Hacker News Reader

Comment by jakobegger | Hacker News Reader

Doesn‘t anybody find it curious that the source of data are range estimates displayed on the cars built-in dashboard?

For all we know, Teslas might be programmed to show more „optimistic“ estimates as the batteries get older.

I‘m not saying this is actually the case, but these stats are effectively manufacturer-provided numbers. Would you trust data like this if it came from any other car manufacturer?

If you have a gasoline car, do you trust the fuel consumption rate displayed on the dashboard? Should journalists trust it? Or should they measure how much gas they put in the tank and divide it by distance driven?

View on HN

When I was reverse engineering car ECUs, one of the interesting little projects was finding fuel consumption display calibration data and calculation algorithm, on 2008+ Subarus. And then I did some measurements trying to get accurate real world consumption, and "sync" the dashboard with it. And while there was ~3% error IIRC, I think it has more to do with getting the most accurate average consumption over the car's lifetime, than fooling the customer on purpose. The calculation is based on the effective injector pulsewidth (ECU can't count actual fuel flow))), and as injector performance deteriorates over time, you get less and less optimistic values with more miles.

> ECU can't count actual fuel flow

That doesn't seem right. Flow sensors exist and could be fitted to the fuel line to the engine for a precise measurement of how much fuel is consumed.

So I'm assuming manufacturers just estimate fuel flow in the ECU because a flow sensor would be additional cost for negligible benefit?

Never saw a separate fuel flow sensor in an automotive application, but fuel injector is technically an orifice flow meter in this calculation method (pressure differential is constant and you calculate how much fuel is flown through orifice by measuring time it is open). Sure, when orifice size is reduced by lacquer depositions or dirt it will cause measurement error. But it will also cause increasing fuel trims (to the point when CEL goes on) and bad driving performance when fuel trims don't apply

> because a flow sensor would be additional cost for negligible benefit?

Exactly. It's basically unnecessary given the other sources of information that the ECU has at its disposal.

It knows the target fuel pressure (some cars even have fuel pressure sensors) and it knows the flow characteristics of the injectors with respect to pressure and pulse width. That alone is enough for a decent calculation of fuel mass flow.

In addition to this, the ECU is also calculating the mass air flow into the engine, either directly via a MAF sensor or indirectly via manifold pressure, VE, and RPM (known speed density).

When it goes to target a specific air/fuel ratio, this is all the information that is really required. However, since injectors tend to deteriorate (clog, stick, etc.) over time, this initial calculation is usually “trimmed” by means of feedback from the oxygen sensor(s).

At the end of the day, there’s a fair amount of information available to the ECU which can be used to refine the fuel flow calculation.

The fuel consumption calculation on my 335i is usually within 0.5 MPG. However, my 2002 Maxima was about 3 MPG off.

> ECU can't count actual fuel flow

Thanks, this is pretty interesting.

It's the first non-bullshit explanation I've seen for why cars tend to display inaccurate average MPG on the dash. I always assumed there was a fully accurate measure of fuel flow into the engine and there was something else broken (or dishonest) about that calculation.

This has come up in the press occasionally and the manufacturers never admit that they don't have a totally accurate measure of flow: https://www.edmunds.com/fuel-economy/your-fuel-economy-gauge...

I currently drive a 2015 Opel Astra. It shows an average of 5.1 liters per 100 kilometers (sorry, German, so it's metric units for me) over multiple months of driving in the dashboard. At the same time, I'm tracking every one of my gas refills using an app which also calculates consumption averages over arbitrary timespans. By that data, I am actually consuming about 6 liters per 100km, which is an almost 20% deviation. I thus stopped to trust the fuel consumption measurements of this particular car.

Before that, I had an old and small 2005 Peugeot 206. The measurements of that car were like exact matches (at least to the first decimal digit) with my manual consumption tracking.

What really baffles me is that the newer and more expensive car is the one with the huge deviation, not the older and cheaper one, as you might expect.

I'd put this in a typical market drift. New generation are built with a lazy cheap mindset, not the old generation "lets try to make something good (probably because there was more love and motivation for said product at that time)".

