FPGA Central

Hi folks.

Last night a friend of mine called. He was a bit upset, as his car
had stopped, with no warning, in the middel of nowhere. He had
waited for an hour for the tow truck, but it never arrived. He had
called somebody to come over and help tow his car to town, but
it turned out his car did not have a tow-rope attachment but some
attach-a-weird-bolt-to-the-bumper arrangement which he did not
understand how worked.

After a couple of hours at the roadside trying frantically to get
out of there, and with no chance to do road-side repairs, he tried
the only thing he could: Fill another 10 liters og gas on the tank,
and see what happened. The car started and he could drive the
10 km to the gas station and fill up.

So what has this story to do with computers?

It turned out that my friend had trusted the car computer which
informed him something like "XXX km to next refueling", leading
him to believe that he could actually get to town on the onboard
fuel.I have no idea why the thing displayed the wrong message,
the fuel level sensor might be proken or the remaining distance
might have been computed based on invalid statistics, but my
friend is not a n engineer, let a lone a computer programmer,
so he trusted what the computer told him. And was caught
completely by surprise when the car stopped.

Why is this computer potentially lethal?

My friend got away from this inciden whith no harm, as it
was not very cold, some -2C to 0C. In three months that
would have been a very different story. The place where the
car broke down is 'weird' in that it is close to the coast but
easily drop to -20C and below in winter, not counting wind-
chill in near-gale-force winds which are common in the area
in winter.

Had the incident occured in mid-winter with an unprepared
person (not bringing thermo suits or thermo blankets in the
car), this computer glitch could easily turn very nasty.

As far as I am concerned, the estimating-the-distance-left-
to-refueling is a gadget is best left out, since the driver will
handle the fuel in a different way, accounting for uncertainties,
if he does *not* get (unreliable) info from the computer.

In other words, this is a classic case of "no info is better
than wrong info."

Rune

Rune Allnor wrote:
> Hi folks.
>
> Last night a friend of mine called. He was a bit upset, as his car
> had stopped, with no warning, in the middel of nowhere. He had
> waited for an hour for the tow truck, but it never arrived. He had
> called somebody to come over and help tow his car to town, but
> it turned out his car did not have a tow-rope attachment but some
> attach-a-weird-bolt-to-the-bumper arrangement which he did not
> understand how worked.
>
Rune,
Your friend is nieve.
http://news.bbc.co.uk/1/hi/england/cumbria/7366371.stm
Bless, Syms.

Rune Allnor wrote:
> Hi folks.
>
> Last night a friend of mine called. He was a bit upset, as his car
> had stopped, with no warning, in the middel of nowhere. He had
> waited for an hour for the tow truck, but it never arrived. He had
> called somebody to come over and help tow his car to town, but
> it turned out his car did not have a tow-rope attachment but some
> attach-a-weird-bolt-to-the-bumper arrangement which he did not
> understand how worked.
>
> After a couple of hours at the roadside trying frantically to get
> out of there, and with no chance to do road-side repairs, he tried
> the only thing he could: Fill another 10 liters og gas on the tank,
> and see what happened. The car started and he could drive the
> 10 km to the gas station and fill up.
>
> So what has this story to do with computers?
>
> It turned out that my friend had trusted the car computer which
> informed him something like "XXX km to next refueling", leading
> him to believe that he could actually get to town on the onboard
> fuel.I have no idea why the thing displayed the wrong message,
> the fuel level sensor might be proken or the remaining distance
> might have been computed based on invalid statistics, but my
> friend is not a n engineer, let a lone a computer programmer,
> so he trusted what the computer told him. And was caught
> completely by surprise when the car stopped.
>
> Why is this computer potentially lethal?
>
> My friend got away from this inciden whith no harm, as it
> was not very cold, some -2C to 0C. In three months that
> would have been a very different story. The place where the
> car broke down is 'weird' in that it is close to the coast but
> easily drop to -20C and below in winter, not counting wind-
> chill in near-gale-force winds which are common in the area
> in winter.
>
> Had the incident occured in mid-winter with an unprepared
> person (not bringing thermo suits or thermo blankets in the
> car), this computer glitch could easily turn very nasty.
>
> As far as I am concerned, the estimating-the-distance-left-
> to-refueling is a gadget is best left out, since the driver will
> handle the fuel in a different way, accounting for uncertainties,
> if he does *not* get (unreliable) info from the computer.
>
> In other words, this is a classic case of "no info is better
> than wrong info."
>
> Rune

Not to mention "marketing info is worse than no info", and "everything
my car tells me is marketing".

