Polar Analysis from flight logs?

I won't speculate on the results of this type of testing, but, around here
the wind varies considerably with altitude, even with calm conditions
in the morning. When doing performance work I would try to work
crosswind. Wind shear effects are less influential. It is a great
help to take a tape recorder and record every bump and time. A lot of
data gets thrown out.

If you take twenty minutes per point, you'll be spending a lot on tow fees.

Dennis Brown

John Sinclair wrote in message ...
I asked Dick Johnson, why couldn't I fly in calm conditions
(morning) and hold a given airspeed (say 60 knots)
and a given heading (say west) for 10 minutes, then
reverse heading to east (to cancel out any wind) and
then analyze the GPS trace to determine my ships L/D
at 60 knots. We have an accurate distance covered and
fairly accurate altitude lost, so why can't we crunch
the numbers?
Dick said the GPS info wasn't accurate enough. I thought
it was a good idea, but I defer to the master.
JJ

At 03:00 28 December 2004, Bob Gibbons wrote:
On this topic of determining L/D from interthermal
cruising, Dick
Johnson did a fascinating and underappreciated study
in the late
1970's of airmass behavior between thermals. Dick flew
in mostly blue
conditions and simply recoded his height loss versus
distance covered
between thermals. Dick's results are reported in SOARING,
June 1979.

Dick found that, on the average, the airmass between
thermals has an
average sink rate related to the upgoing thermal strength.
The
relationship Dick found was; the airmass sink is approximately
10% of
the lift strength.

I have always felt Dick's study explains why it is
so difficult to fly
cross country (in a blue conditions) with a ship having
an L/D less
than about 30:1. The probability of running into the
next thermal
purely by chance becomes too low as the L/D drops.

For this discussion, I think Dick's study shows the
inadvisability of
trying to deduce flight performance from interthermal
measurements.

Bob

On Mon, 27 Dec 2004 20:43:41 GMT, 'Papa3'
wrote:

Mark,

My first post came across as a bit glib - apologize
for that. But, I
actually put a little thought into that subject recently
while writing some
batch analysis specs for GPS log files. The problem
is that a good glider
pilot will not encounter random vertical motions -
even in cruise flight.
He/she will stack the deck in his favor, seeking out
cloud streets or
connecting the best looking individual Cu. Thus,
you can't just average
out the L/D over time on specific segments (e.g. cruise
flight). What you
can do is deterimine which pilot does the best job
of achieving highest L/D
on a given day. Several of the popular flight analysis
programs do this
already.

I'd certainly be interested in any detailed ideas you
might have.

P3

'Mark Zivley' wrote in message
.com...
Obviously ridge flight would not be conducive, nor
wave, but put enough
data together from cruises during thermal flights
and I bet something
could be put together.

Papa3 wrote:

Mark,

How do you propose to isolate the impact of vertical
air motion? For
instance, I can fly the ridge at 100Kts and maintain
altitude (same for
wave
or cloudstreets). I'm sure Rolladen-Schneider (ahem,
DG) would love to
publish the L/D of my LS8 from the average of my
flights for a season:
'LS8, with a measured L/D of 800:1...'

Cheers,
Erik


'Mark Zivley' wrote in message
m...

We all know what the manufacturer's polars look like,
but what about our
individual planes. Has anyone done any work to develop
a program that
would look at some flight logs and determine what
a particular glider's
actual polar is? At one point Ball was making a vario
system that would
determine the aircraft's polar over time just by flying.

For someone who already had some algorithms for computing
wind from
ground track drift during thermals could take this
info and then be able
to back figure from GPS ground speed what the IAS
was during a
particular phase of the flight. By isolating longer
sections of cruise
flight at varios airspeeds it should be do-able.
Question is, has it
been done.

Mark

extra 'hot' in the address to delay the spammers...

Errors in measurement can be made up for with lots and lots of data, so
long as the errors are not biased one way or another. Thus, you should
be able to get an accurate polar even in thermally air, without
spending a fortune on tow fees. Turn on data recording, then fly
absolutely straight and same speed, through thermals as well as sink,
while taking data. Turn off data recording before thermaling back up to
altitude.

The key is to fly so that on average you're not biased toward flying in
lift vs. sink. You should randomize heading (if you always go
up/downwind you'll be in streets), randomize time of turning on/off the
data recording (if you turn on after leaving a thermal and off when you
find a new one, you'll be biased toward sink). If you do this for a
season, for example getting 20 minutes of data in the 1-2 hours of
prestart fooling around at contests, you might have a really good polar
at the end of it.

