GPS Inaccuracy in Cell Phones

On Sun, 12 Apr 2020 13:27:46 -0700, Tom Kunich wrote:


Normal GPS probably doesn't need an accuracy of better than 10 feet or
so.

Too funny by half.

On Sat, 11 Apr 2020 17:56:33 -0700 (PDT), Tom Kunich
wrote:

Strava was consistently 6% shorter than my speedo. MapMyRide wasn't
even in the same ballpark. At first I attempted to explain the
6% loss on the climbing and descending but this course is so flat
that there isn't a total of 100 feet change in altitude over the
entire course. This would make a maximum error of about a 1/10th
of a percent.

In my limited experience, the GPS derived distance traveled is greater
than what is indicated by a cycling computah. My guess(tm) is the
algorithm employed by the GPS software was over-compensating for GPS
errors. Not so short explanation:

If you log the NMEA-0183 GGA and GLL position reports produced by a
commodity GPS, and plot all the points on a map, you'll see a not very
good approximation of a curve or straight line. If you connect each
point in the order in which they are logged (time sequential), the
length of the resultant zig-zag line will be MUCH longer than the path
traveled. At best, any given point is accurate to within a 4.9 meter
(16ft) radius circle. Imagine yourself riding down a straight road,
where the position reports can be randomly placed anywhere within this
circle. That's not going to be even close to a straight line.

However, that's the worst case and without any attempt at error
reduction, error correction, line smoothing, etc. The simplest form
of smoothing the line of travel is to first throw out any wildly
erroneous position reports (usually caused by reflections, diffraction
around buildings, or ground bounce). Next, connect the remaining line
segments, divide each segment in half, and connect all the mid-points
together. Do that a few dozen times, and you will get a fairly smooth
curve. However, since the points can be both ahead and behind the
actually position, the order in which the midpoints are connected is
switched from ordered by time, to ordered by position along the
estimate path traveled. From here, there are a large number of
statistical tweaks that can be used to improve the accuracy. The
resultant smoothed path will be less than the original jagged path (as
long as the path of travel is roughly a straight line). When
smoothing can no longer reduce the path length, then you've done as
much smoothing as possible.

Making 90 degree turns is another interesting problem. If you take
the raw NMEA-0183 data, and plot a path of travel that has a 90 degree
turn, the GPS will usually produce an overshoot on the straight path
before making the turn. It will eventually return to the correct path
of travel, but only after wandering around and adding additional
length too the path of travel. There are processors that look for
such turns and compensate.

That should be enough for now. If you have any questions, feel free
to ask AFTER you've provided my previously requested retraction to
your "nearly 100%" positive test rate claim for Alameda County:
https://groups.google.com/d/msg/rec.bicycles.tech/EuiU0nkYAVM/7M0k1cdxAAAJ


--
Jeff Liebermann
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

On 4/12/2020 6:28 PM, Ralph Barone wrote:
Tom Kunich wrote:
On Sunday, April 12, 2020 at 9:50:47 AM UTC-7, Ralph Barone wrote:

OK, let me try this in more detail. GPS satellites send out a data packet
which contains the satellite’s x, y and z position, as well the the time
from the satellite’s internal atomic clock. If your receiver can get
signals from four satellites, then it can solve the equations to determine
the receiver’s x, y, z and t. If you can see more than four satellites, the
receiver then does a weighted least squares type of fit to get a more
accurate position signal. Look it up if you don’t believe me. The main
difference between civilian and military GPS units is that the military
units have the capability of using the SA (selective availability) signal,
which is a higher accuracy code which is broadcast along with the regular
“low accuracy� signal. There are numerous other small, but important
differences between the two, but SA is the big one.

Cesium "clocks" are not clocks at all. They are an extremely accurate
frequency standard. You still have to use digital electronics around it
to make it into a clock.

And every single satellite has to have synchronized clocks. This accuracy
has to be so accurate that not only to you have to take into account the
distance that the signal has to travel at the speed of light, but the
actual location in its orbit which gives you relative speed hence giving
you the ability to measure the shift in time due to relativity.

In the normal speed regimes it isn't much but it is enough that military
GPS must account for it.

With a somewhat faulty memory I don't believe that they send X,Y, and Z
because they don't know it. Then send a satellite ID and their time. Your
GPS looks up that satellite and its supposed location at that time (with
all of the time corrections) which gives you their position. Comparing
that distance with your clock give you a distance.

So now you have a satellite and its distance from you. That circumscribes
and entire circle that is that distance from that satellite at that time.

Doing this for at least three satellites up to I believe as many as 20
gives you and increasingly accurate position.