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

Symon wrote:
> Rune Allnor wrote:
>> Hi folks.
>>
>> Last night a friend of mine called. He was a bit upset, as his car
>> had stopped, with no warning, in the middel of nowhere. He had
>> waited for an hour for the tow truck, but it never arrived. He had
>> called somebody to come over and help tow his car to town, but
>> it turned out his car did not have a tow-rope attachment but some
>> attach-a-weird-bolt-to-the-bumper arrangement which he did not
>> understand how worked.
>>
> Rune,
> Your friend is nieve.
> http://news.bbc.co.uk/1/hi/england/cumbria/7366371.stm
> Bless, Syms.
Like my drunken spolling

On Tue, 30 Sep 2008 17:18:09 -0700 (PDT), Rune Allnor
<[email protected]> wrote:

>Hi folks.
>
>Last night a friend of mine called. He was a bit upset, as his car
>had stopped, with no warning, in the middel of nowhere. He had
>waited for an hour for the tow truck, but it never arrived. He had
>called somebody to come over and help tow his car to town, but
>it turned out his car did not have a tow-rope attachment but some
>attach-a-weird-bolt-to-the-bumper arrangement which he did not
>understand how worked.
>
>After a couple of hours at the roadside trying frantically to get
>out of there, and with no chance to do road-side repairs, he tried
>the only thing he could: Fill another 10 liters og gas on the tank,
>and see what happened. The car started and he could drive the
>10 km to the gas station and fill up.
>
>So what has this story to do with computers?
>
>It turned out that my friend had trusted the car computer which
>informed him something like "XXX km to next refueling", leading
>him to believe that he could actually get to town on the onboard
>fuel.I have no idea why the thing displayed the wrong message,
>the fuel level sensor might be proken or the remaining distance
>might have been computed based on invalid statistics, but my
>friend is not a n engineer, let a lone a computer programmer,
>so he trusted what the computer told him. And was caught
>completely by surprise when the car stopped.
>
>Why is this computer potentially lethal?
>
>My friend got away from this inciden whith no harm, as it
>was not very cold, some -2C to 0C. In three months that
>would have been a very different story. The place where the
>car broke down is 'weird' in that it is close to the coast but
>easily drop to -20C and below in winter, not counting wind-
>chill in near-gale-force winds which are common in the area
>in winter.
>
>Had the incident occured in mid-winter with an unprepared
>person (not bringing thermo suits or thermo blankets in the
>car), this computer glitch could easily turn very nasty.
>
>As far as I am concerned, the estimating-the-distance-left-
>to-refueling is a gadget is best left out, since the driver will
>handle the fuel in a different way, accounting for uncertainties,
>if he does *not* get (unreliable) info from the computer.
>
>In other words, this is a classic case of "no info is better
>than wrong info."
>
>Rune

While I sympathize with your friend I'll suggest a different point of
view. We had a discussion here recently how people have given over
their risk/danger management skills to other "authority". People
depend on signs, "authority" figures, or computer outputs to tell them
what to do. I think that's a bad thing and people need to take more
responsibility for themselves.

Most automobiles have had fuel gauges for as long as I've been around,
and I think if someone in a hazardous environment ran a car down close
to E and it quit, people wouldn't have blamed the fuel gauge or the
car manufacturer, they'd have blamed the operator for running it so
close to E in a dangerous environment. A lack of planning is not a
good excuse if the consequences of failure are high. Everybody knows
(or should know) that automotive gauges aren't precision instruments.
I think your friend made a mistake in thinking that because a display
is digital then it must be accurate. What gives one that notion?

I hope he uses the incident to learn (and tell his friends!) that the
gauge is only an estimate and isn't precise enough to trust one's life
with.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

Rune Allnor wrote:
> Hi folks.
>
> Last night a friend of mine called. He was a bit upset, as his car
> had stopped, with no warning, in the middel of nowhere. He had
> waited for an hour for the tow truck, but it never arrived. He had
> called somebody to come over and help tow his car to town, but
> it turned out his car did not have a tow-rope attachment but some
> attach-a-weird-bolt-to-the-bumper arrangement which he did not
> understand how worked.
>
> After a couple of hours at the roadside trying frantically to get
> out of there, and with no chance to do road-side repairs, he tried
> the only thing he could: Fill another 10 liters og gas on the tank,
> and see what happened. The car started and he could drive the
> 10 km to the gas station and fill up.
>
> So what has this story to do with computers?
>
> It turned out that my friend had trusted the car computer which
> informed him something like "XXX km to next refueling", leading
> him to believe that he could actually get to town on the onboard
> fuel.I have no idea why the thing displayed the wrong message,
> the fuel level sensor might be proken or the remaining distance
> might have been computed based on invalid statistics, but my
> friend is not a n engineer, let a lone a computer programmer,
> so he trusted what the computer told him. And was caught
> completely by surprise when the car stopped.
>
> Why is this computer potentially lethal?
>
> My friend got away from this inciden whith no harm, as it
> was not very cold, some -2C to 0C. In three months that
> would have been a very different story. The place where the
> car broke down is 'weird' in that it is close to the coast but
> easily drop to -20C and below in winter, not counting wind-
> chill in near-gale-force winds which are common in the area
> in winter.
>
> Had the incident occured in mid-winter with an unprepared
> person (not bringing thermo suits or thermo blankets in the
> car), this computer glitch could easily turn very nasty.
>
> As far as I am concerned, the estimating-the-distance-left-
> to-refueling is a gadget is best left out, since the driver will
> handle the fuel in a different way, accounting for uncertainties,
> if he does *not* get (unreliable) info from the computer.
>
> In other words, this is a classic case of "no info is better
> than wrong info."

