The GPS file problems I experienced in Ashtech Solutions seems to have disapeared when I make *.CAR files of the observations and compute them using MSTAR.

The resulting differences, for the same files between A.S. and MSTAR is about 6 feet. That 6 feet is the amount I needed, according to another GPS point that I located conventionally.

Now . . . what I can't figure out, is why it seems A.S. is giving me this bad data using the same files?

Any thoughts?

John, do you still have an older version of AS loaded on a machine somewhere? Just curious if there is a gremlin hiding in this version that is REALLY hard to find, and guess what, you win!

I'm running 2.6

John

Have you tried letting someone else process and adjust this data?

No.

I just figured it was bad data . . . basically un-redeemable.

The data would "look" better if I got rid of about 1/3 of the sats, but I still didin't trust it.

Yesterday I thought I'd fool with changing the PM2 data to Rinex . . . the data "looked" a little better, but I had the same results.

Then . . . on a fluke, I thought, since I already have the Rinex, I'll change them into "CAR" files for MSTAR.

Well . . . I did . . . and I ran the data through MSTAR and the data for the first day's session came out looking just about exactly where I expected it to be . . . which happned to be about 6 feet from the solution given to me in in A.S. 2.6, using the same files, the same control and the same coordinates on that control.

That Rinex files used in a Magellan postprocessing program, gives computations about 6' diferent from Ashtech Solutions?

Is something like this in any way shape or form "normal"?

John, you noted in a thread on the main board that you do not adjust the network of baselines.

When using a single baseline processor like Solutions, it is a good practice to perform an adjustment of the network prior to examining/using the computed coordinates. Without doing that adjustment, it is possible that one or more of the points in your network has an autonomous position assigned to it relative to other points. The adjustment can also identify weak solutions and/or blunders (incorrectly keyed in control coordinates, antenna height errors, etc).

I recommend that you perform the adjustment and compare the coordinates obtained with what you already have. Without additional information, it appears that you have autonomous positions in your data.

Consider..
Put up two receivers on a given day, dump the data into a new project called A. Process the baseline and export the coordinates. The next day, put up the same two receivers on the same two points and dump the data into a new project B. Process the baseline and export the coordinates. Compare the coordinates from both days. The baseline direction and length will agree closely. The coordinates themselves will likely be shifted a few metres due to the fact that autonomous seed positions were used each day.

Modified By Scott Partridge on 12/2/2004 at 9:32 AM

Scott . . . I "process" the positions.

Although I might often have repeat vectors, I rarely have a viable network. It's actually more like I use the GPS radially, repeatedly.

I was under the impression that for "adjustment" to work well, a good, wel-planned network was needed.

I don't thin I have ever setup a really viable network and I virtually ALWAYS traverse between all my points in the process of a survey, using one method to check the other.

Should an "adjustment" ALWAYS be performed . . . even on radial-type data?



Scott . . . I did EXACTLY what you mentioned about the "A" session and the "B" session and the difference from the "true" location was about 5-6 feet west the first day and about 2' east the second day. The same control GPS with the same coordinates were used to constrain the computations.

I used a CONTROL GPS station in Ashtech Solutions PLUS, I tied into a more local GPS control station(intersection/didn't want to set on it). These were not autonomous solutions.

I make RINEX files of the first day.
using the RINEX files and the same control and the same coordinates on that control . . . the first day's data in A.S. made me about 6'east, while the MSTAR placed me virtually exactly on point.

Upon manipulation of sats and masking, A.S. would also eventually place me in the proper position too.
Modified By John Francis on 12/2/2004 at 9:50 AM

Yes, I would do an adjustment on 'radial' data. True, you do not have a network, but the adjustment will reference your 'radial' vectors to your common control point(s). The example I gave is essentially a simple 'radial' survey. Note that the vectors displayed in Solutions are not actual observations. They are merely graphical representations showing you station pairs where simultaneous carrier phase measurements were made.

If you are interested in sending your raw observation files to me along with the control information, I would be happy to look at them for you. I can assess the data in both Solutions and TGO. If anything, I am sure that Thales support would like to see your data to help you determine the cause of the discrepancy.