Remember not only do you have to correct for speed and distance but
actual processing time in the electronics which includes counting actual operations.

I remember actually writing this stuff but damned if I can remember who for or when.

I also remember that the satellites have to have their clocks corrected
because their orbits are not stable because of atmospheric drag and a
change in relative velocities.

This crap can get so complicated that you could spend your entire life on
nothing else but GPS. There is a lot more that comes into play such as
updating GPS files which hold the exact position at the exactly time.
This is most easily done on ground files. I can't remember if a GPS
receiver keeps the data in files or perhaps like you said that they are
transmitted with the time signal and ID. There would be a hell of a lot
of problems with that. It would mean updating satellite position files
virtually twice a day to keep the satellite correct.

Normal GPS probably doesn't need an accuracy of better than 10 feet or so.


So to summarize, GPS location from your phone does does not depend on the
clock in the phone (although in urban areas where view of the GPS
satellites may be spotty, phones can get a position fix by triangulation
from (known location) cell phone towers.

I've also observed a simple trick built in to our GPS. When we enter a
long tunnel, our position is still shown changing on the screen, despite
the lack of satellite data. The system is smart enough to extrapolate
from our speed when entering. I've seen it demonstrate an error when
I've had to slow down significantly in a tunnel.


--
- Frank Krygowski

On Sunday, April 12, 2020 at 3:28:28 PM UTC-7, Ralph Barone wrote:
Tom Kunich wrote:
On Sunday, April 12, 2020 at 9:50:47 AM UTC-7, Ralph Barone wrote:

OK, let me try this in more detail. GPS satellites send out a data packet
which contains the satellite’s x, y and z position, as well the the time
from the satellite’s internal atomic clock. If your receiver can get
signals from four satellites, then it can solve the equations to determine
the receiver’s x, y, z and t. If you can see more than four satellites, the
receiver then does a weighted least squares type of fit to get a more
accurate position signal. Look it up if you don’t believe me. The main
difference between civilian and military GPS units is that the military
units have the capability of using the SA (selective availability) signal,
which is a higher accuracy code which is broadcast along with the regular
“low accuracy� signal. There are numerous other small, but important
differences between the two, but SA is the big one.

Cesium "clocks" are not clocks at all. They are an extremely accurate
frequency standard. You still have to use digital electronics around it
to make it into a clock.

And every single satellite has to have synchronized clocks. This accuracy
has to be so accurate that not only to you have to take into account the
distance that the signal has to travel at the speed of light, but the
actual location in its orbit which gives you relative speed hence giving
you the ability to measure the shift in time due to relativity.

In the normal speed regimes it isn't much but it is enough that military
GPS must account for it.

With a somewhat faulty memory I don't believe that they send X,Y, and Z
because they don't know it. Then send a satellite ID and their time. Your
GPS looks up that satellite and its supposed location at that time (with
all of the time corrections) which gives you their position. Comparing
that distance with your clock give you a distance.

So now you have a satellite and its distance from you. That circumscribes
and entire circle that is that distance from that satellite at that time.

Doing this for at least three satellites up to I believe as many as 20
gives you and increasingly accurate position.

Remember not only do you have to correct for speed and distance but
actual processing time in the electronics which includes counting actual operations.

I remember actually writing this stuff but damned if I can remember who for or when.

I also remember that the satellites have to have their clocks corrected
because their orbits are not stable because of atmospheric drag and a
change in relative velocities.

This crap can get so complicated that you could spend your entire life on
nothing else but GPS. There is a lot more that comes into play such as
updating GPS files which hold the exact position at the exactly time.
This is most easily done on ground files. I can't remember if a GPS
receiver keeps the data in files or perhaps like you said that they are
transmitted with the time signal and ID. There would be a hell of a lot
of problems with that. It would mean updating satellite position files
virtually twice a day to keep the satellite correct.

Normal GPS probably doesn't need an accuracy of better than 10 feet or so.


So to summarize, GPS location from your phone does does not depend on the
clock in the phone (although in urban areas where view of the GPS
satellites may be spotty, phones can get a position fix by triangulation
from (known location) cell phone towers.

There is no clock per se in your phone. If you power down your phone, wait ten minutes and start it again it comes up with the correct time to the minute. Most phones have an app that mimics a clock. It is set from your nearest cell phone tower.

On Sunday, April 12, 2020 at 5:25:25 PM UTC-7, news18 wrote:
On Sun, 12 Apr 2020 13:27:46 -0700, Tom Kunich wrote:


Normal GPS probably doesn't need an accuracy of better than 10 feet or
so.

Too funny by half.

Why don't you tell us how old you are and what you've ever done for a living?