The news tonight included a car that had been hit by a railroad train.
The driver blamed it on his GPS. http://tinyurl.com/4bfhjn

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
** Posted from http://www.teranews.com **

Rune Allnor wrote:
(snip)

> It turned out that my friend had trusted the car computer which
> informed him something like "XXX km to next refueling", leading
> him to believe that he could actually get to town on the onboard
> fuel.I have no idea why the thing displayed the wrong message,
> the fuel level sensor might be proken or the remaining distance
> might have been computed based on invalid statistics, but my
> friend is not a n engineer, let a lone a computer programmer,
> so he trusted what the computer told him. And was caught
> completely by surprise when the car stopped.

Old story. Which one was more potentially lethal?

http://www.iasa.com.au/folders/Safet...mliGlider.html

-- glen

On 1 Okt, 04:03, Eric Jacobsen <eric.jacob...@ieee.org> wrote:
....
> Most automobiles have had fuel gauges for as long as I've been around,
> and I think if someone in a hazardous environment ran a car down close
> to E and it quit, people wouldn't have blamed the fuel gauge or the
> car manufacturer, they'd have blamed the operator for running it so
> close to E in a dangerous environment.

I won't disagree with that, but rather ask why the guy ran so close
to E in the first place. There might be good reasons (he knew the
situation but something made it worth taking a calculated risk)
or stupid reasons (a car computer displaying a message about the
*driving* *distance*, not amount of fuel, left that it can not
possibly defend.)

>*A lack of planning is not a
> good excuse if the consequences of failure are high. * Everybody knows
> (or should know) that automotive gauges aren't precision instruments.

I told him about my own fuel gauge, which doesn't even have a warning
light when the fuel runs low. The way I keep track of the fuel levels
is to keep track of distance travelled since the last refueling.
I zero the odeometer every time I fill up, and know from experience
how long I usually drive on one tank in different conditions (city/
highway, winter/summer). Not a very accurate method as such, but
it keeps my attention to the fuel levels and I fill up before things
get nasty.

> I think your friend made a mistake in thinking that because a display
> is digital then it must be accurate. *

Agreed. But as a computer layman he had no reason to knwo or
suspect exactly *how* inaccurate the thing turned out to be.

> What gives one that notion?

The fact that it is a *computer* that issues the message? The fact
that the number is given to within 100 m? The fact that the car
is less than two years old? And is of a well-known, well-respected
brand?

The regulars on comp.dsp work with computers and know what these
things can and can not do. I doubt anyone here would get into
that kind of situation because they trusted the computer.

However, the innocent computer layman can not be expected to be
very critical with respect to the displays. Which is all the more
reason to blame whoever decided that the computer should display
this estimate to the driver, not the driver.

Rune

On Oct 1, 1:18*am, Rune Allnor <all...@tele.ntnu.no> wrote:
> Hi folks.
>
> Last night a friend of mine called. He was a bit upset, as his car
> had stopped, with no warning, in the middel of nowhere. He had
> waited for an hour for the tow truck, but it never arrived. He had
> called somebody to come over and help tow his car to town, but
> it turned out his car did not have a tow-rope attachment but some
> attach-a-weird-bolt-to-the-bumper arrangement which he did not
> understand how worked.
>
> After a couple of hours at the roadside trying frantically to get
> out of there, and with no chance to do road-side repairs, he tried
> the only thing he could: Fill another 10 liters og gas on the tank,
> and see what happened. The car started and he could drive the
> 10 km to the gas station and fill up.
>
> So what has this story to do with computers?
>
> It turned out that my friend had trusted the car computer which
> informed him something like "XXX km to next refueling", leading
> him to believe that he could actually get to town on the onboard
> fuel.I have no idea why the thing displayed the wrong message,
> the fuel level sensor might be proken or the remaining distance
> might have been computed based on invalid statistics, but my
> friend is not a n engineer, let a lone a computer programmer,
> so he trusted what the computer told him. And was caught
> completely by surprise when the car stopped.
>
> Why is this computer potentially lethal?
>
> My friend got away from this inciden whith no harm, as it
> was not very cold, some -2C to 0C. In three months that
> would have been a very different story. The place where the
> car broke down is 'weird' in that it is close to the coast but
> easily drop to -20C and below in winter, not counting wind-
> chill in near-gale-force winds which are common in the area
> in winter.
>
> Had the incident occured in mid-winter with an unprepared
> person (not bringing thermo suits or thermo blankets in the
> car), this computer glitch could easily turn very nasty.
>
> As far as I am concerned, the estimating-the-distance-left-
> to-refueling is a gadget is best left out, since the driver will
> handle the fuel in a different way, accounting for uncertainties,
> if he does *not* get (unreliable) info from the computer.
>
> In other words, this is a classic case of "no info is better
> than wrong info."
>
> Rune

People always seem believe computers blindly. If the computer says
something then it must be true.

