Cars (DATA REPLACEMENT) Nissan Skyline R34 GT-R V-spec Improved Physics by Arch

[JackCY]

I had no issues with 1.3 or 1.32 easy to install but I already knew how to so there is that.

What's your best time around Nord. Tourist (100% grip) with the S3?
What's your average consistent time?

 

[Kyuubeey]

I had no issues with 1.3 or 1.32 easy to install but I already knew how to so there is that.

What's your best time around Nord. Tourist (100% grip) with the S3?
What's your average consistent time?

The data.acd was corrupted for me when I extracted it, so I decided to fix it just in-case. I also noticed the rod_lenght and UI issue so why not fix it at the same time.

No idea. I've been too busy testing and devving to drive seriously. I could probably knock out good laps up there with the fastest.

 

[eclip2]

So I've recorded the issue with bouncing wheels. I can clearly feel on the steering wheel that bouncing and it's quite difficult to control the car mid braking, especially on turns.

 

[Kyuubeey]

That's an AWD2 bug. Only happens if you're stationary. Not when driving to my knowledge.

The ATTESA also has a drivetrain brake function: that should aid stability, not hinder it. It's an engine brake on the front axle: doesn't affect rears. If ABS locks the brakes a little it will try to reverse polarity of the drivetrain brake which you might feel in the wheel.

Can you record a video of you *driving*? Then we can tell better.

 

[eclip2]

Notice what happened to the car in the last corner.

 

[JackCY]

The data.acd was corrupted for me when I extracted it, so I decided to fix it just in-case. I also noticed the rod_lenght and UI issue so why not fix it at the same time.

No idea. I've been too busy testing and devving to drive seriously. I could probably knock out good laps up there with the fastest.

Possible, I'm sure I drove the 1.3 because after I installed it I noticed there is 1.32 XD So I had to do it again.
---

Bouncing wheels in AC... tell that to Kunos, the wheels have been messed up in AC since probably day 1. Some cars overheat tyres standing still from all the bouncing, such as RUF on LAC track. Brakes when driving...while Arch probably tried to do it as the "real" thing, in AC this doesn't work well at all. I know this is probably still very much work in progress and from what I saw in data and while driving there are quite a few things I would immediately fix.

eclip2: In last U turn corner it looks as if the car is bottoming out on suspension, the suspension distances for bumpstops are not easily set in AC. It's doable and calculable but even some stock cars suffer from issues in this department. Your video shows a lot of the issues I've encountered myself too.

---

Well any reference laptime you can give on any track. Nord. Tourist is easiest for me to compare as I'm used to driving my own R34 there, less power than your S3, still on stock tyres, stock aero. I'm sometimes on the Japanese Pack server #1 or #2 that are modding open with it if you want to have some moded R34 fun online.

I'm not sure just how fast or slow your S3 is supposed to be, none of the variants really fit me personally although I was able to set the S3 to something more reasonable for my driving, still it may simply be set to your preferences so it would take me more time to fine tune it to know what time it can actually do and ultimately I would end up changing the data instead of trying to fix it in setup menu, change the source of issue not the try working around the result with other settings.

If you want to do some laps online, let me know, same nick and avatar on Steam, or PM me here.

 

[Kyuubeey]

@eclip2

You drive so un-smoothly and coast a lot (Promoting engine brake oversteer) that I will have to put it down to driver issue.

@JackCY

Nearly everything on the car's either from a document or from a measurement. Barring tires. Even the unsprung is calculated from known masses down to the washer and bolt.

Do tell me what mistakes you have spotted and would change immediately. On the stock car, of course. Perhaps I've missed something drastically, but the video-comparison and user feedback shows the car to be at least "4/5" accurate at the very least.

Laptime is nearly impossible to compare unless we know the exact conditions + use the exact semislick compound that they used.

 

[Kyuubeey]

Also I don't understand what you mean with bumpstop distances are not easily calculatable. We kinda measured that, for rear at least. For front I went with anecdote. If you mean setting the susp height to neutral, you can do it with the suspension travel app.