Tom Kunich wrote:
On Sunday, April 12, 2020 at 3:28:28 PM UTC-7, Ralph Barone wrote:
Tom Kunich wrote:
On Sunday, April 12, 2020 at 9:50:47 AM UTC-7, Ralph Barone wrote:

OK, let me try this in more detail. GPS satellites send out a data packet
which contains the satellite’s x, y and z position, as well the the time
from the satellite’s internal atomic clock. If your receiver can get
signals from four satellites, then it can solve the equations to determine
the receiver’s x, y, z and t. If you can see more than four satellites, the
receiver then does a weighted least squares type of fit to get a more
accurate position signal. Look it up if you don’t believe me. The main
difference between civilian and military GPS units is that the military
units have the capability of using the SA (selective availability) signal,
which is a higher accuracy code which is broadcast along with the regular
“low accuracy� signal. There are numerous other small, but important
differences between the two, but SA is the big one.

Cesium "clocks" are not clocks at all. They are an extremely accurate
frequency standard. You still have to use digital electronics around it
to make it into a clock.

And every single satellite has to have synchronized clocks. This accuracy
has to be so accurate that not only to you have to take into account the
distance that the signal has to travel at the speed of light, but the
actual location in its orbit which gives you relative speed hence giving
you the ability to measure the shift in time due to relativity.

In the normal speed regimes it isn't much but it is enough that military
GPS must account for it.

With a somewhat faulty memory I don't believe that they send X,Y, and Z
because they don't know it. Then send a satellite ID and their time. Your
GPS looks up that satellite and its supposed location at that time (with
all of the time corrections) which gives you their position. Comparing
that distance with your clock give you a distance.

So now you have a satellite and its distance from you. That circumscribes
and entire circle that is that distance from that satellite at that time.

Doing this for at least three satellites up to I believe as many as 20
gives you and increasingly accurate position.

Remember not only do you have to correct for speed and distance but
actual processing time in the electronics which includes counting actual operations.

I remember actually writing this stuff but damned if I can remember who for or when.

I also remember that the satellites have to have their clocks corrected
because their orbits are not stable because of atmospheric drag and a
change in relative velocities.

This crap can get so complicated that you could spend your entire life on
nothing else but GPS. There is a lot more that comes into play such as
updating GPS files which hold the exact position at the exactly time.
This is most easily done on ground files. I can't remember if a GPS
receiver keeps the data in files or perhaps like you said that they are
transmitted with the time signal and ID. There would be a hell of a lot
of problems with that. It would mean updating satellite position files
virtually twice a day to keep the satellite correct.

Normal GPS probably doesn't need an accuracy of better than 10 feet or so.


So to summarize, GPS location from your phone does does not depend on the
clock in the phone (although in urban areas where view of the GPS
satellites may be spotty, phones can get a position fix by triangulation
from (known location) cell phone towers.

There is no clock per se in your phone. If you power down your phone,
wait ten minutes and start it again it comes up with the correct time to
the minute. Most phones have an app that mimics a clock. It is set from
your nearest cell phone tower.


Tom, please note that it was YOUR hypothesis that GPS path length errors in
phone mapping applications might be the result of the GPS using “its
internal clock� instead of GPS time.

Secondly, your pedantic assertions that neither a cesium clock nor a cell
phone are a clock because they need to be set also apply for pretty much
any other device which humans call clocks.

Tom Kunich wrote:
On Sunday, April 12, 2020 at 5:25:25 PM UTC-7, news18 wrote:
On Sun, 12 Apr 2020 13:27:46 -0700, Tom Kunich wrote:


Normal GPS probably doesn't need an accuracy of better than 10 feet or
so.

Too funny by half.

Why don't you tell us how old you are and what you've ever done for a living?


Surveying. GPS guided tractors and bulldozers.

Ralph Barone wrote:
Tom Kunich wrote:
On Sunday, April 12, 2020 at 3:28:28 PM UTC-7, Ralph Barone wrote:
Tom Kunich wrote:
On Sunday, April 12, 2020 at 9:50:47 AM UTC-7, Ralph Barone wrote:

OK, let me try this in more detail. GPS satellites send out a data packet
which contains the satellite’s x, y and z position, as well the the time
from the satellite’s internal atomic clock. If your receiver can get
signals from four satellites, then it can solve the equations to determine
the receiver’s x, y, z and t. If you can see more than four satellites, the
receiver then does a weighted least squares type of fit to get a more
accurate position signal. Look it up if you don’t believe me. The main
difference between civilian and military GPS units is that the military
units have the capability of using the SA (selective availability) signal,
which is a higher accuracy code which is broadcast along with the regular
“low accuracy� signal. There are numerous other small, but important
differences between the two, but SA is the big one.

