Some of us like to do fictitious routes, and that is fine, but I prefer to simulate, as closely as possible, the "real thing". So I thought I would explain how I do this in the case of those routes that I put in the "unfinished" category such as the Anchorage to Fairbanks part of the Alaska RR. This is around 350 miles or so and I guess in the real world it is two or more crew districts or divisions but I am doing it with no breaks although the villages along the way will have their house tracks etc. in place. Yes, by fictitious I am referring to ones like Eric mentioned. And lastly, I do not want to criticize anyone's own method, it might be better than mine. I am just explaining what I do and that may interest newbies more than experienced creators.
I used to use UsaPhotomaps in the very early days and it was a good product, freeware, and easy to use but Microsoft abandoned the database from which it worked and about the same time Google came out with Google Earth and later made the Pro version free. So that is my choice for source material. To prepare the markers etc. for a real route you need to have Google Earth and if possible a track chart of the line. Fortunately, I have been able, at modest expense, to acquire CDs of those charts that I did not already have by virtue of working for Southern and CSX.
I won't go into the nitty gritty of Google Earth because it has very good instructions on how to use it. I will advise that while you are digitizing, frequently tap the "R" key as this will realign the image so that you are not looking sidewise and thus encountering parallax errors. Usually, I will start at one end or the other and work toward the distant end. In the case of this 2nd Alaska RR effort, I already had the Anchorage terminal from the first route I did a while back. So I created the new route with MSTS/RGE and while in MSTS/RE I added one piece of track and one road, buried deep in the terrain. This estabished the .tdb and .rdb database files, otherwise your digitized line may shift several meters one way or another. Another issue is that "usually" the most recent aerial imagery is the best to use but even with that, there will be issues where images are spliced together and I guess those who create GE did not expect folks like us to want to precisely define a railroad or road. And then images different in date from what seems best shift the path many meters off, I have seen as much shift was 20 meters, i.e. 65 feet.
In digitizing, I want to put each point within the gauge lines of the track, so it is best if the image has a sharp view of the track and many times I have seen images so sharp you could count the ties. Other times it is just a blur and in that case, I just select the center of the blur. It is impossible to digitize all of a curve, but it is desirable to use many more points on a curve than on a tangent. In surveying a curve, 100 ft. (30.48 meter) chords would be measured with a tape and transit, so points 100 to 200 feet apart on a curve are okay. Sharper the curve, the more frequent the points. However, unless you use a lot of dynamic track you will not have every curve exactly the right radius. On a mainline, it would be unusual to find any curve sharper than perhaps 8 degrees, or 218.25 meters radius. On the Alaska RR, there are a few curves to 12 degrees (145.5 meters) and on the Blue Ridge sub. of CSX there is one 14 degree curve. The magic constant here is 1746. This is a constant that is derived from surveying and it is good (in the real world) for curves up to about 8 degrees. While more precision is needed above that in the real world, I continue to use it for any curve. Refer to any book on surveying to find out where this constant comes from. It may also be in feet, approx. 5,729.7. My book, from the 1957 class I took at Vanderbilt is "Elementary Surveying, Rayner & Schmidt, D. Van Nostrand Co., copyright 1937, 3rd edition, 1955, pages 169---.) Another constant that all of us need to burn in our synapses is 3.28, i.e. 3.28 feet per meter, which can be found in many handbooks of constants. No matter how good or accurate your digitizing is, there is one thing that will defeat us all. Railroad (or highway) curves are not constant. The track chart says it is an 8 degree curve. If, like me, you take pieces of 8 degree curve from Mtracks and place them they will not fit the digitized curve! You just got bit by the Spirals. Aerial imagery may be too course to show it, but it does include the spirals and if you carefully digitize the route and then try to match it, using our fixed curve pieces, they just don't fit. So I put pieces of lesser degree on each end of a curve if I can actually spot the spirals, otherwise I just pick the curve piece that fits best. And of course we don't have every possible curve value unless we use dynamics (which I did in my early KP & BR subs. of CSX) but usually, just picking the closest value of radius is okay. I will use dynamic if I have a curve with multiple tracks and perhaps there are 4, but Mtracks or Kuju or Xtracks only supplies a two track set.