Back in the 1980/90s I used to do a lot of scuba diving, between 100
and 200 dives a year. At that time the first dive computers started
to come out. Looking at the profiles that the computers allowed,
compared to using the (at the time) 20 year old recreational dive
tables, you could spend much longer under water using the computer
than staying within the table limits. In the first season that
computers were used by the club members we had 3 or 4 cases where,
despite obeying the computer instructions, the diver ended up with
symptoms associated with decompression sickness. In the previous 5
years of using tables only, there had been none. As a result the
diving club that I belonged to banned the use of dive computers for
normal diving activities.

The science and algorithms may have improved over the last few years
but I would still treat the results given by these devices with a
great deal of suspicion.

Ian

Rune Allnor wrote:
>
> On 1 Okt, 04:03, Eric Jacobsen <eric.jacob...@ieee.org> wrote:
> ...
> > Most automobiles have had fuel gauges for as long as I've been around,
> > and I think if someone in a hazardous environment ran a car down close
> > to E and it quit, people wouldn't have blamed the fuel gauge or the
> > car manufacturer, they'd have blamed the operator for running it so
> > close to E in a dangerous environment.
>
> I won't disagree with that, but rather ask why the guy ran so close
> to E in the first place. There might be good reasons (he knew the
> situation but something made it worth taking a calculated risk)
> or stupid reasons (a car computer displaying a message about the
> *driving* *distance*, not amount of fuel, left that it can not
> possibly defend.)

You haven't revealed whether the computer (program) made an error or not. As far
as I know the sending unit in gas tanks is pretty much the same as it has been
since they first installed gas gauges. If the sending unit (or the wiring) was
defective then he might well have had the same problem with an ordinary gauge.
If that were the case (his gauge read incorrectly) Would you argue that he would
be better off if they made cars with no gauge?

On the other hand if the sending unit is working to spec then you have a point,
but if I were forced to bet I would bet the sending unit is what failed.

-jim


----== Posted via Pronews.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.pronews.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= - Total Privacy via Encryption =---

On 1 Okt, 18:08, jim <".sjedgingN0sp"@m...@mwt.net> wrote:
> Rune Allnor wrote:
>
> > On 1 Okt, 04:03, Eric Jacobsen <eric.jacob...@ieee.org> wrote:
> > ...
> > > Most automobiles have had fuel gauges for as long as I've been around,
> > > and I think if someone in a hazardous environment ran a car down close
> > > to E and it quit, people wouldn't have blamed the fuel gauge or the
> > > car manufacturer, they'd have blamed the operator for running it so
> > > close to E in a dangerous environment.
>
> > I won't disagree with that, but rather ask why the guy ran so close
> > to E in the first place. There might be good reasons (he knew the
> > situation but something made it worth taking a calculated risk)
> > or stupid reasons (a car computer displaying a message about the
> > *driving* *distance*, not amount of fuel, left that it can not
> > possibly defend.)
>
> You haven't revealed whether the computer (program) made an error or not.

A sensor can measure the amount of fuel left in the tank to within
some accuracy, be it by weight or volume. The computer program is
flawed since it displays a message about *how* *far* the car can
drive on whatever amount of fuel is left.

There is a conversion factor needed to do this conversion from
remaining amount of fuel to 'drivable distance', which is based
on the fuel exonomy of the car, driving style of the driver, time
of year, where the car is used etc.

From my own experience I'd say that the fuel economy can vary
with some 50-100%. When I cruise along the highways, no other
traffic at night in summer, my car easily drives more than 700
on one tank of fuel. In winter, in the city, I don't expect more
than 400-500 km.

Which means that the fuel voulme to distance conversion varies
by 50-100%, rendering the distance estimates all but useless.

> As far
> as I know the sending unit in gas tanks is pretty much the same as it has been
> since they first installed gas gauges. If the sending unit (or the wiring) was
> defective then he might well have had the same problem with an ordinary gauge.
> If that were the case (his gauge read incorrectly) Would you argue that he would
> be better off if they made cars with no gauge?

I am saying that the guy got in trouble because he relied on
unsubstantiated, uncertain, inaccurate, useless and ultimately
superfluous information he never asked for, nor knew how to
evaluate.

Hwoever decided to have the computer display a message indicating
the *driving* *distance* left (not volume of gas left) took on
a responsibility regarding how to plan trips and use the vehicle
they never can back up.

Again, if you can't give accurate information, don't give the
information. No one wants to rely on guesswork, which these
estimates clearly are.

Rune

On 1 Okt, 22:16, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
> Rune Allnor wrote:
>
> (snip)
>
> > Again, if you can't give accurate information, don't give the
> > information. No one wants to rely on guesswork, which these
> > estimates clearly are.
>
> But we have been doing that for years with ordinary gas
> gauges, and most people know it by now. *We get used to
> the inaccurate response of analog (or, I suppose, digital)
> gas gauges and buy at the appropriate time.