 

[eclip2]

As I mentioned previously my driving skills are rather poor I wasn't trying hard on driving nice though, just launched the game and quickly recorded what happens to the car while cornering. Not to mention that only with this modded r34 I have experienced bumping wheels to that extent. As you can see in the video at 2:19 the car started to jump around without any reason.

 

[Kyuubeey]

The V-spec dampers were a little too stiff for the car IRL.

Actually it doesn't jump enough on Nordschleife, not nearly enough. But all the important bumps are the same. So I suspect that it's just because track conditions have changed since then and the small bumps have mostly been evened out.

 

[JackCY]

@eclip2

You drive so un-smoothly and coast a lot (Promoting engine brake oversteer) that I will have to put it down to driver issue.

@JackCY

Nearly everything on the car's either from a document or from a measurement. Barring tires. Even the unsprung is calculated from known masses down to the washer and bolt.

Do tell me what mistakes you have spotted and would change immediately. On the stock car, of course. Perhaps I've missed something drastically, but the video-comparison and user feedback shows the car to be at least "4/5" accurate at the very least.

Laptime is nearly impossible to compare unless we know the exact conditions + use the exact semislick compound that they used.

Also I don't understand what you mean with bumpstop distances are not easily calculatable. We kinda measured that, for rear at least. For front I went with anecdote. If you mean setting the susp height to neutral, you can do it with the suspension travel app.

I've checked these variants, normal, S2, S3 and here is what I mostly saw:
Insane aero lift/downforce balance to the front making the car oversteer. Something like 100%+ front balance. Even with maxed out rear wing on S3 it can still be a bit too high front balance.

The suspensions weren't always leveled out/neutral as the camber setting in setup menu was sometimes widely different from live value. If you measure everything with car on ground AC will take it and won't need adjustment/recalculation, but if you enter say suspension with car measured off ground then the camber won't be same live and in setup menu but more importantly if you enter bumpstop distances at the same time measured with car on ground then they will be widely off. You can check the bumpstops distances by making the limits rock solid. If you have it set "neutral", by that you mean that the car on ground in AC is at same height as the real car? And suspension did not move from the data you entered? Then yes the bumpstops:

BUMPSTOP_UP, BUMPSTOP_DN, PACKER_RANGE

Should work fine with real measurements. Sadly most cars always move, some by a lot (especially original AC cars) and their bumpstops have to be recalculated or suspension moved in data to a "neutral" position.

Max boost and wastegate are equal, this is not how turbos work. Max boost should be higher than wastegate. Plus there is often a turbo controller that you don't need. Lets say turbo max would be 1.0 and wastegate 0.8 normally.

Missing progressive part of springs and bumpstops. AC doesn't calculate it, the MR change gives you a progressive part even if your spring isn't progressive itself since springs in AC are defined at wheel not at spring attachment points since AC doesn't even have any spring or damper attachment points at all. It does everything at the wheel/hub.

As much as you probably wanted to improve braking with AWD system by keeping it active during braking... this does not work in AC and the cars have terrible braking performance as a result. So much so that at times it was as if someone disconnected the rear brakes altogether.

Differentials such as:

REAR_DIFF_POWER=1.00
REAR_DIFF_COAST=0.50
REAR_DIFF_PRELOAD=150

Are mental and unreal. Welded 100% on power? Not even 99% with tiny slip but 100% welded. Also very high preload, probably trying to keep the car stable due to aero oversteer?

The S3 would probably have even higher reference RPM for turbo. With power right now being that of an RB35? Or RB40? I would have to calculate but it is definitely some hell of a brutal power for RB26. 1.2bar 700+hp? Seems unreal to me. RB26 650hp 1.5bar or there about should be a well tuned engine. 700+ 1.2bar is a different motor to me. If it was 700+ at 2.0bar sure why not. There is a whole RB AU forum thread with dynos for decade+ to check power, turbo, etc.

Weight copied from stock Kunos car yet weight balance is recalcuted? Odd. And S2/3 being same weight as stock car is even more odd.

Tyres seemed to grip quite a bit compared to stock. But I could be wrong here.
AWD can't say since the cars seem so imbalanced but the way power is put down right now seemed OK though who knows how that will be when cars don't oversteer from aero etc. at any small input at any speed. Plus locked weird rear diff isn't helping either to say if the front-rear AWD power is OK or not.
On braking and coast the AWD is bad.

