the gas is pumped into the tank. A sunoco station by my house that was built within the last yr n a half actually pumps faster than at other stations. if the gas is pumped in faster then there are bubbles being created causeing the pump to shut off with less in the tank, the gage in the car still shows as bein full but you can easily get a lil more in.zuben i'm still puzzled on how your getting 14 gal in a 13.2 gal tank

[ZubenElGenubi]

Quote, originally posted by coldmm803 »zuben i'm still puzzled on how your getting 14 gal in a 13.2 gal tankHey, I wouldn't have believed it unless I'd seen it! I've filled up with more than 13.2 on several occasions. I could easily get another 1-1½ gallons in after the first shut off. Note that the 14 gallon incident occurred in May (in Houston) around 6:00PM...one of the worst times of the day to fill up.I don't do that anymore becuase I'm trying to calibrate my ScanGaugeII to properly calculate MPG and tank capacity.

[jake75]

I had always heard that the reason mileage above 55 was worse had to do with air resistance. Franly, I do not see much difference at 55 and 70. I don't drive over 70.

[Nomad]

One thought did occur to me on this topic. Most theories put forth so far are not Vibe-specific. So, if they are correct then this phenomenon should logically not be Vibe-specific either. I never noticed this same effect with my 2000 Oldsmobile Alero, but I was not monitoring the MPG as closely as in my new baby.Are people seeing the same basic trends in their non-Vibe cars? i.e. Better gas mileage on fill-ups with more in the tank?

[joatmon]

Quote, originally posted by Nomad » Refilling with a quarter of a tank of gas left tends to get me towards 31 and 32 MPG. Refilling with the yellow "get some gas NOW!" light on tends to get me towards 28 and 29 MPG.math ->Fillup the 13.2 gallon tank, then when the gauge is at 1/4, that means you've used 3/4 of the 13.2 , or 9.9 gallons. 32 mpg on 9.9 gallons means the car moved about 317 miles. Assuming the low fuel light comes on at 1/8 tank (big assuption, mine does not come on reliably at the same level) 7/8 of the 13.2 gallon tank used, or 11.5 gallons to get 29 mpg. 29 * 11.5 means that car moved about 335 miles. This means that if you get the 32 mpg for the first 9.9 gallons and 317 miles, the next 1.6 gallons only push the car 18 miles, giving you 11.25 mpg for the time you are between 1/4 tanks and low fuel light. If you get only get 28 mpg for the tank when filling at the low fuel light, then the car only went 322 miles on 11.5 gallons, so you only get 5 miles on the extra 1.6 gallons, or 3.1 mpg after passing a quarter tank. If the first 3/4 tank is at 32 mpg, and the low fuel light is down to 28 mpg for the entire tank, but say the low fuel light comes on at 12 gallons used, then that extra 2.1 gallons only gets you one extra mile.The only way I can think of consistently getting 11 mpg (or 3 mpg ! ) out of a Vibe is if there is a hole somewhere and half the gas leaks out onto the ground. You could probably drive in such a way to get 11 mpg if you tried, but I assume you don't. I still think that the best way to test the hypothesis that filling up at low fuel light gets worse gas mileage for the tank is to do drive consistently, and fillup at a certain fuel gauge level repeatedly (five or ten tanks in a row), then average the numbers together, and then repeat the process for a different fuel gauge level, and compare the averages.

[ZubenElGenubi]

Quote, originally posted by Nomad »One thought did occur to me on this topic. Most theories put forth so far are not Vibe-specific. So, if they are correct then this phenomenon should logically not be Vibe-specific either. I never noticed this same effect with my 2000 Oldsmobile Alero, but I was not monitoring the MPG as closely as in my new baby.Are people seeing the same basic trends in their non-Vibe cars? i.e. Better gas mileage on fill-ups with more in the tank?Glad you asked! It never occurred to me to check the stats on my wife's car (2001 RAV4) and here's what I got:So, you can see, even if I throw out the high and low fillup values, there is still a large area bounded between 9 & 13.5 gallons that appears to have no trend whatsoever. Even the stock linear trend that Excel created has such a poor fit that it can probably be discounted (and even if you were to consider it, the trend is opposite that of my Vibe!).

Attached files

[ZubenElGenubi]

Quote, originally posted by jake75 »I had always heard that the reason mileage above 55 was worse had to do with air resistance. Franly, I do not see much difference at 55 and 70. I don't drive over 70.This one of the few things I do know about.All other factors being equal, the force of drag is proportional to the square of the velocity. At 70 mph, the drag force against the car is 62% higher than at 55.Quote, originally posted by me...earlier thread »The drag force on any moving object is proportional to the square of the velocity. Here's the formula:D="Drag" force S ="Frontal" area (cross-sectional)V ="Velocity" = Density of airCd = Drag coefficientI definitely see a difference in load and mpg values on my ScanGauge between 55 and 70 mph. Heck, I noticed a difference driving with and against the wind on an errand run of a couple miles.