Coming out of a curve into a long tangent (straight) section (one on the ARR is several miles long) I will use either the 1/2 degree turn, or 1/4 degree turn (yes there are such in Xtracks) to align the tangent as well as I can. For measurement purposes I will string together several 500M pieces but in the end it is usually necessary to put in correction curves, i.e. zig-zags and if so, use the least amount of curve and the biggest radius you can find. Then remove any 500 meter pieces that cross tile boundaries and replace with shorter pieces. It has been recommended that one use 100M pieces across a boundary.
Besides digitizing the track or roads, we should also determine the milepost locations and include the coordinates of these in the marker file. To keep the marker file reasonable in size, I may only include in it, besides the points for the track, the location of the points of a switch (usually in the aerial you can see either the switch stand, or the switch motor & this is close enough to the points) and the direction & size if I can determine it. Manual turnouts to a siding or industry track will usually be #10 size, while power turnouts to a siding or power crossovers will usually be #15 or #20 size. Sometimes these values for control points will be in the track chart, sometimes it is a educated guess, and you could measure the spread of the track beyond the point of frog, if the image is sharp enough and come close. If for exampe, the spread is 1 foot at a point 10 feet past the point of the frog, it is a #10 turnout.
And then we have vertical curves and gradients to deal with. Remember that the %of grade shown on a track chart is the compensated grade, i.e. it is adjusted for curves. Curves produce friction on a moving train which has the effect of increasinig the grade. We don't have that in TrainSim so what I do is take the difference in height between two points on the track chart - some charts show the elevation above sea level, otherwise measure it from GE - and measure on GE the distance between the points, marking points close together on curves. If you come up with a grade difference o 50 feet (or 15.2 meters) and a distance of 12,000 feet (or 3658.5 meters) then 50/12000 = .0042, i.e. .42% grade. Not perfect but real close. Producing the vertical curve is a matter of preference. Some of us have mentioned using 5 meter increments and some 10 meter increments. For a gradual change where I have enough linear space, I use a change of .1% per 10 meters, otherwise .1% per 5 meters. This is subjective and a individual choice. Caution, if you take a measurement of elevation from Google Earth, also take a measurement from more or less the same location in the route under construction. Often there is a large difference. If that is the case then remember that outside of extreme mountain conditions, fills higher than 35 feet or maybe 10 meters would be unusual. Fills of more than 100 feet might be found in mountain terrain.
And then the railroad may not get it right either. North of Anchorage, just northeast of Elmendorf AFB, the railroad used to go close to the base and then wind around a bit and join the main line northeast of the airbase. You can see this on GE images prior to about 2005, especially in the 9/96 image in which there is no evidence of the current line except a fine line through the trees which looks to me like a cleared line for survey purposes. Then look at the current 2024 image and you can still see faintly where the old line ran, and now there is the new relocation. But between 121.8 and 122.8 track charts show a pair of 2 degree curves, but when very carefully digitized, the south most is actually 2-2 degree curves with a short tangent and the 2nd curve is a little sharper than 2 degrees. I suspect a drafting error in the track chart.
And then there is the issue of spirals. Almost all curves so far on the Anchorage-Fairbanks line are different from the value seen on a track chart. This is most likely because spirals are not identified on track charts because for a particular railroad these would be standard. Some predecessors to NS & CSX used a 10-chord spiral. In it's simplest terms this means that the degree of curve was divided into 10 parts on each end, and if the curve was a 10 degree curve for example, it would then be Tangent-1 degree - 2 degrees - 3 degrees - 4-5-6-7-8-9-10 degrees. The length of each part would depend on the available space. So what I do in the case of a 2 degree curve, common on the line under construction, is I come into the curve on a tangent piece aligned as well as possible to the digitized points. Then I start with a 30' piece, 1 degree piece, 1 degree 30' piece and then 2 degrees, adjusting the lengths of each part until the curve aligns well with the digitized points and then I repeat this in reverse until I can lead out of the curve on a tangent piece aligned to the digitized points. For this alignment, I will use 1 or more 500 meter pieces and once I am satisfied, I will removed any long pieces that cross tile boundaries and replace them with shorter pieces, using a 100 meter or shorter piece across the tile boundary.
