So
billvhill,
"So" leaves a lot to be desired as a response when it is at the top of the page with nothing visually preceding it.
Since you have not previously been in the conversation, no one has a clue to what you are remarking about.
Paul in PA
This is from observations made on June 13 and 14 of 2015. I was collecting 20 second averages. I then used a spread sheet to produce 60 second and 180 second points. We don't use kalman filtering so a simple average works in this case.
I collected 348 points on June 13 and 372 points on June 14 from a base that was 2.55 miles away. (All units are US Survey Feet).
The average of all 721 points gave State Plane Coordinates of:
N 6844601.628
E 3090327.36
U 351.73815
The HARN Control Coordinates for this City Control Monument (Kilgore GPS-11) are published:
N 6844601.6326
E 3090327.2509
U 351.81
Based on Translation parameters I've derived for the City of Kilgore Network from HARN to NAD83-2011, I've determined the coordinates for this point to be:
N 6844601.583
E 3090327.37
U 351.7513
Deltas from 721 point average (observed) and translated Control:
N -0.045
E 0.0097
U 0.013
20 second observations:
Std Dev N 0.020
Std Dev E 0.022
Std Dev U 0.048
Spread N -0.065>+0.067
Spread E -0.067>0.074
Spread U -0.130>+0.127
60 second observations:
Std Dev N 0.019
Std Dev E 0.021
Std Dev U 0.046
Spread N -0.057>+0.054
Spread E -0.055>+0.056
Spread U -0.118>+0.116
180 second observations:
Std Dev N 0.017
Std Dev E 0.019
Std Dev U 0.041
Spread N -0.039>+0.036
Spread E -0.043>+0.044
Spread U -0.097>+0.093
I typically use 180 second observation for all boundary and control. The difference in Standard Deviation from 20 seconds to 180 seconds isn't huge, but it is discernable (about a 15% improvement). From what Kent and Bill93 are saying, if I understand:
0.019 usft * 2.44 = 0.0464 usft 95% semi major axis (I picked the higher component)
SQRT(2 * 0.0464 usft ^2) = 0.066 usft positional accuracy between two similarly positioned points.
It should be noted that this doesn't take into account any ppm error from base to rover nor the two theoretical similarly positioned points. And, as has been noted, this control point and the base from which the test was conducted were in very good environments for GNSS.
I hope to spend more time studying the document I linked above as I think it probably explains the concepts behind this pretty thoroughly.
Shawn Billings, post: 384296, member: 6521 wrote: This is from observations made on June 13 and 14 of 2015. I was collecting 20 second averages.
[...]
20 second observations:
Std Dev N 0.020
Std Dev E 0.022
Std Dev U 0.048
Spread N -0.065>+0.067
Spread E -0.067>0.074
Spread U -0.130>+0.127I typically use 180 second observation for all boundary and control. The difference in Standard Deviation from 20 seconds to 180 seconds isn't huge, but it is discernable (about a 15% improvement). From what Kent and Bill93 are saying, if I understand:
0.019 usft * 2.44 = 0.0464 usft 95% semi major axis (I picked the higher component)
SQRT(2 * 0.0464 usft ^2) = 0.066 usft positional accuracy between two similarly positioned points.
If all you have are the standard errors of the Northings or Eastings (without their covariances) and one is consistently greater than the other component, it would make sense to base the error analysis upon the larger standard error, which for the 20-second averages was 0.022 ft.
0.022 ft. x 2.447 = 0.054 ft., which would be the length of the semi-major axis of the 95% confidence error ellipse around the coordinate determined by 20 seconds of data.
The length of the semi-major axis of the 95%-confidence relative error ellipse between two points positioned with the same uncertainties would be:
1.414 x 0.054 ft. = 0.076 ft., which would fail the ALTA/NSPS specification for relative positional uncertainty unless the points were more than 120 ft. apart.
The other element to this is that if one is estimating the uncertainty in coordinates from small data sets, there is an uncertainty attached to the standard error that must be taken into account. Ordinarily, one would expect that repeating occupations significantly different in time is the best strategy for reducing uncertainties and improving reliability.
Kent McMillan, post: 384312, member: 3 wrote: Ordinarily, one would expect that repeating occupations significantly different in time is the best strategy for reducing uncertainties and improving reliability.
There it is; Kent has come up with the solution. Everyone can now start using RTK with confidence!
Kent McMillan, post: 384312, member: 3 wrote: The other element to this is that if one is estimating the uncertainty in coordinates from small data sets, there is an uncertainty attached to the standard error that must be taken into account. Ordinarily, one would expect that repeating occupations significantly different in time is the best strategy for reducing uncertainties and improving reliability.
I ask because I do not know. If a user collected two 20 second observations on the same point, at significantly different times, given the above standard deviations, how would this affect the error estimate of the resulting averaged position?
Shawn Billings, post: 384320, member: 6521 wrote: I ask because I do not know. If a user collected two 20 second observations on the same point, at significantly different times, given the above standard deviations, how would this affect the error estimate of the resulting averaged position?
I would expect that two observations at much different times should tend to cancel biases present in repeats at the same time and would think that the improvements would be from changes in the SV constellations, multipath effects, and atmosphere. This opinion is mainly based upon extensive experience with PPK vectors, however.