Read the story again, and note the message the computer
displays. Have you ever seen a message like that? If so,
when was the first time?

> Most go way above the F when full, and below the E when
> empty. *You didn't mention the regular gauge of the
> car in question, though.

It seems no one sees the crux of the story, which leads me
to believe that this particular computer message is only
used in European cars.

It doesn *not* report 'There are 5 liters of fuel left'
or anything relating to the amount of fuel left in the tank.

It reports '15 km to next refueling', meaning it pretends
to know how far one can drive on the remaining fuel,a
number one obviously can not know since one can not possibly
know what fuel usage profile the car will operate under,
from now on and till it is refueled.

It's a tiny detail, but it makes all the difference.

Rune

Rune Allnor wrote:
(snip)

> Again, if you can't give accurate information, don't give the
> information. No one wants to rely on guesswork, which these
> estimates clearly are.

But we have been doing that for years with ordinary gas
gauges, and most people know it by now. We get used to
the inaccurate response of analog (or, I suppose, digital)
gas gauges and buy at the appropriate time.

Most go way above the F when full, and below the E when
empty. You didn't mention the regular gauge of the
car in question, though.

-- glen

Rune Allnor wrote:
(snip)

> Read the story again, and note the message the computer
> displays. Have you ever seen a message like that? If so,
> when was the first time?

I believe I have seen one, mounted above the rear view
mirror, with the option to select different displays.
(Distance traveled, compass heading, clock, gas mileage)
It would be an option, with the ordinary gauge still
in place on the dash.

(snip)

> It seems no one sees the crux of the story, which leads me
> to believe that this particular computer message is only
> used in European cars.

I meant, does it have the ordinary E...F gauge in addition
to this distance gauge? Note that the usual gauge only
has divisions 1/4 or 1/8 the way between E and F, so would
not be a precision device. If they removed such gauge
then I agree it is a defect.

-- glen

Tim Wescott wrote:
>
> Not to mention "marketing info is worse than no info", and "everything
> my car tells me is marketing".
>

Now I had a vision of my o-scope popping up a dialog that says "May we
suggest the Mark-5 Differential Wave-o-Probe in order to maximize the
accuracy of this measurement mode? Press the 'Ok' soft-key to place
your order for overnight delivery..."




--
Good day!

____________________________________
CRC
[email protected]
NOTE, delete texts: "REMOVETHIS" and
"BOGUS" from email address to reply.

On Wed, 1 Oct 2008 12:25:24 -0700 (PDT), Rune Allnor
<[email protected]> wrote:

>On 1 Okt, 22:16, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
>> Rune Allnor wrote:
>>
>> (snip)
>>
>> > Again, if you can't give accurate information, don't give the
>> > information. No one wants to rely on guesswork, which these
>> > estimates clearly are.
>>
>> But we have been doing that for years with ordinary gas
>> gauges, and most people know it by now. *We get used to
>> the inaccurate response of analog (or, I suppose, digital)
>> gas gauges and buy at the appropriate time.
>
>Read the story again, and note the message the computer
>displays. Have you ever seen a message like that? If so,
>when was the first time?
>
>> Most go way above the F when full, and below the E when
>> empty. *You didn't mention the regular gauge of the
>> car in question, though.
>
>It seems no one sees the crux of the story, which leads me
>to believe that this particular computer message is only
>used in European cars.
>
>It doesn *not* report 'There are 5 liters of fuel left'
>or anything relating to the amount of fuel left in the tank.
>
>It reports '15 km to next refueling', meaning it pretends
>to know how far one can drive on the remaining fuel,a
>number one obviously can not know since one can not possibly
>know what fuel usage profile the car will operate under,
>from now on and till it is refueled.
>
>It's a tiny detail, but it makes all the difference.
>
>Rune

I disagree that it makes much difference. It's a simple translation
from estimated fuel remaining to estimated distance remaining. The
computer may or may not improve the estimate by using recent average
fuel efficiency computation, or it might not. If one decides to
assume that an information source (and I'd argue any information
source) is accurate enough to bet their personal safety on, I think
that's a personal decision which one needs to take responsibility for.
Blaming the information source is poor form, unless there's suitable
reason to believe that the accuracy is reliable. My point in
bringing up automotive gauge reliability in general is that there's no
reason to believe they're that accurate. The fact that the display
is digital or that it's undergone a simple conversion (from volume to
distance) shouldn't change that, and I think it's reasonable to expect
even a layman to understand that.

A couple of weeks ago I co-drove a diesel truck with a 40-foot trailer
(containing three race cars) on a 4000 mile round trip across the
country. The truck has a trip computer just like what you describe,
in that it can show fuel efficiency (average mpg), distance travelled,
or estimated distance remaining on the tank. We'd had it displaying
mpg for a long time, and on the return trip, during one of my
non-driving stints, the other driver switched it to 'distance
remaining' as we passed a sign saying "Next service 75 miles". The
analog gauge indicated less than a quarter of a tank, the computer
indicated something like 87 miles remaining. I thought it was a
mistake to proceed but didn't say anything. If we ran out of fuel my
plan was to continue working on my laptop while the driver who ran the
tank out went for help.

