Needs fixing in V 1.3 (James Ross) Coordinates differ more than 1 km
#1
Posted 18 January 2017 - 08:42 PM
James, What ever fix you have? It might be time to consider a good hard look at fixing this issue to the best of your ability for all who start ORTS routes in the feature. Especially commecial offerings.
Quote
"The OR code that converts to lat/long has been known to be inaccurate for years. I wrote better code in one of my personal projects (which is public and licensed compatibly) but nobody (including me) has bothered to move it/use it, because it's an absolutely pain to test."
The inventor is often the best candidate to fix things he has created.
To quote Lindsay.
"Sorry to bother you James, if the existing code is inaccuarte enough that its essentially unusable (as it appears to be) what would be wrong with at least trying the new code. A problem I am finding is the position as reported in OR differs not only from MSTS but from Goku's route editor as well. It would be LOVELY if all these agreed it would help much as it would enable Goku's editor to be much more used. I believe its CRITICAL for new content to be able to be done for OR, for instance in Australia all the old MSTS content developers have left MSTS for other trainsims that do have decent content developing tools.
An obvious way of testing is to have two independent systems side by side viewing the same route one on MSTS the other on OR, this is something I CAN do."
I personally feel we need to correct this error before continuing with other aspects like creating an official route editor suite. I am also disappointed in Goku not wanting to use other available DEM methods in his TSRE5. The European models for Dem are mostly fine but again Japan comes out horribly skewed.
Robert
#2
Posted 19 January 2017 - 01:30 PM
SP 0-6-0, on 18 January 2017 - 08:42 PM, said:
James, What ever fix you have? It might be time to consider a good hard look at fixing this issue to the best of your ability for all who start ORTS routes in the feature. Especially commecial offerings.
I think you might have confused two different issues.
Firstly, the coordinates in OR are different from MSTS (and incorrect). This is what I've written about in your quotes and it's on my list, but I have very little time for coding in OR with all the other bits I have to keep track of.
And secondly, the projection used by MSTS and OR is pretty bad for some regions of the world, including Japan. This is not something OR can "fix" for MSTS content, but we will not be using the same projection in any new formats/editors we build.
#3
Posted 19 January 2017 - 05:14 PM
It's just a really challenging effort and it seems to be made more difficult by issues with both simulators.
Robert
#4
Posted 19 January 2017 - 11:05 PM
SP 0-6-0, on 19 January 2017 - 05:14 PM, said:
It's just a really challenging effort and it seems to be made more difficult by issues with both simulators.
Robert
The problem I think you will found is basicly casued by the earths shape and that MSTS and OR treat the world is if it is flat (this is the simplest approach) and its therefore impossible to really display the terrain accurately. The conversion from the earths real shape to a flat plain in hamdle by map projections. There is NO perfect map projection, the one MSTS and OR uses is called nterrupted Goode homolosine projection. Its used becsue the mathematics fro the conversion has been widely published and the conversion is relatively simple.
Below is a link to an image of this projection..................
https://upload.wikim...ojection_SW.jpg
Notice in the area of Japan the lat and long squares are very skeyed, that is the shape the world is portrayed at in the sim, Note also the ONLY place on earth that will be portrayed properly is Central Africa.
In future there is two paths, the first is to chose a better projection for such a sim as this, a good choice is Transverse Mercator, this will be the simplest path for OpenRails as it basicly means the only step effected is the actual terrain creation. Both OR and MSTS already display flat tiled terrain (Note 1). The second method is to actually display the world as its true shape, this is the way most flight sims do it. This though is not really feasible in OR as it would mean completely rewriting terrain disply from the ground up and its NO WAY backward compatible with MSTS.
Note 1: The conversion from Geodesic co-ordinates (lat and Long) to Transverse Mercator produces flat tiled terrain.
Lindsay
#5
Posted 19 January 2017 - 11:29 PM
Lindsay
#6
Posted 20 January 2017 - 01:51 PM
Robert
#7
Posted 20 January 2017 - 02:35 PM
All map projections are mathematical compromises. Homolosine is a compromise in that it uses sinusoidal projection for latitudes up to 40° with the homolographic for areas poleward of these latitudes. That is one reason why mapmakers have developed so many different types of projections, each is useful for different mapmaking needs.
An example from the book mentioned above: "Transverse Mercator is used where the north-south direction is greater than the east-west direction. Used as the base for the US Geological Survey's 1:250,000 scale series..."
There is no perfect projection, the earth is not a true sphere >>> https://youtu.be/lzDGvyiJ0rc
and..even if it was there would still be a need for all the different projections.
#8
Posted 20 January 2017 - 03:22 PM
R H Steele, on 20 January 2017 - 02:35 PM, said:
There is no perfect projection, the earth is not a true sphere >>> https://youtu.be/lzDGvyiJ0rc
and..even if it was there would still be a need for all the different projections.
Very true, and I am very glad we only ever model small (relative to the overall size of the earth) sections of it! Transverse Mercator is particularly good because it allows the "center" of the map to be chosen, thus providing the best accuracy and least distortion to whatever area you want.
#9
Posted 21 January 2017 - 12:42 AM
SP 0-6-0, on 20 January 2017 - 01:51 PM, said:
Robert
Its a complex subject, with almost no simply introductory texts, it took me something like 12 months of study (note 1) to get a good understanding of all the principles involved. There are two basic issues, in order to make a map, a (usually) flat representation one has distort the earths surface which is of course curved in 3 dimensions. The second point that complicates the procedure is that the earth is NOT a true sphere, its flattened at the poles and the equator is NOT an even radius. The mathematics to sort this lot out accurately are enormously complex.
The reason why tranverse Mercator can successfully give a good result is the earth is VERY large compared to the amount or area any map (or route) can cover, therefore the distortions are very controled and in fact can be easily specified.
Note 1: With many confusing moments, I DID though manage to sort it all out, well mostly any way.
Lindsay
#10
Posted 21 January 2017 - 12:53 AM
Lindsay