Yesterday, we were setting up control with GPS on a 100 acre cut out from a 156 acre tract we surveyed in '01. Rather than use the RTK, we decided to go with post-processed fast static occupations on each of the control points we created. I used the rover with the data collector at the pole just like I would have with RTK, except no radio link, so the rover was using WAAS corrected positions. For each point I would observe for 3-5 minutes depending on what the point was. As I collected the point in the receiver memory for post processing, the data collector was averaging the WAAS coordinates as well. I localized on the first set up in the field and was surprised how well I was able to navigate to remaining points. We've used WAAS in other receivers with good results, but this was really surprising to me. Today, after post processing, I reprocessed my rw5 file from the data collector, removing the localizations I'd done in the field and developed a point list of the WAAS corrected points. I then did a Helmert rigid body transformation to from the WAAS coordinates to the post processed coordinates determined from my post processing to CORS. The report shows the differences for the 16 points we tied with the rover. Generally the results were subfoot, with some being very subfoot. For jobs that don't require a need for RTK precision immediately in the field, particularly when radio/cell coverage might be an issue, it's nice to know free autonomous results like this are available (units are US Survey Foot):
what kind of receivers are you using?
I think topcon charges a few hundred bucks to enable waas, which is a ripoff imho.
Altus APS-3 first generation receivers. Yeah I agree about the charge.
What is your opinion of the ALTUS system and what controller are you using?
:good:
Hi Shawn,
Would you say this data illustrates precision or accuracy?
WAAS precision can be pretty good if the data is collected in close proximity (with respect to time), but accuracy is a different story. WAAS accuracy can be good, but it's hard to fathom that it's this good.
If your post-processed data is referenced to NAD83/2011 (epoch 2010.0), then you could take the WAAS-corrected positions, which are in ITRF08 (epoch 2013.5), run them through HTDP and see how they compare. It would be interesting to see those results.
It would also be interesting to occupy the points again with WAAS and compare with the previous results.
A couple of years ago, I ran some tests with WAAS collecting data over a 24 hr period, then ran it through a stats program that randomly pulled xxx epochs from the 24 hr data set (~86,000 epochs) and computed the accuracy stats according to NSSDA standards. The results:
Garmin Extrex:
68% = 3.04m
95% = 4.69m
99% = 5.81m
High-performance L1 receiver (similar to what you used) using a Javad antenna mounted on a rooftop (test environment doesn't get any better):
68% = 0.45m
95% = 0.69m
99% = 0.86m
These are accuracy results compared to the antenna location that was referenced to the NSRS.
You can also take a look at the quarterly WAAS performance report published by the National Satellite Test Bed at http://www.nstb.tc.faa.gov/reports/waaspan43.pdf. Table 2-1 illustrates the accuracy of WAAS at each WAAS reference station location, with 95% confidence (horizontal), over a continuous three month period (~7,500,000+ epochs).
Thanks for posting your results. I've been a big believer (and user) of WAAS for many years.
Eric
> Yesterday, we were setting up control with GPS on a 100 acre cut out from a 156 acre tract we surveyed in '01. Rather than use the RTK, we decided to go with post-processed fast static occupations on each of the control points we created. I used the rover with the data collector at the pole just like I would have with RTK, except no radio link, so the rover was using WAAS corrected positions. For each point I would observe for 3-5 minutes depending on what the point was. As I collected the point in the receiver memory for post processing, the data collector was averaging the WAAS coordinates as well. I localized on the first set up in the field and was surprised how well I was able to navigate to remaining points. We've used WAAS in other receivers with good results, but this was really surprising to me. Today, after post processing, I reprocessed my rw5 file from the data collector, removing the localizations I'd done in the field and developed a point list of the WAAS corrected points. I then did a Helmert rigid body transformation to from the WAAS coordinates to the post processed coordinates determined from my post processing to CORS. The report shows the differences for the 16 points we tied with the rover. Generally the results were subfoot, with some being very subfoot. For jobs that don't require a need for RTK precision immediately in the field, particularly when radio/cell coverage might be an issue, it's nice to know free autonomous results like this are available (units are US Survey Foot):
>
>
Good point Eric. It'll require more testing to determine the accuracy of this setup. However these points were observed over a seven hour period with pretty good repeatability. I suspect I'm seeing precision instead of accuracy because the translations aren't what I expected...particularly in easting. That six foot translation should have been about 3 feet. In practical use, I'm hypothesizing that in a work day you could localized on a known point and expect similar precision for the day. Like you I've seen submeter with waas for a long time with great benefit. Sub foot is a new one on me.
Interesting data, but one important aspect is missing: how fast the error changes.
If random 1-second epochs have xx meter standard deviation and you average for 100 second periods, the average error in the 100-sec numbers could be anything from xx meter (if the error varied slowly) to xx/10 meter (if the error varied very quickly as white noise).
There are multiple sources of error. The receiver has its own noise that probably accounts for a lot of short-term changes. That's one reason the Garmin is worse than the better receiver (although their receiver technology has improved since the original Etrex). There are computational effects just below display precision (trig function accuracy or accumulated roundoff) that in an older Garmin amounted to 2 or 3 feet and can be demonstrated in several ways (e.g. indicated distance between waypoints). And then there are the localized ionospheric variations that don't get represented in the WAAS data, and might account for the majority of the error in the high grade equipment.