From your description, it does seems that something out of the ordinary is happening that cannot be readily explained from the information provided.



Modified By Scott Partridge on 12/2/2004 at 10:56 AM

John Francis wrote:

"Should an "adjustment" ALWAYS be performed . . . even on radial-type data?"

Ummm. I was ALWAYS under the impression if you don't do a an adjustment you don't get survey-grade positions. I never gave a single thought to NOT doing an adjustment on every GPS project I have ever done. It just doesn't make sense to me to just go with processed positions. I now understand that after processing the vectors that Solutions will change the autonomous positions of the sites. I also can see why you might think that an adjustment on a radial survey with no triangles might not be necessary, but it just seems like a no brainer to go ahead and do the adjustment anyway. What do you got to loose?

John, I would take a few moments to read the Solutions manual. Here's the first two paragraphs from the "Adjustment" section:


"Adjusting your survey observations is one of the most important tasks to ensure accurate reliable results. A network adjustment is performed to accomplish two results - to test for blunders and errors in the observations (vectors between points in our case), and to compute final coordinates for your survey points which are consistent with the existing control points that you used.

Only data sets with redundant observations (closed loops) benefit from an adjustment. Performing an adjustment on radial vectors (such as those obtained from a kinematic survey with only one base station) will not identify errors in the observations nor improve the accuracy of the points surveyed."


If you're not doing an adjustment, I could understand why you might be missing your points by a couple of feet or more. In the future I would always do at least a minimally constrained adjustment on every GPS project you do in Solutions before using the site positions.

Tommy McClain

... the manual is chock full of useful stuff.

The most important with regards to this discussion is the purpose of the adjustment to "compute final coordinates consistent with the existing control points..."

In John's case, he may have entered in the control point coordinates into his project, but none of the observations are necessarily constrained to that control UNTIL he does the adjustment. This is one of the reasons why an adjustment should always be done, even on a 'radial' GPS survey.

If this is making some people think that they should go back and check their published coordinates for any GPS projects they did relying on the processing results alone without doing an adjustment - yes, they should.

Modified By Scott Partridge on 12/2/2004 at 4:08 PM

Is everyone getting mixed between "processing" and "adjusting"?

These are 2 different things.

An adjustment is made AFTER processing vectors. It is a least-squares adjustment and from what I gather the netwrok has to well planned and setout for it to really work optimally.

I always process my data using a control for constraint. I rarely ever adjust my data, simply because the bulk of me work is further checked(conventionally) and is not of quality geometry.

I beginning to wonder after re-reading your post . . . do you do a processing of your GPS data with NO CONTROL . . . then delineate a control and then "adjust"?

I wonder if, on a session with only 1 "control" and 4 "remote" locations . . . if there would/could be any benefit to hold "processed" locations of higher confidence as more control to THEN perform an "adjustment" on the less confident locations?

One last thought.

If you are setting Control Site coordinates in Solutions for a given point, are you enabling it as 'fixed'?

Do you have more than one control point enabled as 'fixed'?

Page 55 of the latest version of the manual on the FTP site has information about this.

Modified By Scott Partridge on 12/2/2004 at 6:55 PM

I either have a control set as "fixed", or sometimes I let the control float, just to see what the data looks like . . . especially if I plan to return and get more vectors the next day.

Basically speaking, on a job I might have a couple, or even a few repeated vectors, but, in essense, all my GPS gathering has been radial in nature.

The bulk of my GPS work is to get a NORTH and to align my job with the county's aerials. I very, very rarely fail to locate all the GPS points "conventionally", although I Do use the data(even though I don't set up for good geometry), until I have a chance to further verify the location.

I have rarely used the GPS in a "network" situation and I usually have only 1 control for a job.

I would like to try to actually work out a little network someday, but there just isn't enough time for that now.

Through about 4 years of doing this(with the conventional clarification), this is the first, and only time that the GPS seems to have wandered.

Scott,

Thanks. I was wondering if I was the only person who thought that way. Not doing an adjustment on every GPS project just seems like a very good way to get yourself in trouble.


John,

I think everybody understands that "processing" and "adjustment" are two different things. I just think your fooling yourself if you think you get away with not doing an adjustment just because you're doing a radial survey. Even if you're checking those points conventionally.

