Another mid-air (UK)

On Thursday, August 7, 2014 1:30:14 AM UTC-7, Stats Watcher wrote:
At 04:37 07 August 2014, Ramy wrote:

Dan, you kidding about the accuracy of GPS, right? Or maybe

trolling? I

realize you may have never used a GPS in gliders, but I am sure

you've seen

GPS in cars which can pin point in which Lane you are traveling.

OK, I am

not going to fall into this trap.



Ramy





Most automotive GPS systems augment the raw GPS signal with a

'Lock to Road' function. If you have a route set they also often

assume you are traveling the route, so assume you will be and

display you in the correct lane.



You can see this in two ways... On my system if I choose to ignore

the guidance and take a different route the GPS shows me

travelling the the suggest route for a few seconds before snapping

to the new road. If you turn off the 'lock to road' function, it will

often show you several (or even tens) of metres from the road.



Using a modern ( and FLARM is not) GPS engine in an optimum

installation, raw GPS is still +-8m at 95% confidence and 16m at

99% confidence (Current US DoD stats)

Isn't the Flarm GPS WAAS corrected? The WAAS standard of performance is 1.6M nominal horizontal (95%). When the NDGPS system is functional (and GPS engines take advantage of it) this will be reduced to 15 cm.

Even with 8m accuracy, It is more than sufficient for accurate collision avoidance, unless someone considers half a wingspan not accurate enough for collision warning with another glider or obstacle.

Ramy

On Thursday, August 7, 2014 11:46:30 AM UTC-4, Ramy wrote:
Even with 8m accuracy, It is more than sufficient for accurate collision avoidance, unless someone considers half a wingspan not accurate enough for collision warning with another glider or obstacle.

Prompted to comment here by my Geek Fascination Syndrome (GFS)...

I'd guess that the error on each position fix is not significant. Flarm is computing trajectories from multiple position fixes, so a maximum +/- 8M error for each fix will be smoothed out by the statistical distribution of a large number of position fix errors. Some of the errors are + and some are -. Average 100 GPS position fixes and you will get a pretty accurate position.

I'd expect that the sampling rate is high enough to make the trajectory calculations quite accurate. If the sampling rate was not fast enough they could put a faster processor into the units. Plus as Ramy pointed out, FLARM must be alerting for 'possible near misses' because there will be small deviations in trajectory caused by turbulence and pilot control inputs.

The 8M 'error' of GPS is unlikely to be the critical factor that keeps FLARM from fulfilling its mission. I'd bet that the critical factor is the human pilot, that being the most inherently error prone and undependable part of the system.

At 16:06 07 August 2014, son_of_flubber wrote:
On Thursday, August 7, 2014 11:46:30 AM UTC-4, Ramy
wrote:
Even with 8m accuracy, It is more than sufficient for
accurate
collision=
avoidance, unless someone considers half a wingspan not
accurate enough
fo=
r collision warning with another glider or obstacle.=20

Prompted to comment here by my Geek Fascination Syndrome
(GFS)...

I'd guess that the error on each position fix is not significant.
Flarm is
=
computing trajectories from multiple position fixes, so a
maximum +/- 8M
er=
ror for each fix will be smoothed out by the statistical
distribution of a
=
large number of position fix errors. Some of the errors are +
and some are
=
-. Average 100 GPS position fixes and you will get a pretty
accurate
positi=
on.

I'd expect that the sampling rate is high enough to make the
trajectory
cal=
culations quite accurate. If the sampling rate was not fast
enough they
co=
uld put a faster processor into the units. Plus as Ramy pointed
out,
FLARM=
must be alerting for 'possible near misses' because there will
be small
de=
viations in trajectory caused by turbulence and pilot control
inputs.

The 8M 'error' of GPS is unlikely to be the critical factor that
keeps
FLAR=
M from fulfilling its mission. I'd bet that the critical factor is the
hum=
an pilot, that being the most inherently error prone and
undependable part
=
of the system.


+-8m 95% of the time is best case, under ideal conditions and
installations. Do you fly only in ideal conditions and have an
ideal installation? f the answer is 'no' your error will be larger.
Also this is 2D error. As everyone 'knows' and goes on ad-
nauseam, GPS vertical error is significantly worse.

Have you 'any' evidence for the rest of your post or is it just
pure supposition? If you analyse the data rate and bandwidth
requirements of the data link you'll realise what you suggest is
impossible

Whether or not any of this has a thing to do with the original post:
The technology discussed is not perfect, and like a vario is supplementary.
Is your scan 100% perfect? Are your eyes? After 8 hours of flying? Burning from sweat and sunscreen?
Will your vision improve with age? Will you admit it?
Jim

Op donderdag 7 augustus 2014 18:55:16 UTC+2 schreef Stats Watcher:


Error is cumulative, that's how d-GPS and WAAS work. So the guy a mile away from you also has 8 or more meters of error. In exactly the same direction vector as for you. So relative accuracy is virtually +/-0...

On Thursday, August 7, 2014 12:55:16 PM UTC-4, Stats Watcher wrote:
If you analyse the data rate and bandwidth
requirements of the data link you'll realise what you suggest is
impossible

The error correcting computations that I suggest consume ZERO communication bandwidth between FLARM units. The only thing that needs to be broadcast over the data link is the trajectory vector.

In article son_of_flubber writes:
On Thursday, August 7, 2014 11:46:30 AM UTC-4, Ramy wrote:
Even with 8m accuracy, It is more than sufficient for accurate collision=
avoidance, unless someone considers half a wingspan not accurate enough fo=
r collision warning with another glider or obstacle.=20

Prompted to comment here by my Geek Fascination Syndrome (GFS)...

I'd guess that the error on each position fix is not significant. Flarm is =
computing trajectories from multiple position fixes, so a maximum +/- 8M er=
ror for each fix will be smoothed out by the statistical distribution of a =
large number of position fix errors. Some of the errors are + and some are =
-. Average 100 GPS position fixes and you will get a pretty accurate positi=
on.

Not so much as you might expect. The error varies more slowly over
time. If those 100 fixes were taken one every hour or two, then they
would average out much better than fixes taken once per second.

The fact that the error moves more slowly helps the trajectory calculation
as successive samples will have a similar error.

Alan

On 2014-08-07 15:46:30 +0000, Ramy said:

Even with 8m accuracy, It is more than sufficient for accurate
collision avoidance, unless someone considers half a wingspan not
accurate enough for collision warning with another glider or obstacle.

Even the old 100m accuracy when "selective availability" was turned on
was good enough for avoiding antennas or wires etc. It's not as if you
are going to fly as close as you can to them.

Also, note that the errors are not random. Two GPS receivers in the
same area at the same time will show the same error, in the same
direction (to within 6 or 8 m), thus making their relative locations
accurate enough.

At 15:04 07 August 2014, jfitch wrote:

Isn't the Flarm GPS WAAS corrected? The WAAS standard of
performance is
1.6M nominal horizontal (95%). When the NDGPS system is
functional (and GPS
engines take advantage of it) this will be reduced to 15 cm.


Not in Europe as WAAS is US only. A search of the Powerflarm
website returned 0 (zero) occurrences of the term 'WAAS'. So it
seems unlikley.... Same result for NDGPS, so that seems unlikely
too.. If someone finds something different I will obviously be wrong
but until that point it's still +-16m (ie a whole wingspan) at 99%
confidence