repeat occupations with a VRS

On Ohio's VRS I've found little improvement between repeating a shot after 20 minutes or 2 or more hours, and especially if you're looking at accuracies of 0.03m (0.1'). If it's worth getting more than 1 shot I want the difference to be below 0.015m (0.05'), and if it's not I'll repeat the shot again.

This issue is somewhat addressed in the RTN recommendations from NGS, but I was wondering what everyone is using as a criteria.

which basically says if you want 3 cm, then the two occupations should agree to 3 cm.

I agree with this.

Part of the timing between observations has been based on the fact that the change of satellite geometry over time requires a minimum time separation between the first and second observations. And you would not want to repeat the observation on the point 12 or 24 hours later because the geometry of the GPS satellites should be near identical at those times.

But when GLONASS is added to the mix, the orbital period of that constellation is different than for the GPS satellites so I have wondered how that affects strategy.

a reobservation 12 or 24 hours later including both constellation's satellites would be different overall geometry even though the GPS satellites would be in near identical positions.

Actually, the constellation repeats itself every 24 sidereal hours (23h56m solar time). But, as you say, 12 hours is important because the same constellation is on the other side of the earth, but we have spun around to there, so we are seeing the same sats twice a day. Good point.

When I do my analysis, I compute the sidereal time difference for each occupation modulo 24 hours to account for an occupation that might be weeks or months later. I put this in a table in my report showing the repeat baseline agreement. Maybe I should do it modulo 12?

Anyway, I did see a point several years ago that was occupied 2 months apart, 4 hours time of day. So, it was actually at the same constellation configuration. Both occupations had about 9 cm of error in the V, most likely due to multipath repeating itself.

Also a good point about Glonass. But, not everyone is using Glonass, although I think it would be a handicap not to use it, at least where I live in forested/mountainous western PA. Maybe it doesn't matter in flat open country.

:good:

Rather than return to the site to take other VRS observations, why not take a 15 minute static observation and process with OPUS?

> If you want the point to be accurate to, say, 0.03 m horizontally, then what horizontal difference would you accept between the two occupations? i.e. what criteria would cause a third occupation (or more) to be made? I am purposefully not considering the vertical.

Okay, so if the first position was obtained with an uncertainty expected to have a 95%-confidence error ellipse with a semi-major axis of 3cm or less, and the second position is assumed to have been measured by the same process assumed to have the same uncertainty, what difference between the two positions would indicate that the horizontal uncertainty of one or both positions is not likely to be less than 3cm or less? At first impression, I'd say a distance between the two positions of 3cm or more would indicate that the assumed accuracy was wrong.

We work with a minimum of 30 minutes between repeat measures.

I have often wondered, is not the GPS changing constellation by the second . Its orbiting in space and the earth is spinning . seems like way to much changing in the Calculations for it to be correct twice no matter how long between occupations . As long as you break fix and re-occupy.

I choose to Static my control .

> > If you want the point to be accurate to, say, 0.03 m horizontally, then what horizontal difference would you accept between the two occupations?

Okay, on second thought, I'd say that a distance of 4.2cm or more between the two positions, if both are assumed to have uncertainties of the same magnitude, would be the test of whether the assumption were true or not.

[Edit: No, sorry, I used the wrong number of degrees of freedom in computing the test value. I want a third thought on this.]

what I meant to say if it's correct twice no matter the time in between its correct.

A study in the UK would fall in a segment of the GPS constellation I never see under different ionosphere influences than I experience using a completely different form of communication. Other than that it's very similar...
Point being I get different results. I do agree that multipath is one of the main problems with RTK. That is directly related to the lack of education and sound practice exercised by many users.

> > > If you want the point to be accurate to, say, 0.03 m horizontally, then what horizontal difference would you accept between the two occupations?
>
> Okay, on second thought, I'd say that a distance of 4.2cm or more between the two positions, if both are assumed to have uncertainties of the same magnitude, would be the test of whether the assumption were true or not.
>
> [Edit: No, sorry, I used the wrong number of degrees of freedom in computing the test value. I want a third thought on this.]

