Microsurvey Star*Net vs. Carlson Move3

Move3 GUI is better than that of Star*Net.

Related Question for the Resident Geodesists

I started reading up on Move3 recently since I now have Leica Geo Office available for my use and the adjustment process in Geo Office is, apparently, performed by a licensed version of Move3 built into the Leica software.

The Move3 literature repeatedly mentions that it is designed based on the Delft school of adjustment theory; which I assume is somehow related to the works of P. J. G. Teunissen. So what is the difference between this "school” of network adjust theory and the other (whatever they may be) “schools” of adjustment theory?

Have not seen the info on Move3 but will look for it.

I have used StarNet since '95 for all my adjustments and can't
imagine anything more flexible. Now with the new Ver 7 (MicroSurvey)
the interface is better. Agreed .. the converters should be included
with the product.

Related Question for the Resident Geodesists

I also have wondered about this. Based on my google searches, it appears to me that Move3 is the only group using the term "Delft school of adjustment theory", as if it is a distinct method of least squares adjustment.

Donnie

I am very interested in Move3. I did buy a demo ($149, they send you a key to use for 60 days) a few years ago, at that time I didn't like it enough to make the change from Geolab. I had a lot of problems getting it to work they way I wanted it to. But, I do follow them on linkedin, and I see that they are constantly making improvements. I think I will give it another look. They seem to be very responsive to suggestions and bug reports.

I have been using Geolab since 1986. It is a very robust program, and I trust the results. But, it has not been updated since 2001. The developer always promises that there is a new release coming soon, but it never happens. One issue I have with it is that I cannot view network diagrams, it causes the program to blow up. I reported this to him, he said it must be the graphics card on my computer. Well, since then it has done the same thing on 3 computers and my laptop.

To me, it is very unfortunate that such a great program (Geolab) has stagnated. I cannot recommend Geolab to anyone anymore because of this.

Some might think that Move3 is pricey, but for the type of work I do I use LSQ everyday, it is a critical component to my work. I am willing to pay for an excellent program that is being maintained and improved all the time.

I would appreciate hearing from anyone who uses Move3 a lot.

Related Question for the Resident Geodesists

> I also have wondered about this. Based on my google searches, it appears to me that Move3 is the only group using the term "Delft school of adjustment theory", as if it is a distinct method of least squares adjustment.
>
> Donnie

From what I’ve looked at I'm not so sure it's a distinct method of least squares adjustment; rather placing emphasis, working backwards from the residuals, on testing the mathematical and stochastic models assumed in the original adjustment.

>
> Some might think that Move3 is pricey, but for the type of work I do I use LSQ everyday, it is a critical component to my work. I am willing to pay for an excellent program that is being maintained and improved all the time.
>
> I would appreciate hearing from anyone who uses Move3 a lot.

If I recall correctly there was some kind of Trade in for your Star*Net key. Encouraging you to upgrade to Move3.

Ralph

After looking this over I don't see where Move3 does anything more
than the new version of Starnet. The upgrade/tradein from Starnet Pro
to Move3 is $500 and I can't see any gain for the money.

I may be biased with my longstanding use of Starnet 😉 but I really
don't see whats so great for $2500 and I think the GUI in Starnet is better.

I almost always work on a job over an Ortho but thats in the CAD program
after everything is adjusted. I just don't see the advantage of doing
the adjustment over an Orhto.

Full disclosure.. I do use MSCad & Fieldgenius so I was glad when MicroSurvey
took over the program from Ron and dressed up the GUI.

> After looking this over I don't see where Move3 does anything more
> than the new version of Starnet. The upgrade/tradein from Starnet Pro
> to Move3 is $500 and I can't see any gain for the money.
>
The OP stated "Move3 (based on its video presentation) reads native files.". If that's the case then there is a significant gain. No work arounds or jimmy rigs, just a straight up import.

Ralph

Related Question for the Resident Geodesists

James, Donnie

There is nothing mysterious about the "Delft School". Least squares is least squares, That is when it comes to coordinate computation in regular (1D, 2D or 3D) survey networks. The specialty of the Delft School lies in the quality control. A little theory, if you don't mind.