Cesium "clocks" are not clocks at all. They are an extremely accurate
frequency standard. You still have to use digital electronics around it
to make it into a clock.

And every single satellite has to have synchronized clocks. This accuracy
has to be so accurate that not only to you have to take into account the
distance that the signal has to travel at the speed of light, but the
actual location in its orbit which gives you relative speed hence giving
you the ability to measure the shift in time due to relativity.

In the normal speed regimes it isn't much but it is enough that military
GPS must account for it.

With a somewhat faulty memory I don't believe that they send X,Y, and Z
because they don't know it. Then send a satellite ID and their time. Your
GPS looks up that satellite and its supposed location at that time (with
all of the time corrections) which gives you their position. Comparing
that distance with your clock give you a distance.

So now you have a satellite and its distance from you. That circumscribes
and entire circle that is that distance from that satellite at that time.

Doing this for at least three satellites up to I believe as many as 20
gives you and increasingly accurate position.

Remember not only do you have to correct for speed and distance but
actual processing time in the electronics which includes counting actual operations.

I remember actually writing this stuff but damned if I can remember who for or when.

I also remember that the satellites have to have their clocks corrected
because their orbits are not stable because of atmospheric drag and a
change in relative velocities.

This crap can get so complicated that you could spend your entire life on
nothing else but GPS. There is a lot more that comes into play such as
updating GPS files which hold the exact position at the exactly time.
This is most easily done on ground files. I can't remember if a GPS
receiver keeps the data in files or perhaps like you said that they are
transmitted with the time signal and ID. There would be a hell of a lot
of problems with that. It would mean updating satellite position files
virtually twice a day to keep the satellite correct.

Normal GPS probably doesn't need an accuracy of better than 10 feet or so.


So to summarize, GPS location from your phone does does not depend on the
clock in the phone (although in urban areas where view of the GPS
satellites may be spotty, phones can get a position fix by triangulation
from (known location) cell phone towers.

There is no clock per se in your phone. If you power down your phone,
wait ten minutes and start it again it comes up with the correct time to
the minute. Most phones have an app that mimics a clock. It is set from
your nearest cell phone tower.


Tom, please note that it was YOUR hypothesis that GPS path length errors in
phone mapping applications might be the result of the GPS using “its
internal clock� instead of GPS time.

Secondly, your pedantic assertions that neither a cesium clock nor a cell
phone are a clock because they need to be set also apply for pretty much
any other device which humans call clocks.

Shouldn’t have written this while on my way to bed. The first sentence
should have read:

Tom, please note that it was YOUR hypothesis that GPS path length errors in
phone mapping applications might be the result of the cell phone using
“its internal clock� instead of GPS time.

Radey Shouman wrote:
Tom Kunich writes:

On Sunday, April 12, 2020 at 9:50:47 AM UTC-7, Ralph Barone wrote:

OK, let me try this in more detail. GPS satellites send out a data packet
which contains the satellite’s x, y and z position, as well the the time
from the satellite’s internal atomic clock. If your receiver can get
signals from four satellites, then it can solve the equations to determine
the receiver’s x, y, z and t. If you can see more than four satellites, the
receiver then does a weighted least squares type of fit to get a more
accurate position signal. Look it up if you don’t believe me. The main
difference between civilian and military GPS units is that the military
units have the capability of using the SA (selective availability) signal,
which is a higher accuracy code which is broadcast along with the regular
“low accuracy� signal. There are numerous other small, but important
differences between the two, but SA is the big one.

Cesium "clocks" are not clocks at all. They are an extremely accurate
frequency standard. You still have to use digital electronics around
it to make it into a clock.

Pendulum "clocks" are not clocks at all. They are a moderately accurate
frequency standard. You still have to use mechanical parts around it to
make it into a clock.

Sorry, no opinion on GPS, just had to point that out.


Pretty much all clocks consist of an oscillator (eg: the earth’s rotation),
coupled to a decoding device (eg: a stick casting a shadow onto a dial),
with a means to set the clock to conform to some arbitrary standard of when
“now� is (eg: rotating the dial), so calling out our HP cesium clock as
“not a clock� seemed weird, since it did have all three.

As an interesting sideline, we purposely set our HP 100 ns earlier than GPS
time so that when we were comparing the accuracy of the 1 PPS output of the
cesium to the GPS time receiver under test, we could trigger the scope/time
interval counter from the HP and always be guaranteed that the GPS 1 PPS
would show up around 100 ns later. We would then log the delay between the
two signals, subtract 100 ns, and have a plot of the accuracy and jitter of
the device under test.