Here's the basic proposition:

If the North and East components of a VRS position have standard errors of s(N) and s(E), what is the distance that there is a 95% probability that the distance between two successive VRS positions will not exceed?

In other words, if the actual distance is greater than the value calculated from a priori assumptions, it means that the a priori assumptions may well be incorrect.

If s(N) and s(E) are 8.5mm, then the semi-major axis of the 95% confidence error ellipse is 3cm.

If s(N) and s(E) are 8.5mm, then the distance between the two VRS positions, N(1), E(1) and N(2), E(2) is:

SQRT { [N(1)-N(2)]^2 + [E(1)-E(2)]^2 }

So the question is: If the standard errors of N(1), N(2), E(1) and E(2) are all 8.5mm, what is the upper bound of the 95% confidence interval for :

SQRT { [N(1)-N(2)]^2 + [E(1)-E(2)]^2 }

repeat occupations with a VRS new twist

So lets say you have 2 sets of VRS-derived NE coordinates on each of 2 different points on the ground. Would it be possible to estimate the RELATIVE error BETWEEN the 2 different points from just the 4 coordinate pairs? If so, then how? I am thinking about VRS data and ALTA surveys, here.

Actually, I'm in the camp of averaging the occupations, all things being equal, and then evaluating the deltas prior to the actual averaging.

VRS is a finicky critter that is awesome when it's perfect and less than stellar when everything isn't hitting just right. We find that distance to the reference station is just as critical as the cell phone signal component. If either or both of these are weak, then you're pissing in the wind.

If they're both dead on, we have seen occupations of the same object, over the course of two weeks (actually multiple objects), never exceed 0.06', if the object has been measured in at least three times (more than two since you won't know which one is bad if they are different) and then once the deltas are reviewed and found to be within 0.03', averaged. Control and boundary monuments the same.

Once all that is done in the field, it can be adjusted in the office via TBC. When all cylinders are hitting, it's a slick nickle. It did require extra field procedures over the standard rover/base configuration and once we implemented them, we met or exceeded the precision we were used to obtaining.

> Rather than return to the site to take other VRS observations, why not take a 15 minute static observation and process with OPUS?

Because it's instant in the field, and with 15 minutes, you're locked to OPUS-RS, which has hiccups of it's own and many times, you need to "roll your own" with occupation time that small.

I often do fast static, because many of the jobs I do are, first of all, not in areas with VRS, and, second, the points are far enough apart (5 miles to sometimes 30 miles or more), that there is little to be gained.

The project that got me to ask this question was done by someone else. There are about 100 points in a small area (couple of miles long), so VRS makes total sense here. Also, I feel that a 3 minute VRS occupation is more accurate than a 15 minute OPUS-RS occupation, at least more consistent.

You would think that it would be possible to do the exact same thing post processed that the VRS is doing in real time, but that is not the way OPUS-RS works.

The repeat occupation time difference is not so much to make sure it is "correct", but rather to get a different set of satellites to try and see if multipath is present, in my opinion, especially with multiple constellations. In a wide open area, I submit that time difference is not important.

On a side note, I attended a presentation at Trimble Dimensions by Stuart Riley where he described the use in RTK of some of the new systems coming online. One thing that was interesting is that India chose to put one of their two frequencies in the wifi band. In order to use a satellite in RTK you must be able to use both frequencies. SO, the wifi band presents a problem. It should be possible to have many dozens of satellites available in the coming years at any given time. Of course, it depends where in the world you are located. While Glonass, Galileo, and GPS are global, only some of the Chinese satellites are global, many of them are semi-fixed over China. Ditto with India and Japan.

repeat occupations with a VRS new twist

> So lets say you have 2 sets of VRS-derived NE coordinates on each of 2 different points on the ground. Would it be possible to estimate the RELATIVE error BETWEEN the 2 different points from just the 4 coordinate pairs? If so, then how? I am thinking about VRS data and ALTA surveys, here.

Yes, but you'd need to have an a priori estimate of the uncertainties in the VRS-derived coordinates. Two pairs would be a very inefficient means of estimating those uncertainties by themselves.