You could also do the opposite: A good pilot should be flying faster
through sink and slower through lift, and should spend more time in
lift than in sink. The difference between the "polar" measured in
thermal conditions and the factory polar can be a basis of a measure of
pilot skill. A good pilot should give a polar with a worse high speed
end -- because he always flies fast through sink -- a much better low
speed end -- becasue he always flies slow through lift -- and a
positive bias -- the whole polar shifted up.

In principle, all pretty easy to add to a glide computer. Of course we
all have a long list of more important features.

John Cochrane (BB)

BB wrote:
Errors in measurement can be made up for with lots and lots of data, so
long as the errors are not biased one way or another. Thus, you should
be able to get an accurate polar even in thermally air, without
spending a fortune on tow fees. Turn on data recording, then fly
absolutely straight and same speed, through thermals as well as sink,
while taking data. Turn off data recording before thermaling back up to
altitude.

The key is to fly so that on average you're not biased toward flying in
lift vs. sink. You should randomize heading (if you always go
up/downwind you'll be in streets), randomize time of turning on/off the
data recording (if you turn on after leaving a thermal and off when you
find a new one, you'll be biased toward sink). If you do this for a
season, for example getting 20 minutes of data in the 1-2 hours of
prestart fooling around at contests, you might have a really good polar
at the end of it.

You could also do the opposite: A good pilot should be flying faster
through sink and slower through lift, and should spend more time in
lift than in sink. The difference between the "polar" measured in
thermal conditions and the factory polar can be a basis of a measure of
pilot skill. A good pilot should give a polar with a worse high speed
end -- because he always flies fast through sink -- a much better low
speed end -- becasue he always flies slow through lift -- and a
positive bias -- the whole polar shifted up.

In principle, all pretty easy to add to a glide computer. Of course we
all have a long list of more important features.

John Cochrane (BB)

Wouldn't it be simpler and much more scientific to arrange to do a
series of test runs over a LIDAR site, and simply post process the radar
and glider data recorder data ?

Eric Greenwell wrote:
John Sinclair wrote:
I asked Dick Johnson, why couldn't I fly in calm conditions
(morning) and hold a given airspeed (say 60 knots)
and a given heading (say west) for 10 minutes, then
reverse heading to east (to cancel out any wind) and
then analyze the GPS trace to determine my ships L/D
at 60 knots. We have an accurate distance covered and
fairly accurate altitude lost, so why can't we crunch
the numbers?
Dick said the GPS info wasn't accurate enough. I thought
it was a good idea, but I defer to the master.

How long ago did you ask him? GPS is much more accurate in the last
few
years, especially if using the WAAS ablities. But, let's say you know

the distance to only +/- 100 feet (it's typically more like +/- 30
feet), then flying only a mile (5000 feet) would be a 2% error, or
one
L/D point for a 50:1 glider. Good enough for us, I think.

I don't think this is quite right. Consider a 50:1 glider descending
100 meters:
It will go 100 x 50 = 5000 meters
I think vertical navigation errors are typically 1.5 times the
horizontal.
If horizontal accuracy is 10 meters, then vertical will be about 15
So when we've measured this exactly 50:1 glider sinking 100 meters, the
two worst cases we would read for a measurement a
(5000-Herror)/(100+Verror) = 4990/115 = 43.4
(5000+Herror)/(100-Verror) = 5010/85 = 58.9

With errors of 3 meters and 5 meters, it gets closer.
4997/105 = 47.6
5003/95 = 52.6

With errors of 1 and 2 meters
4999/102 = 49.0
5001/98 = 51.0

If you use the GPS for the altitude instead of the pressure altitude,

you might have to fly off a 1000 feet or so of altitude, I suppose.
Maybe Dick was referring to GPS altitude?

Probably. If I understand the technique he uses correctly, he flies a
test glider at a particular airspeed, timing the descent, to get a sink
rate for that airspeed. Then he plots the sinkrates to get the polar.
I can't see why he'd be interested in horizontal position at all.
I don't know what kind of accuracy is possible from the barometric
altimeters. It could be that a skilled pilot/data recorder can get
accuracy below the typical 20 foot tick mark on the altimeter face.

OTOH, you are correct, if I'm reading the FAA docs right, WAAS should
be able to give vertical accuracy better than 2 meters 95 percent of
the time. I don't know if the COTS handhelds can actually deliver that.
On Sam Wormsley's GPS site, there's a link that suggests that a Garmin
GPS 76 maintained about 6 meter vertical accuracy 95 percent of the
time. That's very close to the 20 foot resolution on a barometric
altimeter, and that's absolute position, rather than relative position.
That is, we don't care where we started and where we ended, but rather
how far we descended. If the error offset is pretty much the same at
the beginning and end of a run, then the relative accuracy may be much
better than the absolute position.