Have I created enough confusion, consternation, weeping, wailing, & gnashing of teeth. I'll check for comments and any that seem to require more explantion I will add above this line.
Jerry Sullivan, P.E. (retired) Although my degrees were in Electrical Engr., 95% of my career was Civil Engr., railroad specifically.
I used to use UsaPhotomaps in the very early days and it was a good product, freeware, and easy to use but Microsoft abandoned the database from which it worked and about the same time Google came out with Google Earth and later made the Pro version free. So that is my choice for source material. To prepare the markers etc. for a real route you need to have Google Earth and if possible a track chart of the line. Fortunately, I have been able, at modest expense, to acquire CDs of those charts that I did not already have by virtue of working for Southern and CSX.
I won't go into the nitty gritty of Google Earth because it has very good instructions on how to use it. I will advise that while you are digitizing, frequently tap the "R" key as this will realign the image so that you are not looking sidewise and thus encountering parallax errors. Usually, I will start at one end or the other and work toward the distant end. In the case of this 2nd Alaska RR effort, I already had the Anchorage terminal from the first route I did a while back. So I created the new route with MSTS/RGE and while in MSTS/RE I added one piece of track and one road, buried deep in the terrain. This estabished the .tdb and .rdb database files, otherwise your digitized line may shift several meters one way or another. Another issue is that "usually" the most recent aerial imagery is the best to use but even with that, there will be issues where images are spliced together and I guess those who create GE did not expect folks like us to want to precisely define a railroad or road. And then images different in date from what seems best shift the path many meters off, I have seen as much shift was 20 meters, i.e. 65 feet.
In digitizing, I want to put each point within the gauge lines of the track, so it is best if the image has a sharp view of the track and many times I have seen images so sharp you could count the ties. Other times it is just a blur and in that case, I just select the center of the blur. It is impossible to digitize all of a curve, but it is desirable to use many more points on a curve than on a tangent. In surveying a curve, 100 ft. (30.48 meter) chords would be measured with a tape and transit, so points 100 to 200 feet apart on a curve are okay. Sharper the curve, the more frequent the points. However, unless you use a lot of dynamic track you will not have every curve exactly the right radius. On a mainline, it would be unusual to find any curve sharper than perhaps 8 degrees, or 218.25 meters radius. On the Alaska RR, there are a few curves to 12 degrees (145.5 meters) and on the Blue Ridge sub. of CSX there is one 14 degree curve. The magic constant here is 1746. This is a constant that is derived from surveying and it is good (in the real world) for curves up to about 8 degrees. While more precision is needed above that in the real world, I continue to use it for any curve. Refer to any book on surveying to find out where this constant comes from. It may also be in feet, approx. 5,729.7. My book, from the 1957 class I took at Vanderbilt is "Elementary Surveying, Rayner & Schmidt, D. Van Nostrand Co., copyright 1937, 3rd edition, 1955, pages 169---.) Another constant that all of us need to burn in our synapses is 3.28, i.e. 3.28 feet per meter, which can be found in many handbooks of constants. No matter how good or accurate your digitizing is, there is one thing that will defeat us all. Railroad (or highway) curves are not constant. The track chart says it is an 8 degree curve. If, like me, you take pieces of 8 degree curve from Mtracks and place them they will not fit the digitized curve! You just got bit by the Spirals. Aerial imagery may be too course to show it, but it does include the spirals and if you carefully digitize the route and then try to match it, using our fixed curve pieces, they just don't fit. So I put pieces of lesser degree on each end of a curve if I can actually spot the spirals, otherwise I just pick the curve piece that fits best. And of course we don't have every possible curve value unless we use dynamics (which I did in my early KP & BR subs. of CSX) but usually, just picking the closest value of radius is okay. I will use dynamic if I have a curve with multiple tracks and perhaps there are 4, but Mtracks or Kuju or Xtracks only supplies a two track set.
Coming out of a curve into a long tangent (straight) section (one on the ARR is several miles long) I will use either the 1/2 degree turn, or 1/4 degree turn (yes there are such in Xtracks) to align the tangent as well as I can. For measurement purposes I will string together several 500M pieces but in the end it is usually necessary to put in correction curves, i.e. zig-zags and if so, use the least amount of curve and the biggest radius you can find. Then remove any 500 meter pieces that cross tile boundaries and replace with shorter pieces. It has been recommended that one use 100M pieces across a boundary.