It does NOT have to be on a well planned network or use quality geometry. It can be used on radial vectors or any other job where you have any kind of triangle. Triangles help out with loop closures. No triangles, no loop closure.

As for my own work, I never do radial work unless it is RTK and even then I don't reprocess the raw RTK data. And yes, I always use at least one control point which I have set to horizontal/vertical fixed. Most of the time that fixed control point comes from sending OPUS data from my Z-Surveyor sitting on a new point for 8 or more hours. I have another Z-Surveyor being used as a base too. Its data is also sent to OPUS. I use the 2nd OPUS position as a check on what Solutions computes after doing a minimally constrained adjustment. If it matches (it usually does within 2cm) I fix that control point in the horizontal and vertical and do another adjustment.

I can understand why a person with only two receivers may only do radial surveys, but if you have more than 2 receivers then I don't see any reason why you shouldn't create triangles and do a proper adjustment. I would think that with all that liability your incurring, not doing a final adjustment would be too much of a risk.

Tommy McClain

Maybe I'm missing something here.

I usually set up between 2-7 GPS at a time. BUT, only one usually occupies an established control point.

So, I have Vectors and often move a number of the GPS throught the session, so, in a day's work, I might have 10 or so vectors.

I load all this shistuff into A.S., check the times and look at the data(but I don't manipulate anything), enter the H.I.'s, plug in the control "fix", then enter in the published coordinates for that point . . . then process.

Now, I've been of the belief that the only truly adjustable network, would involve multiple control "fixes" and redundancy. Even though I have, maybe 10 vectors, I've always viewed them as radial since only one control "fix" was involved. I understand that in a "processing" A.S. does some voodoo, with the vector network, and I assumed that some adjustment was involved. I also felt that if I chose to remove vectors, I should then follow with an adjustment, but I never felt that the networking of all the radial points with no redundancy, was worth an "adjustment".

A.S. manual states that without multiple control and redundancy, that an "adjustment" would not have much, if any effect on the processed vectors. I assumed that included all the vectors from that session if there was not more than one control "fix".

In fact, I am starting to believe that it may be something else, but that missing 6' sure sounds like an 'autonomous position' issue. I first noted this some years ago using the "yellow" software. If a user exported the coordinates of points prior to doing an adjustment, then exported the same points again after the adjustment, in many cases the answers would be quite different.

In the older software, you had to pay attention to the 'flow' of the vectors. Just like in a conventional traverse, you can't survey from A > B, B > C, then D > C. There are no coordinates on D consistent with the remainder of the traverse. In a GPS survey, you could manually change the direction of selected vectors so all of them would flow outwards from a known control point. Simply selecting a known point to start from wasn't sufficient to ensure consistency in the coordinates - an adjustment was often required. It was possible to get it to work without the adjustment, but you had to pay close attention to what was going on as far as the network connectivity went. Even simple 'radial' surveys. The software doesn't necessarily carry the coordinates forward in the direction you think it does unless you set it up explicitly to do so.

That is why it is usually safer to do the adjustment, even on 'radial' surveys, since it removes that uncertainty.

There may (and even likely) be other things affecting John's survey that are not immediately obvious, but I like to start with the simplest, most likely explanation first and eliminate it. Less work that way.

I'm struck by the fact that John says he is eliminating a significant amount of data from the baseline solutions by turning satellites off; it seems to fit better. Maybe so, but as a rule, I don't like to remove any specific data from a solution unless I have a strong empirical reason to do so.

Continually raising the elevation mask is an example. I have seen many raise it to thirty degrees or higher to get a baseline to work. In my opinion you are cutting out a lot of valid data when you do that and significantly reducing the redundancy in the integer ambiguity solution. I rarely raise the mask past twenty degrees in even the worst places. There is more benefit to editing individual satellite tracks at that point than there is raising the mask higher.

Of course, it is just as likely that I don't have a clue about any of this and I'm talking out of the top of my hat. (or 'toque' for my fellow Canadian readers)

Modified By Scott Partridge on 12/2/2004 at 10:38 PM

I agree that it "appears" like an autonomous location, but I assure you that the control was set and tagged to fix both horizontal and vertical.

