Help with Confidence, Consistency and Bad Fixes

[Jim Frame]

I'm still pretty confused about the confidence and consistency numbers, and how to use them in recognizing a bad fix.

I recently hit a number of control points in a project with RTK using single-base RTN. The RTN base station is close by, less than 4,000 feet from the site. I use the Stop button rather than timed shots, with a minimum of 2 fixed engines. I've been playing around with observation times, and haven't really settled on a scheme. I started off trying for 180 epochs for control points like this, but lately I've reduced that to 60 when I see that the spreads are small. However, when conditions are dicey I often lose patience before I get to 60, especially when I know I'll have other means of verifying the position.

After the RTK observations I went through the points with a total station, using my normal procedures (D&R, careful setups and a pole-mounted miniprism forsight). I was quite surprised by what I found in a couple of instances. I don't like those kinds of surprises, so I'm hoping someone will be able to let me know if I'm dealing with anything unusual, or if this is just the way things are with RTK.

I'll use 3 of the points as examples. All dimensions are in feet.

The first point, 59, is pretty much wide open, and is representative of what I find with clear sites and good SV coverage. I clicked off 60 epochs in 65 seconds, with 5 or 6 engines fixed probably the whole time. RMS stats were good, 0.03' or less; Confidence 3, Consistency 44.95. When I compared the RTK result with my adjusted terrestrial network position (and after accounting for the small datum shifts between the project coordinate system and the RTN coordinate system), the differences were 0.00N, 0.01E and 0.08 elevation. Pretty darn close enough for me.

The next point, 60, is about 8' feet east an 8' chain link fence and a few feet clear of a leafless tree to the northeast. I recorded 64 epochs in 66 seconds, with similar RMS values as at Point 59. Confidence 2, Consistency 14.25. However, the differences with my total station position are 6.36N, 5.69E and 5.32 elevation. Ouch.

The third point, 56, is under some overhanging evergreen limbs (pine and live oak), though it has decent sky view to the south. I recall having trouble getting fixed here, but the stats say that I recorded 155 epochs in 170 seconds, so I guess it wasn't that bad. RMS values were bigger, around 0.1'. Confidence 8, Consistency 118.5. But the differences were 2.58N, 1.91E and 11.23 elevation. Yikes!

So, what should I do differently in order to produce more reliable results, or at least positively identify unreliable results?

Thanks!

 

[Adam]

How was your cell service? I have had funny things happen when I have poor coverage.

 

[Matt Johnson]

I would alway recommend using your own local base station in multipath environments and when better accuracy is needed. The time it takes to setup your own base station will be more than recuperated with faster fixes, especially if you use 5 Hz corrections. I notice that the mount point you connected to was GPS only as well. If a GPS + GLONASS mountpoint is available results should also be better. You need to be using RTK Verification and Validation for all points where there is multipath potential, especially if corrections are from a RTN or distant base station. I have attached a portion of the Quick Start Guide I wrote which will be published soon.

 

[Shawn Billings]

 

[Shawn Billings]

Matt gave you a good summary. I will add that I use a confidence of 10. For control, I observe for 2 minutes. And I will reiterate, use validate or manually force an engine reset at the end of your observation for one last check of your observation.

 

[Nate The Surveyor]

This leads to a question. I use validate with 3 engines, for practically everything, with any multipath. (This comes from using a Legacy E system for years, and shooting, and re shooting, and trying to break the thing) Often, I let it accumulate 300 epochs. Here is the question. IF I press stop, does this prevent it from doing it's validate thing, prematurely stopping the session? I have hesitated to do this, for this very reason. As a substitute, Could I reset the engines, once, before storing?
Thanks.
N

 

[Matt Johnson]

IF I press stop, does this prevent it from doing it's validate thing, prematurely stopping the session? I have hesitated to do this, for this very reason. As a substitute, Could I reset the engines, once, before storing?
Thanks.
N

Yes and Yes

 

[Nate The Surveyor]

OK, then this brings us to the next question, (and it may be answered elsewhere, but there is alot to learn...)
How much consistency, goes into validate, and, is this a user changeable number? We can select how many engines.... but would there be a method to select, such as we have in phase one, how many resets, like phase one has?
Thanks.
Nate

 

[Shawn Billings]

The likelihood of the validate getting the same bad fix as the initial phase 1 test if very remote. It becomes increasingly more remote with more time separation between phase 1 and validation. Rather than increasing the number of fixes at the end, I would encourage users working in difficult environments to make sure you have two or three minutes between the end of phase 1 and validation.

