Railworks America

Despite the several warnings about staying away from Rail Game 2012's physics, I decided to take a gander to see if even some minor changes could be made to increase realism despite the voodoo magic physics engine that no one understands the workings of (perhaps not even RSC based on some support emails I read).

While searching through the double-stack car's .bin I found the mass of the car and its loading mass, both seemed a bit... off. What I stumbled across was:

Code: Select all

<Mass d:type="sFloat32" d:alt_encoding="0000000000003340" d:precision="string">19</Mass>

and

Code: Select all

<CargoDef>
     <cContainerCargoDef d:id="38250680">
          <MassInKg d:type="sFloat32" d:alt_encoding="0000000000000000" d:precision="string">0</MassInKg>

According to the developer's wiki car mass is in tons and the cargo weight is clearly labeled in kg's. This means that the empty well car running down the rails weighs in at 38,000lbs despite being listed by Husky (just for reference) at 50,500 lbs bare table and of course the cargo weight of 0 kg means that whether or not its loaded makes no difference as you add 0kg to the weight of the car even though it has a load capacity of 169,500 lbs (again as per Husky).

On the plus side they do actually use the proper physics equations, just completely wrong values to put into the equations. For instance, in the same .bin for the double stack car it lists the rolling resistance coefficient as 0.0015 when a University of Illinois study found the coefficient of rolling resistance to be .0010 making the cars in game ~50% harder to roll once they are moving. It seems that improper values were inputted in the beginning back in the Kuju days and rather than fix the problem, they simply adjusted all the values on new rolling stock as it was produced so that everything works well and plays nice together. I think that we could actually turn this into a simulator, but it will take some serious work.

Any thoughts?

I'd be curious to know how many things have incorrect variables (are the engines off too?) and how much of an improvement in terms of realism we'd see if it were corrected. That also makes we wonder... could we make two "versions" of say, a boxcar, one with empty weight and one with loaded weight? Hmm...

We have that now with a lot of cars. I am not sure about box cars......something more to fiddle with tonight.

Testing update for everyone:

I adjusted the mass and cargo weights in the stack cars to their proper weights to some fantastic results. For testing I setup a Free-Roam scenario on Sherman Hill using 4 ES44AC's and 80 stack cars and ran them out of Laramie and up the 0.8% grade on track 2. Using the RSC weights I ran up the 0.8% grade shuttling back and forth between Run-1 and Run-2 to maintain the full 45mph track speed for both the bare table and loaded trains. Not exactly very realistic.

After adjusting to the proper weights I ran the same scenario again and climbing the grade with a bare table train used the same shuttling back and forth between Run-1 and Run-2, but this time with more time spent in Run-2 and less time reducing speed down in Run-1. Loading the train produced good results. Previously I would leave Laramie in Run-2 and accelerate all the way to 30mph before the junction onto track 2. With the properly weighted and fully loaded train, it took Run-4 just to make it to 30mph before the turnout. Then once the train was on the grade and settled out, it took a consistent Run-4 to maintain the 45mph track speed which was a great result!

You are probably now saying but dfcfu342, to run an 80 car loaded stack train up a 0.8% grade it would surely take more than half throttle to maintain a speed as high as 45mph! And you would be correct which leads me to my final point for this post: Train Game 2013 uses a linear throttle where real locomotives us a throttle based on a logarithmic curve. By some rudimentary noggin math, Run-4 on a linear scale of tractive effort would be very close to Run-6 in a logarithmic tractive effort scale.

Tying everything together (beginning to sound like a presentation now isn't it? ) I believe that everyone's issue with the U.S. content feeling overpowered is a two-fold problem. In lower throttle settings too much tractive effort is being put out and most cars seem to be underweight which compound each other.

By fixing the car weights and throttle scaling we could achieve an 80 car stack train with 4 GEVO's running up a 0.8% grade in Run-6. Starting to sound more like a simulator now?

dfcfu342 wrote:Starting to sound more like a simulator now?

Yes, it certainly does!

What about the pre-installed standard scenarios? If you change rolling stock parameters that drastically, chances are you'll break the scenarios as the trains will be much slower and don't meet their pre-programmed objectives anymore.

I for one would love to see more realistic operation and prototypical characteristics, but with TS2013 being more of a game than a simulator we are on our own here I suppose.

I hadn't even considered the scenarios honestly and I have no idea how they are setup I've just played them

This weekend when I get more free time I'm going to finish adjusting all of the values and see what it does to scenarios and then I'm going to dive into rolling friction and air resistance to see if we can get more realistic downhill rolling for the other side of the hill.

Just do the scenarios first to complete them and get the achievement and then change the physics to more realistic ones after that.

