Bug or designed limitation in drivetrain code?

For example. take sunburst and put CVT transmission in it, but have sunburst_engine.jbeam modified so that line 620 is "maxGearRatio":102.7, then try to drive very slowly (use analog throttle, get moving with least amount of throttle first, then try whatever you wish), you see how car rocks up and down, does not want to go.

Most evident after spawning, if one gets car to move it might be less severe, wheel debug shows negative speeds altering with positive speeds, engine debug shows negative and positive torque after each other, while car shakes.


80 meters per hour is speed I would need to get proper gear ratio of tractor crawling, tire radius is over 1 meter, so that means quite big reduction from engine rpm to wheel rpm.

So to be able to simulate anything, crawler transmission of CVT variety would be of course one to be simulated, but perhaps there is precision issue currently, some variable goes out of bounds or is not enough precise, I'm sure developers know exactly what is that issue, just my guesses.

No matter how I try, with gearing set to normal there is none of this issue, so I guess it is some variable in driveline code, but developers will have much better idea of potential problem of course.

edit: Oh and just a heads up if not noticed from other location:
http://www.beamng.com/threads/0-8-0...-thread-read-1-post.33989/page-25#post-521207

 

Not only that, but rather than Using insain reductions in the CVT alone, perhaps to help the simulation out a bit, you can use some of that reduction in the final drive ratio.

 

Sadly issues are much worse with actual tractor wheels.

Not sure if differential ratios will actually provide help, but will test out, I doubt that currently 80 meters per hour at 2000rpm engine speed and over 2 meter tire diameter is possible.

 

Thats a good point. If they are too light you could run into stability issues. Way back in the RoR days, when the wheels were to light, the wheel speed would fluctuate violently from positive to negative velocities. Your issue does sound very similar to this. In RoR, this symptom could be caused by the wheels being too light, having too much spring, or having too much damping. Usually it wasn't a combination as that would just cause and explosion. I do not know how Beam will react to the same issue, but it does sound like you have very similar symptoms.

The biggest cause of this is when you go to increase wheel size, naturally one will end up increasing the ray count of the wheels to compensate for the size difference. This increases the number of nodes by a substantial amount and usually, in turn, ends up creating a light, unstable wheel even though it was perfectly fine before.

 

189.6kg, if I'm getting it right, that is what node debug shows, I guess that is adjusted by node weight in tire file, I'm just helping a bit with this tractor and I just bought BeamNG some week ago, so I hope I understood how to get weight correctly

There is one issue that all 4 are same weight and front wheels should of course have different weight.

I found out that you can't go really far with differential settings, but with transfercase, differential and cvt carefully balanced I can get to 4 km/h with only small issues, good enough for gameplay, but person is always seeking out more, beyond what is possible?
--- Post updated ---

Oh, we cross posted.

Those are interesting points for certain, I know there are 20 rays for wheels, might need increasing and weight balancing, to allow slower speeds, certainly demands more hours of tuning to find out ultimate limits

Also in reality there are 4 ranges in rangebox, I guess we have only two, so limiting to two highest range will allow little cheat here

 

Hmmm... Well... that sounds plenty heavy enough. What is the clutch torque set at? (I don't know if you even need that as a setting when using a CVT in game, but the same symptoms could be caused by way too high clutch torque way back in the day so I figured I would ask)

 

I don't think there is clutch torque involved with CVT, I don't remember seeing one, but there is of course transfer case torque and differential lock torque. I have tested and I get to around 70 000Nm wheel torque, so as 4wd is solid, I have set transfer case torque to 80 000 as well as differential lock torques, as those too should be solidly locked.

However there are other parameters and some probably have such effect which you describe, but I haven't tested enough to be able to figure out what these all do:
"torqueConverter": {
"converterDiameter": 0.5, //probably is equivalent in effect to clutch torque, but not sure.
"converterStiffness":150, //how grabby or loose torque converter is, kind of how much clutch slip allows, but again not sure?
"couplingAVRatio":0.72, //haven't found anything about this, I guess need to search more
"stallTorqueRatio":1.75, //I guess it changes how CVT gets going, need to play with this more, currently I'm thinking it being bit similar to stiffness setting, but maybe comes more to play when launching from standstill?

So indeed there are quite bit of variables in play until we can solidly know there being a bug or what is the limit.