It might be interesting to look at the GPS speed during a descent. If
the airspeed and heading is being held constant, and the airmass is
uniform, then the GPS speed should be constant. Excursions might
indicate shears that would affect the quality of the data. You might
be able to improve accuracy by throwing out segments that indicated
non-steady-state behavior.

Tim Ward


--
Change "netto" to "net" to email me directly

Eric Greenwell
Washington State
USA

At 18:30 28 December 2004, Eric Greenwell wrote:
How long ago did you ask him? GPS is much more accurate
in the last few
years, especially if using the WAAS ablities. But,
let's say you know
the distance to only +/- 100 feet (it's typically more
like +/- 30
feet), then flying only a mile (5000 feet) would be
a 2% error, or one
L/D point for a 50:1 glider. Good enough for us, I
think.

At 50:1 flying a one mile distance should yield an
altitude loss of 100', so I think a +/- 100' GPS error
in altitude could yield a 'measured' L/D of between
25:1 and infinity.

A single measured glide of 10 miles, assuming no other
errors from pilot inputs or net airmass movements,
would be good to only 10%, or +/- 5 L/D points, that's
a bit better, but you'd still need to average a bunch
of runs of 10+ miles to get any kind of accuracy. Best
done in dead calm air the way Dick does.

I have tried plotting speed versus glide angle over
a number of flights. You need to adjust for wind speed
and direction (estimated from thermal drift angle and/or
ground speed differences around a circle), then you
have to convert from TAS to IAS (2% per 100'). I plotted
about 30 points and realized that I had a very low
signal to noise ratio - that is the points made a big
cloud that sort of sloped in the expected direction.

I suppose with enough data points you could get a resonable
average L/D performance for any given speed, but I
did the experiment for a different reason, I was interested
in the VARIANCE in performance that I could expect
in real world conditions, including pilot technique
and long stretches of good/bad air.

If I know the difference between expected performance
and 99% worst case performance I know how much cushion
to carry on final glides - particularly as I get closer
to home, and the ground, and have fewer options to
recover. It also may help a bit to know how early you
can start a long final glide, though this is much more
subject to observable conditions like streeting and
average thermal strength and therefore less informed
by bulk statistical analysis.

9B

Andy Blackburn wrote:

At 18:30 28 December 2004, Eric Greenwell wrote:

How long ago did you ask him? GPS is much more accurate
in the last few
years, especially if using the WAAS ablities. But,
let's say you know
the distance to only +/- 100 feet (it's typically more
like +/- 30
feet), then flying only a mile (5000 feet) would be
a 2% error, or one
L/D point for a 50:1 glider. Good enough for us, I
think.


At 50:1 flying a one mile distance should yield an
altitude loss of 100', so I think a +/- 100' GPS error
in altitude could yield a 'measured' L/D of between
25:1 and infinity.

My example was for GPS distance, and pressure altitude, to indicate that
a distance measurement wasn't a problem. That's why a I later referred
to flying off at least 1000 feet if GPS altitude was used.

What I don't know is how much error change one can expect in GPS
altitudes taken 5 or 10 minutes apart. The difference (GPS start height
minus GPS finish height) might have a much smaller error than the
altitude itself, which would allow shorter glides (500 foot loss if the
differential error was only 5 feet, for example).


--
Change "netto" to "net" to email me directly

Eric Greenwell
Washington State
USA

I have tried using a $10,000 carrier phase GPS receiver with 0.1m
precision (post-processed) for glide testing. The GPS data was so
precise, you could clearly see the antenna move a few cm when the wing
was raised for takeoff. Even when flying in the calmest conditions,
with no discernable airmass movement, the vertical motions are
significant.

In analysing the data, I could not precisely fit a straight line to the
data points from 1-2 min glides at constant airspeed, even after
correcting for slight airspeed variations using total energy. Johnson
does not have this problem because he only has two data points, one at
the beginning and one at the end of each glide. When you have about a
hundred data points, one every second, you can really see the problem.
You need a lot more data points to average out the noise.

Based on this, I doubt that you could get useful data from a less
precise GPS, with a slower sampling interval, in uncontrolled
conditions. There is just too much noise to get useful results without
an impossibly huge data set.

Eric Greenwell wrote:
Mark Zivley wrote:

We all know what the manufacturer's polars look like, but what
about
our individual planes. Has anyone done any work to develop a
program
that would look at some flight logs and determine what a particular
glider's actual polar is? At one point Ball was making a vario
system that would determine the aircraft's polar over time just by
flying.

For someone who already had some algorithms for computing wind from

ground track drift during thermals could take this info and then be
able to back figure from GPS ground speed what the IAS was during a

particular phase of the flight. By isolating longer sections of
cruise flight at varios airspeeds it should be do-able. Question
is,
has it been done.

