Micro Blog(s) - 2017

Little update from my side:
We have decided to not allow any shift anymore. Previously, the game did not use any safeguards against weird or almost impossible shifts, for example it let you shift into too low gears, overrevving the engine or let you shift without using the clutch (essentially assuming unlimited force to engage the gear and applying gear damage at the same time).
We now moved away from this and implemented several safe guards and changes to this:

• There's a new "Gearbox safety" option which prevents you from using gears that would exceed the engine's maximum rpm, the game will simply reject the shift
• This also applies to switching non manual transmission modes (for example going from D to 2/1)
• In addition to that we made the switch to rejecting shifts that did not use the clutch at all. While we are still working on this and we might change the design slightly, the idea is to prevent users from "cheating" by not using the clutch while tolerating the applied gear damage. We also try not to interfere too much with users that don't press the clutch far enough.

 

A little update on the previous blog:

• There has been some confusion about whether the gearbox safety will be enforced or an option. Of course it will be an option We won't restrict your joy in destroying your favorite gearbox if you so wish
• The reason the safety option has been added is that overrevving now causes serious damage, so it tends to destroy your engine pretty quickly if you don't watch out. To prevent frustration about that, the gearboxes will protect themselves a bit better when that option is enabled
• In addition we switched from outright rejecting shifts to "cutting power" for a short time. This makes it much less frustrating to miss a shift while still penalizing you for it

 

Working on things, loving the cappuccino freddo

 

Work, 10+h days, not holidays guys

 

Just a little detail to show you, I added adjustable caster to most of the cars in the tuning menu. Adding more caster usually results in stronger self steering forces, making drifting more predictable. However too much can be a bad thing, upsetting the balance of the car due to weight jacking, as well as negatively affecting bump steer and steering ackerman.

Also if you look closely, you will see that we added a smoother way to implement the alignment changes, this way the suspension should no longer become damaged or break from extreme alignment setups.

Video - Click to Play - Direct Link

 

Finally got around to fixing that DCT bug where you could drive in the wrong direction for a brief moment when switching from R to D (or vice versa) under somewhat unrealistic circumstances

 

Buildbot broke, we needed to reinstall its OS and compilation environment, took us until 4am yesterday. Testing out the Jenkins Pipeline jobs to build the game. (This is an accelerated quick build, the complete build takes around 20 minutes):


And the steps to upload the game to steam:


Lots of work on things you will never see ^^

 

Something new in something old.

Video - Click to Play - Direct Link

 

Work in progress

Video - Click to Play - Direct Link

 

Just a little feature for you drag fans out there

Video - Click to Play - Direct Link

 

A tiny little detail, but like most real life transmissions the M mode now does actually shift down automatically if your rpm drops to almost idle.
This ensures the car won't bog down too much and makes driving in M mode much easier.

Video - Click to Play - Direct Link

 

The lateral damping of our tire carcass was too low before, but we found a way to increase it by 4-5x by using specialized damping. This has also made the tires feel more direct, predictable, and smooth through ffb.

Left is old, Right is new. I used the most wobbly old bias ply tires to make it easy to see the difference.

Video - Click to Play - Direct Link

 

A little more information on how the N2O system works:
The basic components are:

• N2O tank with a given capacity and actual content
• Purge valve nodes
• N2O controller which just acts as a little proxy to enable/disable things
• Of course the added power can be selected either via hardcoded values in jbeam or for example with a slider in the tuning tab

You have two basic modes of operation:

• Automatic injection
• Manual injection

The automatic injection has a minimum RPM, a minimum gear and only actually injects any N2O with full throttle and when not on the clutch.
It can be armed/disarmed (arming also triggers a short purge visual) via a key binding, everything else then happens automatically.

The manual injection on the other hand does not have these safety guards, does not purge and just lets you inject N2O whenever you want really.
This has a couple of drawbacks. To prevent slipping of clutches the actual torque from the N2O system is taken into account while calculating the clutch torque, however, this uses the minimum RPM from the automatic injection as a baseline. So when you manually inject at lower RPMs, you'll likely slip the clutch.

(Purging is just a visual thing btw)

For those that don't know how N2O works and why the above issue exists: N2O injection essentially adds a constant power offset to your engine (hence the "xyzHP shot" names), however, since essentially power = RPM * torque (with the right units) having an additional 100hp at low RPMs requires a much higher additional torque compared to another 100hp at 5000 RPM.
To ensure the maximum powertrain stability we always want the clutch torque to be as low as possible, so making it work with insane amounts of torque at very low RPMs does not make sense.

Using N2O in the game also touches various other parts that we are still working on. (For example engine thermals, efficiency, durability etc)
Don't expect miracles though, we work with what we have engine wise, so you're not getting some sort of new and highly detailed engine simulation

 

Fans of trucks and solid axle vehicles will appreciate the newly added torque reaction features. Important results of that are:

• Chassis twist, very relevant in heavy trucks
• Asymmetric tire loading from driveshaft torque. Due to the driveshaft torque twisting the solid axle, these cars will tend to spin the rear right tire, and tend to handle differently left vs. right under power
• Engine rocking when revving due to flywheel inertia
• Engine rocking due to the opposite reaction of the driveshaft torque

Video - Click to Play - Direct Link

 

Here's a follow up video to the mysterious earlier video:

Video - Click to Play - Direct Link

 

Hot-loading for Mesh is fixed too

Video - Click to Play - Direct Link

 

Currently working on small improvements for the spawn algorithm of External camera. Here's some debug visualizations of two of its worst enemies:

One of them is curved tunnels:

The other is... fast tunnels:


Red lines are the reasons why specific spawn configurations were discarded.
Blue lines are those discarded spawn configurations: each blue line represents the predicted camera path during its panning movement. If the camera was static, it would be just a point, rather than a line.
The black line represents the chosen spawn point. Green sphere is the original position, white is its current position, red will be the final position, if everything goes according to the basic predictions we performed at the beginning of the take.

 

Sneak peak, work in progress

 

This is a statistical profiler, and is one of the tools a developer can use to optimize software:


It sneakily gathers stats about what the program is internally doing at intervals, for a period of time. If this profiler thingie caught BeamNG.drive doing the same thing (in the same function of code) 90% of the times it has taken a look, then it would mean we are probably spending 90% of the time there.

More samples means more reliable stats. In this case, the profiler has taken 62.000 samples while I was recording a replay of an ETK K-series doing burnouts and donuts.

Once you know what the cpu time is used for, you have a better idea of where it might be worth to spend your time optimizing. As you analyze and rework the chosen code, you run the profiler again to verify the results are indeed what you expect (they not always are).

In this case, the changes resulted in roughly a 0.6% increase in framerate for that specific replay recording scenario in my computer.

 

So, in the first days of the repository our CDN provider Akamai really helped us push the load - so everyone can have fast ingame downloads

The peak bandwidth at one point was 1939 MBit/s

Since its release, we had 387 thousand downloads with it