Besides digitizing the track or roads, we should also determine the milepost locations and include the coordinates of these in the marker file. To keep the marker file reasonable in size, I may only include in it, besides the points for the track, the location of the points of a switch (usually in the aerial you can see either the switch stand, or the switch motor & this is close enough to the points) and the direction & size if I can determine it. Manual turnouts to a siding or industry track will usually be #10 size, while power turnouts to a siding or power crossovers will usually be #15 or #20 size. Sometimes these values for control points will be in the track chart, sometimes it is a educated guess, and you could measure the spread of the track beyond the point of frog, if the image is sharp enough and come close. If for exampe, the spread is 1 foot at a point 10 feet past the point of the frog, it is a #10 turnout.
And then we have vertical curves and gradients to deal with. Remember that the %of grade shown on a track chart is the compensated grade, i.e. it is adjusted for curves. Curves produce friction on a moving train which has the effect of increasinig the grade. We don't have that in TrainSim so what I do is take the difference in height between two points on the track chart - some charts show the elevation above sea level, otherwise measure it from GE - and measure on GE the distance between the points, marking points close together on curves. If you come up with a grade difference o 50 feet (or 15.2 meters) and a distance of 12,000 feet (or 3658.5 meters) then 50/12000 = .0042, i.e. .42% grade. Not perfect but real close. Producing the vertical curve is a matter of preference. Some of us have mentioned using 5 meter increments and some 10 meter increments. For a gradual change where I have enough linear space, I use a change of .1% per 10 meters, otherwise .1% per 5 meters. This is subjective and a individual choice. Caution, if you take a measurement of elevation from Google Earth, also take a measurement from more or less the same location in the route under construction. Often there is a large difference. If that is the case then remember that outside of extreme mountain conditions, fills higher than 35 feet or maybe 10 meters would be unusual. Fills of more than 100 feet might be found in mountain terrain.
And then the railroad may not get it right either. North of Anchorage, just northeast of Elmendorf AFB, the railroad used to go close to the base and then wind around a bit and join the main line northeast of the airbase. You can see this on GE images prior to about 2005, especially in the 9/96 image in which there is no evidence of the current line except a fine line through the trees which looks to me like a cleared line for survey purposes. Then look at the current 2024 image and you can still see faintly where the old line ran, and now there is the new relocation. But between 121.8 and 122.8 track charts show a pair of 2 degree curves, but when very carefully digitized, the south most is actually 2-2 degree curves with a short tangent and the 2nd curve is a little sharper than 2 degrees. I suspect a drafting error in the track chart.
And then there is the issue of spirals. Almost all curves so far on the Anchorage-Fairbanks line are different from the value seen on a track chart. This is most likely because spirals are not identified on track charts because for a particular railroad these would be standard. Some predecessors to NS & CSX used a 10-chord spiral. In it's simplest terms this means that the degree of curve was divided into 10 parts on each end, and if the curve was a 10 degree curve for example, it would then be Tangent-1 degree - 2 degrees - 3 degrees - 4-5-6-7-8-9-10 degrees. The length of each part would depend on the available space. So what I do in the case of a 2 degree curve, common on the line under construction, is I come into the curve on a tangent piece aligned as well as possible to the digitized points. Then I start with a 30' piece, 1 degree piece, 1 degree 30' piece and then 2 degrees, adjusting the lengths of each part until the curve aligns well with the digitized points and then I repeat this in reverse until I can lead out of the curve on a tangent piece aligned to the digitized points. For this alignment, I will use 1 or more 500 meter pieces and once I am satisfied, I will removed any long pieces that cross tile boundaries and replace them with shorter pieces, using a 100 meter or shorter piece across the tile boundary.
Have I created enough confusion, consternation, weeping, wailing, & gnashing of teeth. I'll check for comments and any that seem to require more explantion I will add above this line.
Jerry Sullivan, P.E. (retired) Although my degrees were in Electrical Engr., 95% of my career was Civil Engr., railroad specifically.
Comment