In reality S2/3 would likely have 4 wheel steering disabled. Probably rear active diff gone too which S1-S3 in AC don't have either so good there. Personally for low power rear active ERS+brakes hack diff in AC kind of works because it's not so invasive but at higher power levels it was always slower (lap time wise) than traditional passive LSD, the handling was worse, less predictable, steered a little different and wrong, not the way active diff should as the torques were applied differently. For GTR (R35) that has passive LSD + brakes ... AC can do that but for R34 with active rear LSD... AC can't do that correctly :/ Many R34 don't have the active rear anyway and who tunes an R34 with ALSD likely removes the active diff too and replaces with a good passive LSD.

ALSD is "nice" but sort of a gimmick for cruising that is hard to perfectly tune and use for track/race use.

To me the springs and dampers are soft, but then I'm looking at a road car stock version so that's to be expected. I wouldn't say they are stiff.

Also suspension seemed to change toe with travel even more than Kunos's R34. I thought this "new" suspension was meant to address these sort of issues not make them worse?

I didn't see that much in terms of a need for shader patch/physics hacking. What changes in the data actually require it beside: TORQUE_MODE_EX, which I would have to look up if there is where to know what it's supposed to do or fix and then check if it actually does that or not, EXTENDED_RAYTRACING same thing, not sure there are any more physics hacking extensions. Or that the way it's hacked is actually correct or more correct than default AC behavior, would have to first have access to code of both to be able to see what it's actually doing/calculating before saying one is better than the other. Some suspension things in stock AC used to be and probably still are broken/off, no doubt about that.

I don't think the car would behave drastically different without physics hacks.

The hub weights seemed fine to me.

---

Post some lap on a track, time or video, maybe you have it set for you and can drive it around fast. For me it seemed far off from what I would consider driveable and the car behaved murderous. Maybe I'm driving it wrong but considering the data I saw to me it seems the problem is mostly there not in my driving. So I would like to see a fast race lap with the S3, or any other version really (S3 seemed easiest to drive for me since I could tune it in setup menu), to see that the car is actually driveable by someone.

Nord. Tourist is easiest for me to compare on, 18C air 24C track, 100% grip, mid clear. There are 2 Jap. pack servers online with same settings too if you want to meet up online with moded R34

 

[Kyuubeey]

@JackCY

I'll read it a bit later, thanks for your thoughts.

From the first few paragraphs I read, there are almost no real issues whatsoever.

Aero balance is correct +0.5kg or so at 200km/h and at all wing angles. You should be aware that drag influences aerodynamic wheel loads more or less as much as the aero surfaces will on a car like this.

Setup range and alignment is correct +-0.05deg camber, toe is accurate +-0.01deg.

Tire grip is same as KS: the grip dropoff is more accurate than KS tires. Bumpstops are accurate and also implemented accurately. Mass is accurate to the design specifications: I don't know the real mass to the gram. CoG is also accurate to the design specifications: I don't know the real CoG to the gram. KS is completely wrong CoG wise lol. Unsprung masses are accurate to the 0.01kg but the CoG calculation as always is surely off somewhat, so maybe give or take 1kg at most.

The LSD's are to manufacturer spec, but I am a bit dubious always about "100% lock" LSD myself. Then again, there are dozens of plates...

Turbo doesn't matter, there is no "turbo simulation" in AC, like real pressures and cm^3. You can do it almost however you want and this is how we matched it with my testers. I think if I did it more "properly" it would just be more work for no actual gain in the end. If you knew about GT-R's you would also nitpick the fact that the two turbos do not output the same pressure, but in our implementation it is largely irrelevant.

You cannot replicate toe and camber curves for Nissan's multi-link rear in AC: while also retaining internal forces correct. I settled for a compromise that offers only noticeable distortion near the very ends of travel, and preserves the operating range as well as possible. It's literally impossible to improve it except by doing another compromise, or by fixing the internal forces in AC. I'm still searching for perhaps a better compromise, but this is the one that doesn't mess with the driving too much, according to tester feedback from R32/R33 at least. The practical bump progressive understeer behavior is also similar to IRL: it could very well be oversteer which is not correct.