The next day's data also would appear to be autonomous(moving about 2 feet east of the correct position), but I assure you again that the same control point was used and and set fixed.

Maybe these are autonomous locations for another reason. Maybe it's nothing more than the fact that I'm running windows 98 and haven't completely cleaned out my hard drive for over a year(although, if I remember right, I got the same results on my XP laptop).

I just performed an adjustment and the data did in fact kick back to the west(and a little north), 5.8 or so feet, which brings it to where I needed and in agreement with the MSTAR program.

Now . . . I have never ran the GPS before without having and running checks before, AND, if I remember right, I had a similar problem(I think) back in 2001, where the data just didn't fit.

In my county, we have about 30 or more NGS monuments, 65 permanent county GPS control and a couple hundred county temporary control, along with 1/2' pixel aerials and various other controls from the surrounding counties(in our county). I'm not at a loss for control and this is why I always "check in". This is for sure the first time this has ever happened and possibly the second.

Now, I'm not sure why the A.S. manual seems to indicate that an adjustment would not be worthwhile without considerable redundancy and repeating vectors, for certainly, this job had neither(at least on the session that I ran the adjustment on).

As far as my operations, very often, in the past, I would occupy more than one control, but I would always let one monument cotrol everything, then shift and let another monument control, because I wanted to see the results myself . . . I was never dissatisfied, except now maybe twice in 4 years.

I wished someone from Ahtech would ring in and explain the disclaimer in the A.S. manual about using the adjustment.

In my simple mind, two things happen when you perform an adjustment.

1. The integrity (or strength) of the network is validated. If fact, the Chi Square test is often referred to as the "Goodness of Fit" test - how well do the observations fit together. A priori error estimates are compared to the observed errors, among other statistics.

2. The network is tied to known control

I would agree that for a simple 'radial' survey, there is little in the statistical data that would have much meaning unless vectors had been occupied multiple times.

I do find it fascinating that with up to seven receivers deployed, there is rarely an opportunity to consider other than single vectors from a base. I would think that there must be simultaneous data available between rover pairs in some instances.

Even if you never use any baselines other than the ones established from the base station to the rovers, the adjustment will still reference all of your data to your control point quite efficiently. It is this single function of the adjustment that is applicable to your procedure.

It could simply be the order in which John did the processing and entered the control coordinate information that caused his discrepancy. Which means that others could be doing the same thing.

An important caveat to this entire discussion is that in a 'radial' survey used to tie in legal monuments, redundant observations - made at a different time of day so the satellite constellation has changed - are a must.

a must.
Modified By Scott Partridge on 12/3/2004 at 10:24 AM

I think that way back when . . . when I first began using A.S., I misread th disclaimer to mean that radial surveys AND surveys with just one control point wouldn't benefit from an adjustment.

This mis-reading goes back about 1-1/2 years.

Prviously, I used MSTAR, which gave no such option and I used the ProMark Xcm units a lot with multiple control(of which I could only use one). The data ALWAYS came back "flush", in a greement with the multiple control.

I rechecked a lot of those with A.S. and they also came back "flush" with the multiple control.

The idea of using the "adjustment", along with the mis-reading of the disclaimer had me not using the "adjustment", except in the rare occasions when I actually specifically set up a network.

AS I said, I always(virtually) check all my data conventionally, so I feel confident, respectively in my data for the past year or so.

The rare occasions when I didn't/couldn't check the data conventionally, I had other checks, none-the-less(older surveys of the particular points).

Tis leads me to wonder about the results from MSTAR, especially since MSTAR came completely in agreement with the adjusted data from A.S., although MSTAR has no least-squares capablilities.



update:
I just ran checks on my couple of big jobs I just ALLMOST completed and did an asjustment. O.08 feet shift on one(60 acres) and a 0.15' shift on the other(120 acres).

I've checked a few other jobs and it looks like I dodge a few bullets and my habit of conventionally locating all my points kept me out of trouble on the few occasions that things didn't work out right.
Modified By John Francis on 12/3/2004 at 3:57 PM