A Starnet question

The screen shot doesn't show anything on my phone. I am guessing it failed on the low end? That indicates your error estimates are on the high side. If that's the case changing them will have no significant impact on your results. It can also mean your network geometry is poor or redundancy is lacking.

You have a lot of redundancy, but I can't tell if it is spread around. For instance, umpteen measurements of one variable and one measurement of the others will give you unrealistically good error estimates for those other variables because there is an assumption that the redundancy is not concentrated in only a few variables.

You have the optimum solution with the data the program is given and assumptions it is making. Changing all the standard error estimates by some identical factor will not change the solution, just the reported error ellipses.

This is very similar to Dave Karoly's question in another current thread. I'll paste my comments from there to here, too for reference:
[There is an] assumptions built into the programs regarding whether it is to accept your estimates of std errors or to come up with its own scaling and use yours only for proportioning the error between measurements

See my post in this thread:
Relative Positional Precision - Why did I fail?

Also see late in this thread for the discussion that taught me this
Least Squares Ellipse Confidence Calculations?

You are right, your Chi-square "failure" is because your error estimates are greater than actual. I always like to make them pass, but the coordinates will not change by doing so.

I set my errors to the instrument specs and centering errors at 0.003'. If settings need to be adjusted, I mess with the centering errors. This strategy works for me most of the time. What are your estimates?

Thanks for your help. I am learning the finer details of the program.
I was using 0.007 0.007 0.020 (in metres) for the co-ordinates of some static observations. I tightened these up a bit.
Instrument errors are 2mm plus 2ppm and 3 secs for HA, 10 secs for VA. (I hadn't noticed the zenith setting I may tighten this up as well). 1 second instrument.
I am investigating importing GPS vectors but I am not sure if this would make any difference, There is no redundancy in the GPS baselines, they are all from one base station. Also, Starnet does not support the co-ordinate system I am using so I would need to convert the coordinates after.

You are looking for the overall error factor to be about 1, and you want the individual component error factor to approach 1 as well. Examine those component error factors for clues as to which error estimates to fiddle with. The error factor of your zeniths is 0.284, so I think changing your estimated error for that from 10" to 3" (30% of your first guess) is probably appropriate. Try that first. The error factor on your coordinates is 0.597 so try error estimate of roughly 60% of your first guess (say ....0.004 0.004 0.01) .

You can set up custom projections in StarNet, so it will support your co-ordinate system with a little effort.

Employing your vector data does have advantages, and its not hard to do. Each vector comes with vector quality data which StarNet will use to weight the observation appropriately.

The redundancy is not that the GPS baselines are redundant with GPS baselines but that the GPS baselines are redundant to your traverse geometry for the same points.

Also GPS baselines are inherently redundant because they are the least square results of many epochs of baseline measurements.

Paul in PA

The co-ordinate system I am using requires a shift grid. I know Starnet can use this as it would need it for the OSTN02 projection which is the standard projection in the UK.
But I can't find out how to use a shift grid in the custom projection tab.

I can convert OS to site grid easily though so I might just do that afterwards. The local scale factor with the OS is significant though so I will need to scale the distances measured by EDM.

Thanks again for your help with this.

I figured out my TBC problem was due to a settings issue, or the nut behind the wheel (me).

The vertical centering error probably should be closer to 5mm I would think. It's probably better than that. 2cm or even 1cm is very loose, I bet you can measure up to at least 5mm or better.

squowse, post: 337944, member: 7109 wrote: The co-ordinate system I am using requires a shift grid. I know Starnet can use this as it would need it for the OSTN02 projection which is the standard projection in the UK.
But I can't find out how to use a shift grid in the custom projection tab.

I can convert OS to site grid easily though so I might just do that afterwards. The local scale factor with the OS is significant though so I will need to scale the distances measured by EDM.

Thanks again for your help with this.

Actually, will Starnet automatically adjust my EDM distances using the local scale factor? (If I set the projection details)

Dave Karoly, post: 337951, member: 94 wrote: The vertical centering error probably should be closer to 5mm I would think. It's probably better than that. 2cm or even 1cm is very loose, I bet you can measure up to at least 5mm or better.

The errors were to reflect my estimated accuracy of the coordinates produced by GPS processing.

Paul in PA, post: 337882, member: 236 wrote: The redundancy is not that the GPS baselines are redundant with GPS baselines but that the GPS baselines are redundant to your traverse geometry for the same points.

Also GPS baselines are inherently redundant because they are the least square results of many epochs of baseline measurements.

Paul in PA

2 cautions on that statement.
A vector will have an associated covariance matrix. The longer the occupation the less realistic the numbers are. This is specifically because it will continue to use the root of the number of epochs long past diminishing returns.
In addition, treating epochs as redundant ignores the presence of the same bias and blunder of the observation in every epoch.
IMI RTK vectors are sideshots...

Yes

Cool I will give it a go next week and see how I get on.