From: "Dunham, David W" 
Subject: Differential GPS

Scott,
     I think I may have your wrong e-mail address here, so I will check your
address at home, and maybe forward this to you from there.
    Most of what you say about GPS in "How I Beat S/A with a $150 GPS" in 
your
Web site, http://nashville.com/~dega/gps00.htm
is very good and interesting; I'm glad that cheap receivers like your
Eagle Explorer can now record and average data.  In general, we recommend
recording data for two hours or more to ensure an accurate result.  This
has been discussed in previous issues of ON, I think in the March 1997
issue (I don't have that one here), and extensively in two long articles
about GPS in ON Vol. 6, #6 (Dec. 1994) - you might put a reference to those
on your Web site.
   In your "suggested uses" at the end, you describe differential GPS
measurements ("base - rover" mode) for measuring multiple sites
quickly.  What you describe needs to be tested - and it must be stressed
that DGPS only works if the same satellites are being recorded at both the
base station and by the rover, and for processing the data, it is almost
always necessary that the same make and model receiver be used
by both.  You should do some tests, like the extensive ones that you
did with your single GPS unit (that is, convince one of your colleagues
to also get an Eagle Explorer and see what you can do together).
One test like this is described in the Jan. ON (Vol. 6, #13), pp. 281-3.
Since graze sites are usually within a distance of 5 km or less of each
other, the same satellites will usually be visible from all sites so that
the process should work, as long as large buildings, cars, or other
obstructions don't block the view of the sky from both base and rover
GPS antennae.  To be sure, two runs should be made to check for
consistency in the differential results.  It is best if the GPS unit 
actually
identifies the satellites that are used to determine the position, and
records that for each saved point.  Then you can check to be sure
that the same satellites are used by both base and rover. If not, then
2 or 3 runs are necessary to see if the results are consistent.
    Too bad about your Aldebaran experience in South Dakota.
Whatever you do, you should always collect data so that, if
absolutely necessary, someone could return to the observing site.
Count telephone poles along the road to an intersection if nothing
else - most telephone poles have identification numbers that you
could also write down, then pace your telescope's distance
along and perpendicular to the road from the telephone pole.
Although accurate positions of your sites may be lost, there
is some hope of using your data, since the graze was quite well-
observed at several expeditions east and west of South Dakota.
I guess that you were along a north-south road, in which case,
an accurate longitude and height above sealevel can be determined
from the USGS map, even if you only have a poor GPS position for
the latitude.  Then, in a careful reduction of the other expeditions,
a match can be made with shallow features that also produced
events at your stations in South Dakota; by doing calculations for
different latitude values, an interpolation can be performed to
get what must be the correct value.  Your other observed points
should have less weight than the sites from other expeditions with
good positions, but they should be able to fill in some useful detail
of the profile that wasn't covered elsewhere.
     David

From: Scott Degenhardt 

David,

According to my research,  two GPS recievers within 200km of each other
will "see" the same satelites because of their 16,000 mile high orbit. 
I do, of course, plan on testing the differential GPS method.  I am
already confident that it will work, for I already tested the concept
within several hundered feet in one of my tests posted in my article.

Thanks for the input.  I would like to hear from others who try my
methods.  I would especially be interested in any failures since I only
had 100% success.

Scotty

From: "MALEY, PAUL D. (JSC-DO)" 
Subject: GPS problem

David:

I am back from the Jordan conference and saw the GPS solution email. I
could not believe that what the author says was true. So I immediately
sent it to Trimble and Chuck Gilbert in turn sent me a very fast reply
message saying the web site data made erroneous assumptions about S/A.
Chuck has sent the reply to the author and asked him to contact him. I
really wish that you had not emailed everyone with this information
without letting me check it out first. Chuck's primary remark to the
author: 'you are in danger of causing other people to unknowingly
collect data of insufficient quality for their applications.' Wait til
Chuck gets this sorted out with the author and then send out another
email that corrects the situation before it gets out of hand.

From: "Dunham, David W" 
Subject: Re: Differential GPS

 - Yes, but the problem is with satellites that are low in the sky, near the 
5 deg. or
whatever limit of the GPS receiver.  With different obstructions at 
different sites,
the signal strength from the low satellite can be different so that it would 
be
selected for one location and not the other, even if they are only a few 100
meters apart (or even less, as you found out near the satellite dish 
antenna).

But there are other problems, too - I hope that Chuck Gilbert has gotten
in touch with you to find out what really is possible, and not possible, 
with
inexpensive GPS receivers. Also, we look forward to more test results.