On Sunday, April 12, 2020 at 10:09:00 AM UTC-5, Tom Kunich wrote:
On Saturday, April 11, 2020 at 6:49:59 PM UTC-7, Ralph Barone wrote:
Tom Kunich wrote:
After trying to use these cell phone apps that us GPS to give distance
traveled I've looked up data trying to discover why these apps are so far out of kilter.

This isn't just my assessment, before when my friend still lived in
Castro Valley he was using some app that always had noticeably less
mileage than my or his wife's speedometers.

As I said before, Strava was consistently 6% shorter than my speedo.
MapMyRide wasn't even in the same ballpark. At first I attempted to
explain the 6% loss on the climbing and descending but this course is so
flat that there isn't a total of 100 feet change in altitude over the
entire course. This would make a maximum error of about a 1/10th of a percent.

Then I figured that it had to do with latitude corrections not being
properly made. While that may be part of the problem it isn't very likely
since rather than the old geosynchronous orbits that were originally
used, they are using an entire spate of GPS satellites that are in all
sorts of orbits and at a distance that allows them to make complete
orbits twice a day. They sent out their precise point in the orbit and if
you have exact time and several satellites you can calculate (triangulate
if you will) your position. The best GPS units can handle 20 satellites
at once and can get your position down to mm.

Question: could this error be due to time since the time of the GPS apps
is derived not internally but from the cell phone center which can be
several miles from you and transferred through several cell towers?

No. GPS predates cell phone systems that send out time. GPS receivers get
their time signal from the satellites themselves.

This seems pretty likely to be the problem to me. Being just one
microsecond off could increase the error band to 1131 feet on a single
satellite and it is unlikely that that apps track more than two or three
satellites at a time. On a 3 satellite lock this would interestingly
enough give an error band of about 6%.

Since this ride is local it would always be using the same cell CPU to be
getting the time and so you would expect a pretty consistent error always
in the same direction while on a long distance ride the errors would be
divided between plus and minus errors and would average out to little or
no discernable errors.

Likewise the error on my friend's app for the Palomares ride always gave
approximately the same sort of error relatively reliably.

Military grade GPS that has an entirely different approach to getting
time. I have designed clocks that lost one microsecond in a week. This
requires a crystal oven that can regulate crystal oscillator temperatures
to within a tenth of a degree C or better per 24 hour time period. The
crystal is cut on a particular bias of the grain which I cannot remember.
Crystals grow like a tree with a grain. This high accuracy crystal is
frequency controlled by its size, the placement of the electrodes and
temperature. The highest accuracy ones are extremely wasteful of these
crystal structures. If you have precise location of where your position
is you can use this to get precise time from your transmissions from the
satellites and set the time precisely the first time. From that point you
can find your exact position and then again with multiple satellites
reset your timer as often as you find a significant timing error. These
GPS systems use up to 20 satellites encoded in a manner that commercial
devices cannot unlock. This is how you can target any location from any
other location with mm distance errors. This means that the ONLY accurate
means of touching that spot is from perfectly vertical from the target.
Also since all of these satellites pass an un-named precisely known point
at a precisely known distance you can reset the satellite timers which
are necessary for this system.

This is the sort of thing I was paid to know and use. This was why I
could get a quarter of a million bucks a year and have people standing in
line for my services. Too bad about the concussion which caused me a
couple of years out of the field at just exactly the wrong time - when
one generation of managers that knew me were retiring and another which
didn't took over. Resume's mean very little in the immediate area you're
looking, because most of the hiring is done from recommendations of high
level managers. But the resume alone is enough to get me jobs in Illinois
or Texas or Arizona or Florida (???) were I willing to relocate. Last
week I had two offers from aerospace firms in San Diego but I can't get
the wife to move away from her children and grandchildren. And I really
don't care to go back through the effort of showing boards of directors
and CEO's what I can do. When I first recovered I should have immediately
taken the jobs I was offered as some of these managers were still working
and wanted me at places like Lawrence Labs in Berkeley and Sandia in
Livermore. But quite frankly with my spotty memory I didn't think I still
had it. Finally I was talked into taking a job for a friend and I told
him that I would not allow him to pay me. But as soon as I sat down with
the schematic and idiot code that the previous programmer had made it was
all as fresh as yesterday. My total pay for 8 months was gas money to the
job location and a $2,000 H-P laptop which I haven't even opened since the end of that job.





While talking about how GPS predates cell phones explain my Sony watch which always has the correct time which I bought in the 70's and is still working today.

That is quite funny.

You made my day. :-)

Andy