As you know we must distinguish between:
* PRECISION => the propagation of measurement noise (not errors) into the coordinates. Precision is something you cannot do much about. It is a result of your instruments' specifications and, of course, you should take your observations under good conditions.
* TESTING => procedures to identify possible errors in your observations. In general one tries to avoid "false positives" or "false negatives" as much as possible. False positives occur if an observation is rejected while in reality there is nothing wrong with it. False negatives occur if an observation containing a real error is accepted by the testing procedure. Bottom line: testing results have a statistical character. You may only say "this observation is correct" when you add to it "with a 95% probability". Note that tests referring to individual observations are one-dimensional (data snooping). In addition two-dimensional tests (e.g. on state plane coordinate pairs) or three dimensional teste (e.g. on 3-D coordinates or GPS baselines) or multi-dimensional tests (e.g. on the network as a whole, typically Chi square or normalize Chi-sqaure) can be performed.
*RELIABILITY => deals with two interesting questions
Q1: having performed the testing procedures with care, I now know that the result cannot be interpreted as absolute truth. Some observations may contain errors which I have not identified. My common sense says, if the error is really big I will definitely detect it, but how big should it be at least, for my test to say 'Rejected"? This is called "internal reliability" and is expressed by the so called marginal or critical detectable error. Its value can be computed for each individual observation. Note that here you can a lot to improve the quality of your work, typically by applying a better network design and/or by a better precision of your measurements.
Q2. The second question follows from the previous. Let us assume that an error is made in an observation with a value just below the critical (or marginally detectable) value. This means, I haven't been able to detect it by testing. I may then ask myself: what influence has this (assumed) error on the coordinates? This is called external reliability and it can be expressed in various ways. For instance MOVE3 gives the maximum coordinate shift for each critical (or marginally detectable) error.

Now back to the Delft School: As we have seen, precision, reliability and testing are interrelated. This relation should be consistent and should not in any way depend on whether you look at one-, two, -three or multi-dimensional tests. You may also translate this requirement into the following formulation: the reliability in connection with one-, two-, three-, or multidimensional tests is always the same.
That's all of the typicality of the Delft School in a nutshell. In effect it means that your QC results are consistent all the way. The founding fathers of the Deft School are Prof Tienstra, Prof Baarda and indeed Prof Teunissen (whose class mate I was back in the eighties).

Gijs Boekelo

Related Question for the Resident Geodesists

A very reliable friend who once participated in development of one of the routines the Move3 still employs says that there are big reasons to upgrade to Move3.

Howdy,

I did notice that the manual for the software is available for download at: http://downloads.move3.com/Manual_MOVE3.pdf. I do not know whether this is the most recent. It is nice they made it available for review.

Reading through it I had some questions. I did note that the "Literature List" did not include the works of Teunissen. I also noticed mention that it did not necessarily use correlations in GPS vectors. Interesting that this information is not used even when available. Their treatment of the issue of WGS 84 especially now that it differs so greatly from NAD83 is vague. I did not see mention of an ability to transform vectors between datums (WGS84 and ITRF) and epochs.

Perhaps a more careful review will show some novel aspects of the program. The manual does provide some good information for those interested in least squares and the analysis if results.

Like most users, once familiar with a package it is difficult to justify learning a new one. I resist moving to more automated data snooping.

Cheers,

DMM

Here's a youtube video describing some of the move 3 features.

[flash width=560 height=315] http://www.youtube.com/v/u1ZV4izFJvU?version=3&hl=en_US [/flash]

Ralph

As I recall, Star*Net adjusts on a grid, then projects onto an ellipsoid if specified by the project options. Move3 (and Trimble and some others) perform the adjustment directly on the reference ellipsoid. The difference is reportedly insignificant for projects less than a few hundred kilometers in any dimension.

I use the Leica OEM version of Move3 inside LGO, it seems to be decent, BUT you don't get the full version.

If Move3 doesn't suit you, I would take a look at Columbus, from what I have seen very good support and frequent updates, reasonably priced too.

SHG

Jim

In terms of computed coordinates your comment (on the grid vs on the ellipsoid) holds, I guess. For the quality control I prefer to work in a model (which by definition is a simplification of the real world) that resembles the real world as close as possible. In a 3D ellipsoidal model observations can almost always be entered in their 'raw' format, hence there is no need to project, correct, reduce or otherwise compromise them (a rare exception is for instance the correction due a deflected vertical).

So the quality of what you try to descibe and try to improve has a one-to-one relation with the original 'undeformed' observations. That, in my view, is an important advantage of working in a robust 3D model. In addition, the possibility to process hybrid networks (partly 3D, partly 2D, may be even partly 1D) without compromising the observations in any way is a positive side effect of the ellipsoidal approach. Finally, this approach puts projections and projection systems on the place where they belong, i.e. outside the least squares process.

Gijs Boekelo