David

From: Scott Degenhardt 
Subject: Re: Differential GPS

David,

Bellow is a link from Trimble Navigations Tutorial on GPS and DGPS.  It
states that DGPS correcting works well up to 200km.  I think our 2-5
mile spread on a graze should prove to work well for correcting by fixed
rover method:

If you find a spare half hour every one should "take" this tutorial.  It
starts at:

Scotty

BTW, I called Chuck today and he never returned my call.  He speaks of
some "assumptions" I have made, but the only thing I reported were
actual observations, not assumptions.  I took great pains to be sure
that my method is tried and true.  I have discussed it thoroughly with
the Surveyers at Arnold Air Force Base and they have also blessed my
results.  I know it would be impossible to "unpublish" my results so I
waited until I was sure.

From: Joan and David Dunham 
Date: Sat, 13 Sep 1997 18:26:21 -0400
Scott,
     Thanks for your message.  I'm glad that the surveyors at
Arnold AFB agreed with your results.  The key to single-station
GPS is a suitably long stay time and the ability to store and
average hundreds of readings - it's good that a unit like your
Eagle Explorer can do that.  A number of studies have shown that
a stay time of 2 1/2 to 3 hours is virtually always effective,
more than 99%, in achieving an IOTA-acceptable result.  Shorter
times can often also achieve good results, as your observations
showed, but that is not guaranteed for an unknown site.  Anyway,
I hope that next week, you and Chuck Gilbert can get in touch
with each other to molify any concerns that he may have about
your work.
     Thanks for the Trimble DGPS web sites - I'll look at them
when I get a chance.  I look forward to results of some DGPS
tests that you might do.  I think the key to success is to use
the same equipment at base and rover stations, the ability to
record time-tagged data, and use of the same satellites by both
base and rover - some repeat tests relative to known benchmarks
will verify what can be, or can't be, done.  Does your GPS
receiver display and/or record the satellites that it uses for
each solution?
At 08:17 PM 9/12/97 -0500, you wrote:
>David,
>
>Bellow is a link from Trimble Navigations Tutorial on GPS and DGPS.  It
>states that DGPS correcting works well up to 200km.
That's true for fully differential-capable units that record all the
data for each satellite observed, and then can post-process the data
to use only the satellites in common to both base and rover.  But
for less-capable units, I think much shorter distances are needed to
virtually eliminate the possibility that a low satellite will be used
at one station and not the other.  Probably 10 km or less would 
satisfy that.  So it's NOT a good idea to leave one unit at home as
a base and use the other as a rover at graze sites 100 km away; it's 
always better to select a benchmark, or at least a landmark (such as
a road intersection) whose position can be measured well from a USGS
map, in the graze zone for the base station.  In any real use with 
units not specifically designed for DGPS use, at least 2, and if 
possible 3, runs (sets of measurements for a minute or two at each 
station) should be made to check for consistency in the differential 
results.
David

Date: Mon, 15 Sep 1997 20:13:40 -0500
From: Scott Degenhardt 

David,

I spoke with Chuck Gilbert of Trimble Navigation.  We cleared
up some things, most of which seemed to be misunderstandings
in the intended use and accuracy.

Here are the issues we discussed:

1) I am assuming that the points in my 400 foot circle that I
   plot due to scrambling are centered on my actual location.
   Chuck stated that this isn't necessarily true, that it
   could vary by as much as 6 to 8 feet depending on your
   longitude and latitude.

   Here's where I realized that he was thinking we were
   looking for sub-meter accuracy.  I explained that we were
   looking for 50 foot accuracy!  He said that my method
   should easily reach that kind of tolerance for integration
   times of 3 hours as discussed in earlier communications.
   This means that a single unit integrating at one spot for
   at least 3 hours for now should be a valid reading using
   my methods.  I would like others to give me feedback on
   how it works for them.

2) He mentioned that the patterns of scrambling change over
   the long term, so only time will tell if the above method
   remains valid.

3) This one is the biggest concern.  It affects the "base-
   rover" method.  First, let me explain the issue.

   a) Each of the GPS satellites contain their own unique
   scramble in their signal.  The total of the scramble error
   on the display is the "average" of the satellites it is
   using in its calculation of the displayed position.

   b) Even though 10 satellites may be "visible", the GPS
      will use the number of satellites needed to reduce the
      "GDOP".  GDOP stands for geometric dilution of
      precision.  This number represents the accuracy that
      the GPS can determine your location.  If a pair of
      visible satellites are farther apart than another set
      that are visible, the triangulation from the farthest
      apart pair yields a higher geometric precision than the
      closer together pair.  The software in the GPS will
      choose the most separated sets of satellites so it can
      get the lowest GDOP.  As you can well picture, this
      means that satellites closer to the horizon as well as
      a few over head will give you the biggest
      triangulation, lowering the GDOP.