An independent third party can do accurate measurements and determine the difference between the dashboard measurement and the actual capacity; I like to think that if the difference was too high, people would've noticed by now. If the difference is too high and beyond the range of a simple measurement error, then I'm pretty sure it's fraud like VW's emissions scandal and Tesla would get millions if not billions of fines from the EU. But I'm sure Tesla wouldn't let it get that far - it could probably be fixed via a software update.

I fill up every time I refuel and I take notes of the quantity and odometer and the average matches the dashboard.

What manufacturer and model is that?

I had a 2008 VW Golf TDI Gt Sport (the 140PS PD engine) and that consistently lied. My newer 150 Euro6 (I think) engined VW Beetle TDO gives even better MPG so perhaps it is lying again... Potentially time to measure it versus how much I put in.

How do you measure the distance driven? Perhaps the tachometer is also lying?

Very good point. Odometers (tachometer measures RPM, odometer measures distance) are not as reliable as some may believe, and even if they're perfectly accurate, there are so many variables that it's hard to control. Smaller tires spin faster, which drives up the miles read. Tires get smaller as they wear, so the odometer can be calibrated for new tires but will quickly become inaccurate.

But the odds that your odometer is accurate is very slim in reality: http://www.azfamily.com/story/19769079/tests-show-odometers-...

I take a look at my estimates in the dashboard on my Mazda2 (Germany) and also calculate the metrics manually every time I fill the gasoline. Funny thing is, that I tend to use less in the real world then my car makes me believe.

My car tends to value the long range drives on a highway way higher then my average driving habit while commuting. So the higher consumption outliers on the highway tend to skew my results.

There is a website you can use to keep track of this: http://www.fuelly.com/

I trust my range indication in my petrol car to be conservative. It has about 30km more range than indicated - imagine the uproar if you planned a journey it said you could do but ran out of fuel? It’s quite handy as there are some areas I’ve driven through have no fuel stops for over 100km, and I’m sure the US has quite a few areas like that.

That's because of the "fuel reserve" in your gas tank. You can Google that term to learn more.

I have a Volkswagen Golf GTE. My dashboard claims it can run ~47km on a charge, and it adjusts whenever you change your AC settings. When I bought it, VW claimed 50km of range.

In the 60.000km's I've driven that car, I NEVER were able to get even close to 40km on a charge.

I know it's anecdata, but I found the dashboard estimate on my VW (heh) strangely correct. It's off by +/- 5% in most cases. Since it doesn't take upcoming traffic/terrain/wind into account, I find it more than reasonable.

Anecdotally, I run a 2016 model year ford fusion 4cyl turbo and its fuel economy gauge has never gotten above 14.5mpg in the year I've had it. Not exactly numbers to boast about. I trust it.

I trust the fuel consumption in my car, yeah. If it says I've been using 12l/100km on average and I've travelled 500km, you can bet I'm filling up exactly 60l at the pump.

That really depends on the car. I don't own a car so often use rental cars. Most cars' consumption is within 10% of the shown values but at least every fifth car I drive has vastly different numbers.

I think it is not necessarily due to the car but because of the calculation methods they use. If you drive motorway and city, consumption is accurate. But as soon as you start driving in the mountains or a mix of different terrains, data often vastly differs.

If the car has a Mass Airflow Sensor (MAF), then it can more accurately calculate the fuel usage. The majority of older / cheaper cars use a Manifold Absolute Pressure Sensor (MAP) and have to make some guesses at what the actual fuel usage is.

I've had a car that consistently showed 0.5 l/100 km less than the real consumption.

The Tesla is strange because that number on the dash is "rated miles", or the best you're going to get. So you can't just drive to empty and get a number you can compare with others, unless you drive it in quite a certain way.

My S60 says something like 215 miles when fully charged, but when driving on the Interstate the actual range is more like 160 miles, with the cabin heater on. It can vary significantly due to rain as well, a recent trip that had a leg almost entirely in very heavy rain caused a significant drop in range.

Can't you measure how much energy is used during charging?

were you driving it manually, or using Autodrive? Were you driving it conservatively, or heavy on the accelerator?

Did you stay at a consistent 55 - 60mph, or were you going faster?

There are a lot of factors in play for a electric vehicle.

Reply on news.ycombinator.com