Engines are from curves/graphs, except S1 is simply boost-up.

Shader patch physics are currently needed because I plan to implement more features as they are available for DWB.

I'll check the thing about progressive spring and stop, but it shouldn't be a real issue currently.

 

[Kyuubeey]

PS: The ATTESA diffbrake isn't there to improve performance. I doubt it improves performance at all. It's really just a stability aid more or less, and it does work how I want it to in AC, so...

 

[Kyuubeey]

@eclip2

Hey. I'm driving now the S3 on Nords Tourist and I noticed your issue, with the wheels hopping. I've never had this happen on any car in fact. And it happens on the same spot. I will have to investigate if it's from the tire raytracing, colliders, or what. But it is very strange because the physics collider should not be hitting anything.

That's definitely not okay however. I'll investigate. GT-R's with similar suspension height and stiffness didn't do this there, so it's very strange indeed.

EDIT: Turning raytracing off seems to have fixed it. During that one test. I will report it to be sure and test further.

 

[mclarenf1papa]

Missing progressive part of springs and bumpstops. AC doesn't calculate it, the MR change gives you a progressive part even if your spring isn't progressive itself since springs in AC are defined at wheel not at spring attachment points since AC doesn't even have any spring or damper attachment points at all. It does everything at the wheel/hub.

You can't replicate non-constant MR behavior with the progressive rates in AC anyway, so not a big deal...

And there are currently 39 - though not sure about the count for the current public release - physics "hacks" as you so call them. Torque mode was a Kunos thing that never got put into AC to fix the incorrect application of suspension forces (I promise that seeing the code wouldn't help with understanding if it truly fixes the issue or not). Raytracing is pretty self-explanatory as it's just extended raytracing for the tire's contact height to attempt to improve the single-point contact model (it's always been a bit buggy - I don't really have anything to do with that particular feature's development, so not totally sure what it's doing).

 

[Kyuubeey]

If I had a graph for MR I'd implement that properly via the new DWB additions.

I mean I can guesstimate it, or put geo into a simulator and just graph it and it should be relatively accurate I believe. Not like they change super drastically from one susp to another on roadcars.

Thing is the front I made is a bit weird in how it's done, where I attempt to retain the camber/toe curves first, then I fix the anti and try to again retain the curves. You can't *really* do it 100% in AC no matter what you try due to it's design. So the only way to get MR accurate for that is to get someone to measure it. No one wants to jack up their car etc. just for that.

I don't think simulations will output the true result: shouldn't. So I will make do for now, perhaps when I implement more DWB changes I will make it a very smooth, "artificial" MR curve based on similar suspensions, or leave it constant. Both are as accurate/inaccurate.

 

[AadHofman]

Post some lap on a track, time or video, maybe you have it set for you and can drive it around fast. For me it seemed far off from what I would consider driveable and the car behaved murderous. Maybe I'm driving it wrong but considering the data I saw to me it seems the problem is mostly there not in my driving. So I would like to see a fast race lap with the S3, or any other version really (S3 seemed easiest to drive for me since I could tune it in setup menu), to see that the car is actually driveable by someone.

A decent lap with S1 around the Nords (8:01.1). I was too cautios around 5-6 minutes marks, basically in a crash avoidance mode, so lost some time there.

 

[Kyuubeey]

Very nice lap. Perhaps I will try one myself, later. Does it still count if it's a slightly changed dev-build?

 

[AadHofman]

Of course not.

 

[Sanch0z]

I've managed to pull 8:03 around Nordschleife on V-spec (quite similar to GT-R R33 8:01 time by Motoharu Kurosawa considering 200m difference between full lap and real life hotlap distance) and 7:55 on S1. All with 100% grip, clear weather and blanked semislick tyres. I think it is possible to go under 8 minutes, but it requires big skill. The mod itself is excellent. R34 feels more alive now. It became much more agile and tail happy. Though, A-LSD is more annoying than helpful, especially while braking, but ATTESA is real helpful (it can really save your ass on corners like flugplatz and shwedenkreuz if car goes sideways).