   What all this means is if the base station is seeing
satellites 4,8,11 and 15 and an obstruction blocks the rover
satellite so that it has to use 4,8,13 and 15 then the
calculation of the effects of scrambling may be a different
answer.

4) I explained to Chuck that the preferred observing site is
   almost always in a field or open area away from structures
   (as to get away from light, etc.).  This should reduce the
   chance of obstructing the "visible" satellites to both the
   base and the rover GPS units.  I also reminded him that
   his own Trimble Navigation On Line Tutorial on GPS and
   DGPS (http://www.trimble.com/gps/fsections/aa_f1.htm)
   states that two units within 200km of each other will be
   seeing the same sets of birds.  I explained that the
   typical graze line was only separated by 1 to 5 miles.  He
   felt a little better about that.

5) HOWEVER, we both agreed that the biggest obstruction
   problem that you can never get rid of is the person
   standing with the GPS in their hand!

   There is an easy fix for this, though.  If we establish
   ahead of time that you will take readings at the base-
   rover locations while ALWAYS facing due South (with your
   back to Azimuth Zero)  and the unit as far from your body
   as is practical, then you will assure that you are both
   obstructing the same part of the sky at the same time.

Some of the first tests I will perform is to purposely block
some satellites at the base and different ones at the rover
and see what this error might add up to.

For now, though, I intend to report all of my single station
GPS readings (while continuing to check there accuracy
manually also) using my method described (and soon to be
updated with the above info) at the link:

http://nashville.com/~dega/gps00.htm

Sincerely,
Scott Degenhardt

From Chuck_Gilbert@Trimble.COM Tue Sep 16 16:50:42 1997
Date: Tue, 16 Sep 1997 10:25:09 -0700

Hiya Scott,

Just a few comments on your email.  

> 1) I am assuming that the points in my 400 foot circle that I
>    plot due to scrambling are centered on my actual location.
>    Chuck stated that this isn't necessarily true, that it
>    could vary by as much as 6 to 8 feet depending on your
>    longitude and latitude.
That is correct, well stated.  

> Here's where I realized that he was thinking we were
>    looking for sub-meter accuracy.  I explained that we were
>    looking for 50 foot accuracy!  He said that my method
>    should easily reach that kind of tolerance for integration
>    times of 3 hours as discussed in earlier communications.
Sorry, no.  I was never under the mistaken impression that you require
sub-meter accuracy.  I have always understood fully that you only
require 50 feet.  My comment on three hours of averaging is not as
stated above.  My comments (on 3 hrs) follow:  

"After averaging continuously for three hours, the result will often be
within 50 feet.   Of course, this implies that sometimes, it will NOT be
within 50 feet.  The only problem is that with the technique in
question, you simply never know when it is or is not really within 50
feet.  
In the end it is simply what may be called a "Clint Eastwood"......
"Do you feel lucky today?"  

However, no reputable scientist who fully understands the dynamics of
the entire system would ever believe or state that only 3 hours of
averaging is always within 50 feet of truth."  I apologize if I was
unclear in  the telephone conversation.  

> 2) He mentioned that the patterns of scrambling change over
>    the long term, so only time will tell if the above method
>    remains valid.
Correct

> 4) I explained to Chuck that the preferred observing site is
>    almost always in a field or open area away from structures
>    (as to get away from light, etc.).  This should reduce the
>    chance of obstructing the "visible" satellites to both the
>    base and the rover GPS units.  
> 
> I also reminded him that
>    his own Trimble Navigation On Line Tutorial on GPS and
>    DGPS (http://www.trimble.com/gps/fsections/aa_f1.htm)
>    states that two units within 200km of each other will be
>    seeing the same sets of birds.  
I agree that the above statement is true when taken as a generalization.