I haven't heard of it being done, and I can't imagine how one would
compensate for air motion, both vertical and horizontal, just using
the
GPS info. Both motions change with location, altitude, and time.
Perhaps
if the flight record included the airspeed, like some varios can
supply,
there would be some hope of doing it. I don't think you could count
on
the vertical motion averaging to zero during the cruises, since we
typically adjust our path to include as much up air as possible.

You can get some good info using a flight recorder, but you have to
do
it when the air is calm. If you are really interested, invest in a
few
high tows and make the measurements. Take a look at this test done on
a
DG 800:

http://groups.yahoo.com/group/ASA-NewsGroup/message/59

You don't have to be a group member to read the message.


--
Change "netto" to "net" to email me directly

Eric Greenwell
Washington State
USA

"Tim.Ward" writes:

the distance to only +/- 100 feet (it's typically more like +/- 30
feet), then flying only a mile (5000 feet) would be a 2% error, or
one
L/D point for a 50:1 glider. Good enough for us, I think.

I don't think this is quite right. Consider a 50:1 glider descending
100 meters:
It will go 100 x 50 = 5000 meters
I think vertical navigation errors are typically 1.5 times the
horizontal.
If horizontal accuracy is 10 meters, then vertical will be about 15
So when we've measured this exactly 50:1 glider sinking 100 meters, the
two worst cases we would read for a measurement a
(5000-Herror)/(100+Verror) = 4990/115 = 43.4
(5000+Herror)/(100-Verror) = 5010/85 = 58.9

To find and eliminate the random errors just have a second GPS receiver on the
ground under the area of the flight test. Comparing of the records would show
the GPS errors. (Any motion on a stationary receiver is GPS error.) We are
able to get better than millimeter accuracy (for earth plate motion studies)
using basically this technique, as you don't care where you are absolutely only
relative to the other point/receiver. We are interested in the polar at
higher speeds than plate motion speeds so some accuracy will be lost.

Steve

At 00:00 29 December 2004, Eric Greenwell wrote:
My example was for GPS distance, and pressure altitude,
to indicate that
a distance measurement wasn't a problem. That's why
a I later referred
to flying off at least 1000 feet if GPS altitude was
used.

What I don't know is how much error change one can
expect in GPS
altitudes taken 5 or 10 minutes apart. The difference
(GPS start height
minus GPS finish height) might have a much smaller
error than the
altitude itself, which would allow shorter glides
(500 foot loss if the
differential error was only 5 feet, for example).

Clarification noted - but distance measurement is a
problem with GPS with respect to polar calculations.

Without knowing the technique Dick Johnson uses, or
the specs on a specific pressure transducer, it's hard
to know if measuring pressure altitude through a digital
transducer is more or less accurate than the traditional
method. I'd guess it's a close call, but that has nothing
to do with GPS.

The main source of error, is being able to turn GPS
ground speed (or distance) into IAS reliably by subtracting
wind speed and adjust for altitude. An even greater
source of error is trying to use fixes from a typical
soaring day with airmass movements and pilot control
inputs, airspeed changes and flightpath deviations.

The empirical evidence is that there is way too much
randomness from the above noted effects to tease out
a anything much beyond just how much randomness there
in fact is on a typical flight.

Maybe if you did fifty 10-mile runs on a dead calm
day across five different airspeeds, you'd get less
scatter - but I think that's more or less what Dick
does, except he measures IAS directly, rather than
having to figure it out from ground speed.

9B

Andy Blackburn wrote:

Without knowing the technique Dick Johnson uses, or
the specs on a specific pressure transducer, it's hard
to know if measuring pressure altitude through a digital
transducer is more or less accurate than the traditional
method. I'd guess it's a close call, but that has nothing
to do with GPS.

I don't know the specifics of Dick's instrument, but a digitally logged
unit has got to be better, maybe a lot better, than trying to read a
mechanical altimeter with a little vibrator on the panel!


The main source of error, is being able to turn GPS
ground speed (or distance) into IAS reliably by subtracting
wind speed and adjust for altitude.

I think the pilots trying this are not doing that, but instead rely on
their airspeed to give them IAS. This does require calibration of the
ASI for real accuracy.

An even greater
source of error is trying to use fixes from a typical
soaring day with airmass movements and pilot control
inputs, airspeed changes and flightpath deviations.

I agree completely, and expect the best data to come on the same kind of
day Dick uses; that is, not a soaring day!

It's this air mass movement that drives the Akafliegs to using "sacred
cows" to measure the motion of the test glider relative to the "cow".



--
Change "netto" to "net" to email me directly

Eric Greenwell
Washington State
USA