The trip computer counted down 'miles remaining' far faster than the
miles actually went by, and the driver was quite stressed out about
whether we'd make it to a filling station (especially one that had
diesel) or not. We managed to pull up to a pump with the computer
showing 1 mile remaining on the tank. We don't know how much fuel was
actually left in the tank. I'd been thinking that there is likely
error in the computation and I wasn't willing to bet which way the
error was, i.e., it could run out of fuel with 10 miles remaining left
on the display, or the display could conceivably show 0 or a negative
number for a while depending on how the estimates and computations
were done.

During that time we had a pretty significant headwind (actually the
remains of hurricane Ike), so whatever fuel efficiency number the
computer was using to estimate 'miles remaining' could have been more
than the actual mpg achieved. In any case, I think it unreasonable
for a person to assign decision making for their personal safety to a
system which is unlikely to have enough information to make that
decision. Fuel efficiency has a lot of factors that affect it, and I
don't think an alert layman should just trust that a computer in an
automobile has enough accuracy or information to know just how much
further one can go.

There is a saying that it is futile to try to "idiot proof" a system
because they keep making bigger idiots. I don't think your friend is
necessarily an idiot (he may be, I don't know him, but I'd suspect
not), but I do think his trust in machinery is misplaced. Especially
when his personal safety is involved.

We do all wind up having to trust equipment, or computers, or various
devices, from time to time. Sometimes the trust is misplaced,
sometimes systems really are made to be reliable and accurate. Knowing
the difference is pretty worthwhile, though, I think. If in doubt,
take the safe route.

I'm glad it worked out for your friend.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

>On Tue, 30 Sep 2008 17:18:09 -0700 (PDT), Rune Allnor
><[email protected]> wrote:
>
>>Hi folks.
>>
>>Last night a friend of mine called. He was a bit upset, as his car
>>had stopped, with no warning, in the middel of nowhere. He had
>>waited for an hour for the tow truck, but it never arrived. He had
>>called somebody to come over and help tow his car to town, but
>>it turned out his car did not have a tow-rope attachment but some
>>attach-a-weird-bolt-to-the-bumper arrangement which he did not
>>understand how worked.
>>
>>After a couple of hours at the roadside trying frantically to get
>>out of there, and with no chance to do road-side repairs, he tried
>>the only thing he could: Fill another 10 liters og gas on the tank,
>>and see what happened. The car started and he could drive the
>>10 km to the gas station and fill up.
>>
>>So what has this story to do with computers?
>>
>>It turned out that my friend had trusted the car computer which
>>informed him something like "XXX km to next refueling", leading
>>him to believe that he could actually get to town on the onboard
>>fuel.I have no idea why the thing displayed the wrong message,
>>the fuel level sensor might be proken or the remaining distance
>>might have been computed based on invalid statistics, but my
>>friend is not a n engineer, let a lone a computer programmer,
>>so he trusted what the computer told him. And was caught
>>completely by surprise when the car stopped.
>>
>>Why is this computer potentially lethal?
>>
>>My friend got away from this inciden whith no harm, as it
>>was not very cold, some -2C to 0C. In three months that
>>would have been a very different story. The place where the
>>car broke down is 'weird' in that it is close to the coast but
>>easily drop to -20C and below in winter, not counting wind-
>>chill in near-gale-force winds which are common in the area
>>in winter.
>>
>>Had the incident occured in mid-winter with an unprepared
>>person (not bringing thermo suits or thermo blankets in the
>>car), this computer glitch could easily turn very nasty.
>>
>>As far as I am concerned, the estimating-the-distance-left-
>>to-refueling is a gadget is best left out, since the driver will
>>handle the fuel in a different way, accounting for uncertainties,
>>if he does *not* get (unreliable) info from the computer.
>>
>>In other words, this is a classic case of "no info is better
>>than wrong info."
>>
>>Rune
>
>While I sympathize with your friend I'll suggest a different point of
>view. We had a discussion here recently how people have given over
>their risk/danger management skills to other "authority". People
>depend on signs, "authority" figures, or computer outputs to tell them
>what to do. I think that's a bad thing and people need to take more
>responsibility for themselves.
>
>Most automobiles have had fuel gauges for as long as I've been around,
>and I think if someone in a hazardous environment ran a car down close
>to E and it quit, people wouldn't have blamed the fuel gauge or the
>car manufacturer, they'd have blamed the operator for running it so
>close to E in a dangerous environment. A lack of planning is not a
>good excuse if the consequences of failure are high. Everybody knows
>(or should know) that automotive gauges aren't precision instruments.
>I think your friend made a mistake in thinking that because a display
>is digital then it must be accurate. What gives one that notion?
>
>I hope he uses the incident to learn (and tell his friends!) that the
>gauge is only an estimate and isn't precise enough to trust one's life
>with.