 

Oh, well thats neat. I didn't know it did that. That's pretty awesome!

If that's the case, then that takes out a lot of the possible stability issues right there for any possible torques or spring values... So I guess we are back to step one. Try the normal sunburst CVT settings and then just change the solid gear ratios (through the range box's) to be much more extreme of a reduction... see what that ends up doing for you.

 

What I have found out is that if there is not much of issue in High range, with Low range there still can be lot of issues, 3:1 reduction in range box can be enough if CVT range and differential ratios are set close to limit, not sure if that has any value though as with less reduction in CVT and more in range box it seems to be possible to get lower speeds possible.

It seems I can't use generic profiling app correctly as all I get is gray box without graphs and I don't understand how to add graphs to that, but it is only one that would show gear ratios, which might offer bit more insight in such situations.

 

I got some specs if anyone is enough bored and likes to play around, overall axle ratio 21.098:1, first gear range top speed 7.4km/h with 2200rpm or bit more, specs are not defining maximum reachable rpm, tires 20.5-25 SGG (L2), allowed metric size 550/65 R25 XLD (L3).


What I have found out is that going much more than 6:1 differential ratio is going to cause problems, CVT tolerates quite large range without problems, but I would not go past 10:1 and trying to keep min+max not much more than 4 times reduction at low end over high end so for example max 1 and min 4, range box and transfer case are then bit mystery how much those allow, but wouldn't dare go much past 3:1 to both allows already quite big drop.

oneWayViscousCoef affects quite bit to issue, that might have something similar in effect as clutch torque which was mentioned earlier, search did not reveal any mentioning about this, but it seems to have similar effect to clutch torque as setting it low and no wheelspin, increase value and you get more wheelspin and less slipping in torque converter(?).

Torque converter stifness I still don't quite have found out.

Torque converter diameter seems to cause engine to stall quite easily if set too high, if set causes again some slipping in drivetrain, I guess at torque converter, but this seems not to have much to do with issues mentioned in first post, oneWayViscousCoef however does have some effect on that.

If you set differential ratio to 10 or above, there might come issue were rear tires spin wildly and whole drivetrain becomes unstable, but of course it is a sum of all those parameters.

 

IRL, there are not many differentials that would go much higher than that - for the same reasons

I recently designed a powertrain setup in-game for a mod that simulates a real monster truck. It uses gear reductions in 3 places to achieve realistic final drive ratio.Transfer case, differential, and planetary gears. A similar result could be achieved without all of that, and just the f/r differentials being set at something like 23:1ish. But that was with 66" tires

So, the game can go well beyond what is expected as 'realistic'.
For simulation of real world parts, I'd say is perfect, and not bugged or limited by design. Anything will break if you increase it's values too much.

Also, be sure to check the Powertrain documentation... it explains a lot of the values and what they do.

 

Thanks from the link, somehow I had missed whole wiki thing, stupid me

Game certainly goes quite far, there is reduction in wheel hubs IRL (I guess monster trucks have such too?), which I think is not in game, but when balancing all the places it does marvels indeed.

I doubt that many would find it enjoyable to 'drive' 80 meters per hour, so it might not be needed to be possible, but similar way I doubt you can make monster truck to go that slow, there are limits and I think they are well reasonable limits, now that I have learned more about how to adjust ratios.

Mod I'm altering has originally quite wild amount of power and grip, with bit of researching and adjusting I have managed to make it quite close to my liking and yes, more realistic values I find and can input to game files, better behavior seems to become.

I'm not quite sure what I was thinking with 102.7:1 CVT ratio though

 

Yea, I am sure that CVT was going nuts in the simulation. I don't even know how that would work in real life

 

You just need 1cm pulley and 102.7cm pulley, first one is a challenge, make it bigger and then 2nd becomes a challenge, but in theory it is possible

 

That would be one hell of a crazy contraption that I would love to see.

 

I wonder here if BeamNG supports animated CVT parts, you could then put CVT unit on flatbed of T series as it does not matter where CVT is, as long as driveshaft enters and exits from the unit, it would be beautiful to see ratio changing realtime, too bad I have no idea how one could actually build such, but indeed it would be sight worth seeing

 

You compound them. You would never build 100:1 reduction in real life. But you can build a 20:1 reduction and feed it into a 5:1 reduction and get 100:1 overall.