Starting a new map, a how to guide

[martinvk]

Many a time I've read questions about how to start a map so since I've started a new one, I thought I would share some of what I do. It is by no means the only way to work but if you haven't developed your own method yet, I offer it as something to try.

After working on my Betuweroute for about one year, most of the track is placed. Now comes the tedious part, adding scenery and landscaping.

In order to keep my interest in Trainz, I decided to start a new map. It will be smaller than the 160km Betuweroute but not so small that there is no room to maneuver.

This time I'm doing the Arnhem-Tiel route which just celebrated its 125th anniversary. This would be your first task, deciding on a route. If it will be fictive, then you can do anything you want as long as it looks plausible. If it will be real, then you need to find out as much as you can before starting. It would be a shame to get halfway along only to discover that there is too much missing information. Not that you need to know everything, you can always use a bit of poetic license to fill in the gaps.

For those not into DEM data, one of the great resources for current routes is Google Earth. With it you can see the locations of the major landmarks, major buildings and the type ground cover. The only real drawback is elevation. In my case, the route is relatively flat so that is not an issue.

The first thing I do is put placemarks in GE at all major points of interest like track junctions, start and end of curves, station platforms, bridges etc. I then export the data and import it into Excel to create a reference list. I only keep the name of each placemark as well as its latitude and longitude value. Be sure to select the lat and long of the placemark and not that of the viewport in which it was located.

Now in your first baseboard place the World origin and configure it to a value close to one of your placemarks. Not too close so it doesn't get in the way of your route but not so far that you have to place many baseboards to reach the start of your route. Using my Lat/Long Reader, I place reference marks at each placemark I made. Since the Reader is a Surveyor only item, I can place as many as I need without having them show in Driver.

Once all reference markers have been placed in my map, I'm ready to place track and major objects. For each interesting spot I go back to GE and zoom in on it so that I have at least two placemarks visible and then measure the horizontal distance of the view. I like to have no more than 1000 meters, preferably closer to 500 m. I do a screen capture and paste it into my photo editor.

First step is to crop the image to only show the ground, removing the window border and then resize the resulting image to a horizontal size of 1024 pixels while letting the vertical change proportionally. Next I change the vertical size to 1024 by adding blank space to the bottom and not by stretching the image. Save as an uncompressed TGA file.

In GMax I create a square plane object the same size as the measurement I took in GE. Next I rotate it 90 degrees left. Then I map the TGA image I saved. Finally save, export, create a config file and voila, a new guide. Once they are all done, I have a handy reference guides for each interesting point in my map.

Now it's your turn. :wave: Me, I'm busy placing track and other objects to populate my map.

 

[nicky9499]

gee that was REALLY helpful. I'm just about starting a very short branchline which me and my friends go to play after school sometimes. I know the route quite well, but the info here really gave me a good start!

 

[geophil]

I would use geographic coordinates (i.e. lat/long) in Surveyor with utmost care.

As we all know, the Earth is not flat, but Trainz landscape is. Hence, we have a task to transform the curvature of the Earth to a plane. This is called a projection. To add to the problem, such a projection always comes with a certain error. People have invented quite a few of those projections in order to minimise error for a specific purpose.

For all forms of coordinates, including lat/long, in addition to the projection, we will be dealing with a mathematical model these coordinates refer to. The common models for the Earth are either some sphere or some ellipsoid with varying radii and for the ellipsoid some varying displacement. This is called a geodetic datum.

For Google Earth we know they are using an ellipsoid with the geodetic datum WGS84.

What does Surveyor use? Does anyone know? When lat/long coordinates were invented in Surveyor, what was the purpose? Map projection? I am not aware of any indications in that direction. Or were those coordinates a by-product of a helper function for sunrise and sunset? The latter could mean that we have a very simple model here - with some consequences.

The little problem arising from this uncertainty is that we can't be sure of the coordinate transformation we have to apply. We know there is some built-in projection and we can access some sort of lat/long in Surveyor. How are these related to WGS84?

As we don't seem to know the answer, couldn't we simply do without such "details"?

Of course we can, as long as we do our own thing, i.e. invent our own transformation. We could say we transform WGS84 coordinates to Trainz lat/long by not changing them at all. And we will not look too closely when measuring distances in Surveyor and compare with the real world.

And we will not incorporate a DEM. The DEM would bring trouble. A DEM will be in one of the well know map projections, not the unknown one of Surveyor. None of the known tools supports Surveyor lat/long, not even HOG.

What kind of error will we get if ignoring all this math stuff? For well known projections the error can run up to several hundred metres. For an unknown projection, the expected error is, well, eh, unknown. :wave:

geophil

 

[nicky9499]

another tip;

Whel building your route, do not do everything together.
This means, for example, I lay my track until location A. I find that there is a nice hill suitable for a small town. So I click the object tab, and start placing buildings, roads, signs and whatever.

Building your route this way will make completion seem impossible given the length of your route. Rather, lay all the track first, include stations, industries and signals. This is hard to do at first, like say you see an object you fancy, and fear that you would forget it's name later on, so you place the object. Then you see another interesting name, and the cycle goes on.

By finishing the line (or a large part of it), you know where the end is and it won't seem so long.

 

[martinvk]

another tip;

Whel building your route, do not do everything together.
...like say you see an object you fancy, and fear that you would forget it's name later on, so you place the object. ....