Oh, Eric. You are being sooooo unreasonable. This fuel gauge has to b
super accurate. Its digital.

Steve

On Wed, 01 Oct 2008 21:17:45 -0500, steveu wrote:

> Oh, Eric. You are being sooooo unreasonable. This fuel gauge has to be
> super accurate. Its digital.

Not only that, but a read-out in terms of km-to-empty tells you nothing
about the absolute error with respect to full-tank, the way the thickness
of a guage needle and coarseness of guage markings can. In fact, if
you're travelling relatively short distances, which is usually the case,
you get completely the wrong sense of comfort, because the natural
comparison is distance-to-empty vs distance-to-travel, where there is no
indication that the less vs greater comparison is swamped by the
estimation error of distance-to-empty.

I agree with Rune's original comment: this is a bad wy to design a fuel
guage.

--
Andrew

On 2 Okt, 03:30, Eric Jacobsen <eric.jacob...@ieee.org> wrote:
> On Wed, 1 Oct 2008 12:25:24 -0700 (PDT), Rune Allnor

> >It reports '15 km to next refueling', meaning it pretends
> >to know how far one can drive on the remaining fuel,a
> >number one obviously can not know since one can not possibly
> >know what fuel usage profile the car will operate under,
> >from now on and till it is refueled.
>
> >It's a tiny detail, but it makes all the difference.
>
> >Rune
>
> I disagree that it makes much difference. *It's a simple translation
> from estimated fuel remaining to estimated distance remaining.

Depends on what you mean by 'simple'. The arithmetics, yes.
Finding the correct conversion factor, no.

>* The
> computer may or may not improve the estimate by using recent average
> fuel efficiency computation, or it might not. * If one decides to
> assume that an information source (and I'd argue any information
> source) is accurate enough to bet their personal safety on, I think
> that's a personal decision which one needs to take responsibility for.

My point is that the simple fuel gauges indicating the remaining
*amount* of fuel is safer, simply because it

a) reports the state with far less percieved accuracy
b) leaves it to the user to estimate how far one can
get on the remaining fuel.

> Blaming the information source is poor form, unless there's suitable
> reason to believe that the accuracy is reliable. *

Not the source (the sensor is probably the same it
always was), the presentation.

My point in
> bringing up automotive gauge reliability in general is that there's no
> reason to believe they're that accurate. * The fact that the display
> is digital or that it's undergone a simple conversion (from volume to
> distance) shouldn't change that,

It does. As you can tell from your story. Once the information,
either by contents or presentation, changes how the users operate
a piece of equipment, the damage has been done.

> and I think it's reasonable to expect
> even a layman to understand that.

The 'intelligent' or 'critical' consumer is a mythical creature
on a par with the Sfinx or Phenix. Humans have been around for
some 50000-100000 years, but only for the past couple of decades
have the coution of 'criticise information sources' been round.
And only in the Western cultural sphere.

In all other ages, in all other cultures, people have respondes
(and still respond) to authority. My point is that it is human
nature to believe what they are told by what they percieve
as an 'authority.' 99.999% of the population do not have the
knowledge or skills required to evaluate what a computer
tells them, which is why they take the messages at face value.

Rune

Tim Wescott wrote:
> [snip]
>
> Not to mention "marketing info is worse than no info", and "everything
> my car tells me is marketing".
>

This morning there was an ad on TV for a blood glucose meter which
claimed having voice output,as well as visual, made the meter more accurate.

On Thu, 02 Oct 2008 05:29:39 -0500, Richard Owlett wrote:

> Tim Wescott wrote:
>> [snip]
>>
>> Not to mention "marketing info is worse than no info", and "everything
>> my car tells me is marketing".
>>
>>
> This morning there was an ad on TV for a blood glucose meter which
> claimed having voice output,as well as visual, made the meter more
> accurate.

That's what you get for watching TV in the morning, I suppose...

Sorry, cheap shot. I've just never understood how people managed to do
that, but I know that many do.

Cheers,

--
Andrew

On Oct 2, 1:13 am, Andrew Reilly <andrew-newsp...@areilly.bpc-
users.org> wrote:
> On Wed, 01 Oct 2008 21:17:45 -0500, steveu wrote:
> > Oh, Eric. You are being sooooo unreasonable. This fuel gauge has to be
> > super accurate. Its digital.
>
> Not only that, but a read-out in terms of km-to-empty tells you nothing
> about the absolute error with respect to full-tank, the way the thickness
> of a guage needle and coarseness of guage markings can. In fact, if
> you're travelling relatively short distances, which is usually the case,
> you get completely the wrong sense of comfort, because the natural
> comparison is distance-to-empty vs distance-to-travel, where there is no
> indication that the less vs greater comparison is swamped by the
> estimation error of distance-to-empty.
>
> I agree with Rune's original comment: this is a bad wy to design a fuel
> guage.
>
> --
> Andrew

I have to agree with Rune. The fundamental difference is between
accuracy and precision. If a given number is reported to 5 decimal
places - then most people will intuitively believe that the given
number is accurate to 5 decimal places as well. You see this a lot
when people are reporting dB values.