Pretty sad that you seem to be figuring things out better then the company that made it. But then again, this is the same company that released Donner Pass in it's awful state. Yes they are redeeming themselves by fixing it. But that's extra time and effort to fix something then getting it right to begin with. It basically cuts down the profit margin because of the extra time (=money) they are dedicating to it. Not to mention another blow to their reputation. At least I got Donner Pass for 50% off, so I'm not as angry as some that paid full price.

ca2kjet wrote:I'd be curious to know how many things have incorrect variables (are the engines off too?) and how much of an improvement in terms of realism we'd see if it were corrected. That also makes we wonder... could we make two "versions" of say, a boxcar, one with empty weight and one with loaded weight? Hmm...

Hi,

There are several 'issues' in the rail vehicle physics that we inherited from Kuju (or whoever designed the original rolling stock). My (highly speculative) theory at this point is that the underlaying physics calculations aren't any worse than they are in MSTS -- if they are any different at all (it's the same company which developed both titles, so maybe they reused their physics calculations?)

Previous experimentation has already shown that there is HUGE potential for much more realistic train performance in RW. There was a very long thread about that here a few months ago, I think it started as a discussion on (engine?) braking settings and soon developed into a highly techincal and fascinating debate on what could be possible. (I don't find it any more, my search didn't turn up the right topic -- does anybody remember its title?)

MT and LD versions of cars are not only possible, but in my opinion crucial if one also wants to get the braking settings correct. The reason here is that the BIN file doesn't use a fixed brake force parameter, but rather a "% of weight" as brake force -- which makes it impossible to create load sensing brakes. So using two BIN files, one for MT and one for LD, each with its appropriate weight and brake force setting, will have a dramatic effect on train performance. Add to that modified charge, apply and release rates on all cars in the consist and the engine too and you're looking at a totally new running experience.

The engines' basic parameters, tractive effort and power, at least on the GP9 and F7s which I've run a lot, generally seem to work okay. At least most of the time I am able to get about the same max. speed and tonnage results as for example Al Krug's gradient/tonnage calculator suggests.

HOWEVER -- all of these changes will affect existing scenarios, quite likely breaking many of them since their AI timing will be changed substantially. This is experimental stuff that will quickly create compatibility issues with 'out of the box' content.

Cheers,
Michael

I'm one of those crazy people who doesn't play scenarios... so bring on the better physics!

That Topic was called "dodgy physics", Michael. But I got nothing from Search, either. Maybe some difficulty from the server switch. Of course, you could start looking thru the General Discussion topics.

EDIT: Found it! viewtopic.php?f=6&t=3737&hilit=Physics Have fun!

I've had issues searching as well. Searching for sd40t-2 results in nothing, even though donner pass has been out a while now and I know some people had tried to change the sounds on them.

While it might break the existing scenarios, I wouldn't worry about it too much. This is the same issue as in MSTS, change the physics and now you can't get that train over the hill. It would require those writing new scenarios to specify whether or not they use the new physics.

Kevo

Realistic physics depend on hard core programming as well (that's what the Dodgy Physics thread reveals, ultimately), not only values of tables to be adjusted (not to mention the need for the correct adjustment of a dozen of parameters for every vehicle). So, IMHO, forget about it with RSC products, they are not meant to it, X-Box generation doesn't want to wait the SH train (in the main SH scenario along the route) to last 1 mile to reach 15mph (currently is only 0.1 mile), nor take 1.5 mile to stop the same train (currently is only 0.3 mile, if not less). Just have fun with the diversity of fleet, routes and the really awesome graphics (the shadows now on TS213 are impressive indeed).

Just as a side note, despite of born as Kuju products, MSTS physics (currently of course, thanks to a lot of modders) are much better than RW physics, but still far from realistic once the improvement is also limited to the hard core programming thingy. The question is to choose or not to choose between a simulator or a game, and is more than clear what the RSC option is (despite of game's name).

Thanks for everyone's input! I know a true-to-life simulation will never be possible without actually tearing into the inner game engine but I'm making do with what I have and making it as realistic as possible with what we have to work with.

Today's latest update is that I finished adjusting all the Sherman Hill cars to their proper weight values and then set about the task of setting them all to a brake-force of 13% of their loaded weight which from what I gathered on the internet is the current standard value for freight cars. I ran the Granite via Wycon scenario and to my surprise it ran without a hitch! No crashes or scenario failures and the AI trains even ran more realistically (They actually accelerated from a stop instead of instantly reaching speed).

I will test a few more scenarios this weekend and if they all run to completion as well then I will start on setting the rolling resistance values to true values and see if we can get trains to actually roll downgrade and require some real driving to keep under control

If everything goes smoothly (fingers crossed) would anyone be willing to test my altered .bin files on their machine to see if they can replicate my results with working scenarios?

I'll test them, I've been looking for almost real physics in the game.