This is one of the reasons I created those little map pins and map flags. I place then where future objects will be. That way I can continue placing track without having to remember where other things should go when I get around to placing them.

 

[martinvk]

I would use geographic coordinates (i.e. lat/long) in Surveyor with utmost care.

As we all know, the Earth is not flat, but Trainz landscape is. ...
geophil

What, you are saying it isn't flat :hehe:

Seriously, I don't worry too much about the difference, especially over the relatively short distances I model. I think the error in my measuring is greater than the true error between Trainz and the Earth.

Now if someone were to do the Trans Siberian or the Trans Canada, then they would have to take the difference into account.

 

[jjqburch]

Geophil, just out of interest, do you know what geodetic datum that maps.live.com website uses? Or flashearth.com? I was just wondering because i want it to match my dem.

Thanks for you help, i have found your many posts a great source of advice.

JB

 

[geophil]

Well, let me underline my scepticism with a few figures now.

I compared the well known UTM projection with the unknown Surveyor projection to determine the error.

Internally, Trainz uses a flat landscape with a Cartesian coordinate system. UTM and all the other large scale map projections do the same. The common way for prototypical routes in Trainz is to associate the UTM projection (or a similar one) with the Trainz internal Cartesian system. Keep in mind that the lat/long system in Surveyor is add-on functionality and not an integral part of the Trainz terrain model.

For UTM we have exact formulas how to convert WGS84 geographic coordinates as obtained from Google Earth to the Cartesian coordinate space we can use in Trainz. UTM coordinates are in metres.

My example is located in the Lehigh Valley, Pennsylvania, USA (because that was the most recent route with reasonable size I have played around with).

Here is my test:

I placed the Trainz World Origin object at -76° 16' 39.30" 41° 45' 45.12" (decimal minutes converted into integral minutes and decimal seconds). UTM coordinates are 393800 4624199.

A first measurement takes place at the nearest UTM 1000m grid line intersection to the SW. Desired UTM coordinates are 393000 4624000. With the Trainz landscape being a Cartesian one I can simply count the 10 m lines to get there. To further simplify life I have highlighted 1000m grid lines with appropriate ground textures. I put a lat/long reader object at the grid line intersection in question and obtain: -76° 17' 13.98" 41° 45' 38.94". These are the lat/long coordinates Surveyor provides for this point. Let's convert to UTM using the official formulas: 392996 4624021. That's a difference of 21m and we are only a few hundred metres from the origin.

Next point at 383000 4628000, 10km to the West and 4 km to the North, place a lat/long reader, read the values and convert to UTM proper: 383050 4628167, an error of 174m.

Next point at 373000 4638000, read lat/long reader coordinates and convert to UTM again: 373221 4638312, error 382 m

Last point at 367000 4652000 which is 38.6 km away from the origin. Convert lat/long reader to UTM and obtain 367488 4652386, or an error of 622m.

What do we learn: The error increases with the distance to the World Origin. That's no good. It prevents us from building larger modules. It also prevents us from merging modules, built with this approach. In my earlier posting I already mentioned that this approach also would not let us use a DEM and the error detected explains why.

Therefore I recommend not to rely on lat/long coordinates as obtained by Surveyor. The model of the Earth behind it and the projection is over-simplistic.

This has nothing to do with the lat/long reader object which is simply retrieving the coordinates from Surveryor.

If a route builder wants to associate Google Earth placemarks with coordinates in Trainz and does not want to employ any of the tools, then I suggest to take the Cartesian coordinates instead which you can switch on in the trainzoptions file (-loc) and convert to UTM with some external program or website (I have been told there are a few around).


@jjqburch:

Google Maps, Microsoft Visual Earth (maps.live.com), Yahoo Maps and Ask.com are all based on a common standard spherical Mercator projection (not transverse, not ellipsoidal) with re-projected WGS84 coordinates. (Flashearth is the same as it only reproduces images of the aforementioned.)

geophil

 

[martinvk]

...
I compared the well known UTM projection with the unknown Surveyor projection to determine the error....
geophil

Well that is certainly an error to take into consideration. On an overall point of view, it will certainly mess with an accurate model. However on a smaller scale, I think the error can be managed. Since it appears that the error is linear with distance (1.6%) , it could be considered as a form of compression the way that physical modelers squeeze their layouts into confined basements.

To place local details around a specific set of lat/long coordinates, I find the Surveyor system good enough. The fact that they are all 1.6% too far from the origin doesn't bother me too much.

One point when placing a guide point. The Lat/Long reader shows the values to 3 decimal places but that still results in an on the ground error of 1.8m for the latitude and 1.4m for the longitude location, that being the distance the Reader can move before the numbers change. Therefore the Reader could be out of position by about 2.28 m, measured on the diagonal. So even if the image you map to the guide is very accurate and the guide well scaled, remember that things could still be displaced somewhat.

Before we get to worried about that, remember also that the basic grid is 10m so fine details tend to get blurred anyway.

Thanks to Geophil for the lesson in projection systems and coordinates. Since Trainz is simulator, I guess we should also simulate some of the accuracy.

I'm reminded of the story about the dancing bear. We shouldn't be amazed how well it dances, we should be amazed that it dances at all.

From an outsider's point of view, that Auran was able to add a lat/log system to the original game that comes within 1.6% is pretty amazing, all things considered. I remember when the only way to locate objects was using the elastic rulers. I ended up with so many in my map that Trainz just about froze.