My analog fuel gauge changes depending whether I'm going uphill or
down, and I seem to go through the top 1/2 of my fuel tank faster than
the bottom half. With an analog gauge you tend to get some context
with the reading and you can't say I've got 10.34562 Liters left in my
tank.

Just my thoughts.
Cheers,
Dave

Andrew Reilly wrote:
> On Thu, 02 Oct 2008 05:29:39 -0500, Richard Owlett wrote:
>
>
>>Tim Wescott wrote:
>>
>>>[snip]
>>>
>>>Not to mention "marketing info is worse than no info", and "everything
>>>my car tells me is marketing".
>>>
>>>
>>
>>This morning there was an ad on TV for a blood glucose meter which
>>claimed having voice output,as well as visual, made the meter more
>>accurate.
>
>
> That's what you get for watching TV in the morning, I suppose...
>
> Sorry, cheap shot. I've just never understood how people managed to do
> that, but I know that many do.
>
> Cheers,
>

BUT I engineer, turned on local news to find out if I would freeze or
drown if/when I went out ;/

On 2 Okt, 08:52, Rune Allnor <all...@tele.ntnu.no> wrote:

> The 'intelligent' or 'critical' consumer is a mythical creature
> on a par with the Sfinx or Phenix. Humans have been around for
> some 50000-100000 years, but only for the past couple of decades
> have the coution of 'criticise information sources' been round.
> And only in the Western cultural sphere.

That should, of course be

"only for the past couple of *centuries*
have the notion of 'criticise information sources' been around.
And only in the Western cultural sphere."

All that started in Europe in the 18th century, with the age
of enlightenment.

> In all other ages, in all other cultures, people have respondes
> (and still respond) to authority. My point is that it is human
> nature to believe what they are told by what they percieve
> as an 'authority.'

Just to illustrate my point further, if you consult the
dictionary for terms equivalent to "question / challenge
authority" you find stuff like

- cantankerous
- disloyalty
- insubordination
- rebellion
- treason

and so on. Hardly terms anyone would want to become
associated with.

It is a fundamental human trait to conform to
whatever (percieved) authority figures are around.

Rune

Eric Jacobsen wrote:
> On Wed, 1 Oct 2008 12:25:24 -0700 (PDT), Rune Allnor
> <[email protected]> wrote:

>>It doesn *not* report 'There are 5 liters of fuel left'
>>or anything relating to the amount of fuel left in the tank.
>>
>>It reports '15 km to next refueling', meaning it pretends
>>to know how far one can drive on the remaining fuel,a
>>number one obviously can not know since one can not possibly
>>know what fuel usage profile the car will operate under,
>>from now on and till it is refueled.
>>
>>It's a tiny detail, but it makes all the difference.

> I disagree that it makes much difference. It's a simple
> translation from estimated fuel remaining to estimated distance
> remaining. The computer may or may not improve the estimate by
> using recent average fuel efficiency computation, or it might
> not.

Perhaps not so simple as it's an additional layer of extrapolation.
By their nature extrapolation processes are "unstable" -- i.e., both
heavily dependent on seed information and divergent from ground truth
in the not-too-long term -- and are thus often best left to the most
sophisticated processor available. Which in the present case is most
probably the driver with all the implicit knowledge his experience
affords rather than some faddish piece of silicon.

> If one decides to assume that an information source (and
> I'd argue any information source) is accurate enough to bet
> their personal safety on, I think that's a personal decision
> which one needs to take responsibility for.

And who can't hear must feel? We are (or are becoming) engineers and
stand a good chance of having been in the basic mindset that implies
since childhood, so it's sometimes hard to realize that ordinary
people don't scrutinize their assumptions like that. (Nor are we
necessarily good at it in nontechnical arenas.) Humans in general are
very much at the mercy of framing (or presentation, in Rune's words)
and also don't transfer things learned in one frame to another all
that readily. Anyone interested in this discussion should read
Taleb's "Black Swan" on issues of knowledge, learning, and
prediction.

> My point in bringing up automotive gauge reliability in general
> is that there's no reason to believe they're that accurate.
> The fact that the display is digital or that it's undergone a
> simple conversion (from volume to distance) shouldn't change that,
> and I think it's reasonable to expect even a layman to understand
> that.

As noted above it's naive to assume that understanding in one
instance will suffice to engender recognition of technically similar
but spatially, temporally, or socially distant cases. If we call
saying "but the assessment shouldn't change" the "moralist" approach
then I think Rune's point is that safety is a matter of pragmatics --
of observing over judging. And we do observe that the assessment
changes; that gratuitous levels of mediation between device and
operator may be hazardous; that this sort of "tiny detail" is as
important a source of disaster as the big gaping hole. On these
matters, see e.g. Perrow's "Normal Accidents" or Rochlin's "Trapped
in the Net" which is online at
http://www.pupress.princeton.edu/books/rochlin/ .


Martin

--
Values of beeta will give rise to dom!