 

[Nate The Surveyor]

I just re watched your video, Shawn. So, if you are storing the point, after say, 300 epochs, instead of say, 20 epochs, then this provides your "Maximum Separation" more separation, and somewhat improves the shot's certainty?

 

[Shawn Billings]

Right on, Nate.

Matt and I have slightly diverging opinions about the confidence counter in phase 1. Matt likes five, I like ten. But it really isn't quite as important in tough spots as validation with a couple of minutes between the end of phase 1 and the start of validation. I like ten because I want to reduce the chances of validation detecting a bad phase 1 fix, having waited through the whole process to get there. Matt likes to get through phase 1 faster. In the end, it's probably a wash AS LONG AS YOU ARE USING VALIDATE.

 

[Nate The Surveyor]

What I'm thinking about here, is doing topo, where I don't care about accuracy more than a foot horizontally, and 0.2' vertically, but faster time is better, BUT I want no shot to be 2' off vertically, and 3 ft horizontally is not going to kill us....
Back in the pre EDM days, We used to do something using precise angles, and a short, precise base line, called a "Tri-angulation". The work phase was "Triangulate that distance" We tried to stay with a triangulation ratio, of 10:1 or less, (9:1 was ok, but 15:1 was not) This built a long skinny triangle, to measure the longer distance. We essentially would be performing a B/B int, to measure across valleys, or rough terrain, that was hard to measure. Mis measuring the baseline distance by 0.01' translated into 0.10' distance error, in this paradigm. (assuming the small angle at the small end of the triangle was perfect....) We'd wind this angle 6 to 10 times, depending on the distance being Triangulated.
In this same sense, LONGER observation times, translate into longer travel of the satellites, reducing the chance of a bad location. (Yields longer distance across the "triangulated base")
If I am doing topo in the woods, and only doing it with 10-20 epochs, is there also a way to make validate work more vigorously, at the end of the observation, (like it does at the beginning) I have left confidence counter at 10 at the beginning as well. But, I'm looking for an understanding of where we begin to "Work against ourselves". Would it be productive, to be able to change this validation number, at the end of an observation?
Thanks.
N

 

[Nate The Surveyor]

From Wiki: https://en.wikipedia.org/wiki/Triangulation "The point can then be fixed as the third point of a triangle with one known side and two known angles."

 

[Shawn Billings]

for topographic work in woods, I use the Last Point white box button. If I start in the open and work my way into the woods on fairly even terrain, I can use look at the Last Point and see if the fix looks right. It's not as automated as our verification and validation procedures, but if you're wanting two work through canopy quickly without waiting for those checks and you can make realistic estimates about how far (horizontally and vertically) you are from you last point, then this makes for a quick, down and dirty check.

 

[Matthew D. Sibole]

I have a different approach than Shawn and Matt. I typically have my confidence set at 25 when in high multipath areas with a validate with 2 fixed engines. This is for boundary corners and control. Reason being is I don't want to sit there for 5-10 min while it is in phase 2 and then it go to validate and dump them all. I have never gotten a bad shot (even with validate turned off) with a consistancy of 25. I will typically have it set to auto-restart 2 times as well giving me 3 points per pin that I use cluster average on.

 

[Nate The Surveyor]

for topographic work in woods, I use the Last Point white box button. If I start in the open and work my way into the woods on fairly even terrain, I can use look at the Last Point and see if the fix looks right. It's not as automated as our verification and validation procedures, but if you're wanting two work through canopy quickly without waiting for those checks and you can make realistic estimates about how far (horizontally and vertically) you are from you last point, then this makes for a quick, down and dirty check.

Shawn, I wish it ALSO showed the bearing, to the last point. In other words, it shows the distance, and the elev. BUT it lacks the 3rd dimension, Side to side. To "Keep on eye" on if it fits or not.
N