I once did some long-term averages with a WAAS-corrected Garmin, and I saw the error of short-term averages drifting around over a longer period. On one day you might get pretty stable readings for a couple hours, and then see the xx meters movement in 20 minutes. The next day the pattern would be different. The average of 3 or 4 days with an hour per day tended to be within a meter of a known WGS84 position, and the ultimate for any amount of averaging never got below a half-meter due to the computational problems.
> That is outstanding; best I've ever heard with WAAS. Now I'm inspired to try that again out on the baseline.
>
> If WAAS is that good; imagine other implementations of SSR/RT-PPP; like Fugro, Veripos, and RTX. Already being released on a number of rovers this year; gonna be a whole new game.
Hmmm, sounds like the kind of thing that might give a dumb dirt surveyor like me quite an education.
(lurker out)
Thanks for your comments.
To your comment about rate of change, in their ag receivers, Trimble claims 6"-8" accuracy with WAAS if it's collected within 15 minutes of the last position. They call this pass-to-pass accuracy. However, I think accuracy is the incorrect term. It's really precision because if you go back to the same point a week (or a day for that matter) later, you won't be within 6"-8"...not even close to that.
The error rate of change of WAAS receivers that you mention is also dependent on a well-behaved ionosphere. Increased solar activity can disturb the ionosphere. WAAS accuracy is highly-dependent upon being able to accurately model the delay of the GPS measurements as they pass through the ionosphere.
I don't know if Garmin's receiver technology has improved much. They used to use a Sirf GPS chip which is not that bad. I don't think they don't use Sirf any more. They likely use a local (Taiwan/Chinese) chipset like MTK. There's so many to choose from these days.
However, the limitation of consumer receivers (and mobile phones) generally isn't the GPS receiver itself, but rather poor GPS antenna that they use. Sometimes, the antenna is just a trace line on a printed circuit board.
I'm a bit pessimistic that you'd see 1 meter accuracy from a Garmin (or any consumer receiver) after logging for a few days. It might hit less than a meter occasionally as the positions wander around, but if you apply NSSDA standards in computing the 68%, 95%, 99% confidence values over three (or more) days of data, I think you'd see close to what I saw in my experiment a couple of years ago.
Eric
> Interesting data, but one important aspect is missing: how fast the error changes.
>
> If random 1-second epochs have xx meter standard deviation and you average for 100 second periods, the average error in the 100-sec numbers could be anything from xx meter (if the error varied slowly) to xx/10 meter (if the error varied very quickly as white noise).
>
> There are multiple sources of error. The receiver has its own noise that probably accounts for a lot of short-term changes. That's one reason the Garmin is worse than the better receiver (although their receiver technology has improved since the original Etrex). There are computational effects just below display precision (trig function accuracy or accumulated roundoff) that in an older Garmin amounted to 2 or 3 feet and can be demonstrated in several ways (e.g. indicated distance between waypoints). And then there are the localized ionospheric variations that don't get represented in the WAAS data, and might account for the majority of the error in the high grade equipment.
>
> I once did some long-term averages with a WAAS-corrected Garmin, and I saw the error of short-term averages drifting around over a longer period. On one day you might get pretty stable readings for a couple hours, and then see the xx meters movement in 20 minutes. The next day the pattern would be different. The average of 3 or 4 days with an hour per day tended to be within a meter of a known WGS84 position, and the ultimate for any amount of averaging never got below a half-meter due to the computational problems.
Hi Shawn,
It would be an interesting to compare the WAAS coordinates with the post-processed coordinates. If you have the 1Hz coordinates from WAAS for each session, it would be interesting to run it through an analysis to compute the 68%, 95% and 99% confidence values. I've got a custom-built piece of software that will perform that calc.
Otherwise, even taking the averaged WAAS coordinate from the session, running it through HTDP to shift it to NAD83/2011, and comparing to the post-processed coordinate would be interesting to see.
As I mentioned to Bill93, you can carry the WAAS iono model forward and play some tricks to smooth it out via modeling. Some high-end receiver designers do this to optimize WAAS performance for non-aviation users, but the model is highly dependent on a well-behaved ionosphere.
The ionosphere is a little squirrely these days. We're actually at the solar max this month of the current 11-year solar cycle. It affects high-performance WAAS receivers, but I'm not really sure how much. The FAA has some cool new WAAS tools that I'm looking at to try to gain a better understanding.
Eric
> Good point Eric. It'll require more testing to determine the accuracy of this setup. However these points were observed over a seven hour period with pretty good repeatability. I suspect I'm seeing precision instead of accuracy because the translations aren't what I expected...particularly in easting. That six foot translation should have been about 3 feet. In practical use, I'm hypothesizing that in a work day you could localized on a known point and expect similar precision for the day. Like you I've seen submeter with waas for a long time with great benefit. Sub foot is a new one on me.
Coast Guard DGPS in the Real World - Wow!
Old Trimble Pathfinder Pro L1 only.
Used Coast Guard DGPS (150 mi away), one hour occupation.
Post process (actually just average of data) with 0.2 ft accuracy result.
30 min maybe 0.5 ft
PS Make sure you have matching Epochs if comparing to record positions.