> I explained that the
>    typical graze line was only separated by 1 to 5 miles.  He
>    felt a little better about that.
> 
> 5) HOWEVER, we both agreed that the biggest obstruction
>    problem that you can never get rid of is the person
>    standing with the GPS in their hand!
Absolutely, in practice, the user's head is usually a bigger proiblem
than nearby trees.  Users are looking for and actively avoiding trees.
Unfortunately, most users are not so diligent at keeping their bodies
out of the way of a handheld GPS unit's view of the horizon.  
> Try it, hold something in your hand so that you can read the LCD, and
> see how you would have to contort yourself so as to not obstruct the
> horizon.  (Explain to your spouse prior to this exercise that you HAVE
> NOT lost your mind...:-)
> 
>    There is an easy fix for this, though.  If we establish
>    ahead of time that you will take readings at the base-
>    rover locations while ALWAYS facing due South (with your
>    back to Azimuth Zero)  and the unit as far from your body
>    as is practical, then you will assure that you are both
>    obstructing the same part of the sky at the same time.
This is not a good idea.  To begin with, for users in the northern
hemisphere, the sky to the south is the part of the sky that is most
densely populated with satellites (due north for northern hemisphere
users has the lowest satellite density).  The procedure described above
maximizes the probability of failure.  

Second, no matter which way you face you better be sure to face the same
direction within, say, about one second of arc.  (This is beginning to
sound like synchronized swimming.)  Also make sure that both people are
very much the same size.  

Think about the trigonometry here.  The satellite is about 20,000 km
away; your body is 4 feet away.  As you move your head slightly (say two
inches) to avoid a mosquito, you have just exposed about 100,000
kilometers of orbital path.  It is highly unlikely that your partner did
an identical jig at that same time.  
In short, this is good in theory, but not a practical field technique.  

In short, for RELIABLE scientific data that can be mathematically
supported and validated, there is not yet any practical substitute for
differential processing.  Fortunately, real-time differential data is
transmitted over the entire north American continent by satellite and in
limited coastal regions by Beacons.  Additionally, 

Some good news is that for the 50 foot accuracy that you require, it
should be very achievable in a few years when S/A is turned off;
especially when averaging multiple positions).  

> Some of the first tests I will perform is to purposely block
> some satellites at the base and different ones at the rover
> and see what this error might add up to.
Such tests are academically interesting.  At any given moment there are
many possible answers.  For example, if you have 8 satellites visible
(typically you can see 8 - 10) there are 163 permutations of 4 or more
satellites out of the eight available.  
Therefore, at any given moment, there are 163 different  positions that
could be computed.  Each on of these is valid and will be within 100
meters 95% of the time.  (Note, however, that if the PDOP is higher than
about 4 - 6, you may wish to assume that the noise will increase by the
factor of the PDOP.  e.g.  Positions from a  PDOP of 8 will be two times
noisier than those computed from a  PDOP of 4.)

> For now, though, I intend to report all of my single station
> GPS readings (while continuing to check there accuracy
> manually also) using my method described (and soon to be
> updated with the above info) at the link:
Thank you for the opportunity to send a few words.  I hope that they are
helpful and educational.  Be safe.  I hope that others who may use your
data in the future also have access to information about how the
position was derived.  

Cheers,
Chuck Gilbert

Subject: Re: GPS & IOTA
From: David Wilson
>If you want to see a sample of my work on the subject, visit
>http://www.erols.com/dlwilson/gpsavg.htm
Thanks for the Web site with all your GPS averaging info. - 
it looks interesting.  I'm copying this to the others to
encourage them to look.  
David D.
Joan and David Dunham
7006 Megan Lane
Greenbelt, MD 20770
(301) 474-4722
dunham@starpower.net 

Subject: Re: GPS & IOTA
I have yet a new spin on DGPS that I just learned about this week!  You can
now get realtime differential corrections live on the internet!  Ric Honey
(jehoney@usit.net) succesfully did this and said the position stayed within
a 5 to 10 foot circle ALL the time.  I believe he got this through an SA
watch page at:
http://joe.mehaffey.com/sawatch.htm
Maybe he can reply to all of us on his exact procedure.  I CC'd this to him.
Scotty


From: Ric Honey 
Subject: Re: GPS & IOTA

Here is another link that will provide additional information.
http://www.wsrcc.com/wolfgang/gps/dgps-ip.html
I use a Garmin GPSIII+, their connection cable and SA Watch software.
With an internet connection, SA Watch can link to and pass the DGPS
correction signals to the GPSIII+ which the unit then uses to calculate
the a more accurate position.
The internet based data is generated in Colorado, so I suspect better
accuracy's could be obtained from DGPS transmitters nearer to your
position, but the receiver in this case is a $20 shareware fee.


Return to the
Main Index