Update 1.16 Physics Changes

[hasslemoff]

Not a final conclusion yet but a 80 % conclusion that there has been a change, surprised I was as after testing everything from LMP's to Classics, RH to CH tyres, ABS off and on, comparing data logs, replays and motec analysis plus reading all the differences that people were feeling.

90% of the replays I had were all as you say hot lapping, trying to get the most out of the car and track were any minor change would not be noticed as you automatically adapt to the situation as you are use to slight changes as the car and driver are on there limits each corner.
Basically the change will not be noticed as its that small and people are use to minor changes while driving on there limit each lap, some might of still notice'd it but it seems by the responses that it does not seems to be the case, as with me.


I'm not fully committed to there is change yet as I'd like to do more testing and as people have said that its easier to drift, i'm not to see this or feel it yet, but that could be down to again been used to catching a drift while been on the edge of grip, this part I still notice no difference.


One combo I noticed a difference which should even be visible in a replay, but I would like to test it more before making a true statement, also this might help the people who like me are still unsure or have not felt a difference if I explain what is happening and how the change has effected the car.
Luckily I only ran this combo a few weeks ago and even have a video from 1.15 with a cockpit view that might help in what i'm trying to describe. found here

• Car Ferrari 250 GTO no tune
• Tyres CS
• No ABS
• Damp track 20% real track settings
• Free run mode

If you have a replay or drive a car regularly that you are not hot lapping is the best way to notice if the change is there, basically just a random lap and yes No ABS shows the effect more even if it is small.

Quick rundown on what I noticed on that combo between 1.15 and 1.16.

Each corner the brake is used to turn the car into the corner so as the steering is smooth, you will notice this in the above cockpit view video that the end comes around with no extra steering input, this is best seen in corners 4 and 5 on entry to the corners, driving like this is where I noticed a change.
The lap is not drove like a hot lap but drove quick but not on my limit.

The change I noticed in two corners and cannot replicate even by changing my brake point and approach, in other words manipulate the car to make it look the same with the same set up as 1.15.

Turn 2 is were I first noticed the difference, hitting the same brake point at the same speed with the same amount of brake pressure the car turns in more only a bit but enough for me to not hit the apex the way I want so I can get on the gas when I want I have to wait a split second longer.
To produce the same turn in for turn 2 it feels like I would have to drop the rear BB by 1 or 2 to get the car pointed the way I want.

Turn 6 the up hill right hander on 1.15 the car always felt unsteady on braking and seemed to always need slight steering input before turning in this is shown on the replay, 1.16 its not unsteady I can keep the wheel straight 95% of the time while braking. But the car turns in slightly more.
Again with the car in the same settings I can not manipulate it to do the same unless I brake way late and then that totally messes the corner up.
Again I would drop the rear BB by 1 to get the turn in I want, but would have to raise the rear BB by 1 or 2 to make it unsteady on braking up the hill.

Ill try and get some videos and data for those corners later in the week, but they are what are changing my mind, you might think two corners are not a reason to change your mind but the car will not do what it did in 1.15. Again im only 80% and until I get time to look through it properly I wont say it's definite.

If you are on the fence or feel no difference but would like to see if there is a change and you are like me and mainly push the car and driver to the limit, try a car that you just drive around in It took a damp track, no abs and the ferrari 250 gto on CS tyres for me to notice anything and its mainly the braking and turn in where you would notice any difference.
Hope it helps.

 

[tarnheld]

I'm just going to post the charts with no explanation, refer to my previous post (mainly parts 2 and 3) for how I read them - I don't want my bias to get in the way.
I jest.

Note that corners are considerably shorter with this combo (MX5 TC at Riffelsee), about 1 - 2 seconds being a reliable mode value, corresponding to a steering and yaw rate wind-unwind frequency of 0.5 - 1 Hz.

As much as i appreciate discussing these matters with data backup, the diagrams doesn't convince me that there was a physics change for the following reasons:

• The steering angle FFTs of both laps have differences of ~10dB, just as much as the yaw rate FFTs. So if the steering angle is the input and yaw rate is the output why need a physics change for explanation when input differences suffice?
• The FFT can give results that can be misinterpreted.

Just to give you an example for the second point, here is another diagram:

As you can see, i added the throttle input to the other stuff. As you know the throttle input is mostly constant, but there are spikes when shifting. The FFT of the throttle input has therefore components in all frequencies (the frequency is plotted linear here, and cut off at 10Hz). As you see, there is a ~10dB magnitude difference of throttle at 6.9Hz between the two laps -- easily misinterpreted as meaningful when it's just random differences in the spikes.

I'm open for any hard facts showing a physics change -- but i stay in the 'don't know'-camp for now.

If you want a simple test that produces data without noise, I recommend looking at data between 1.15 and 1.16 using the follow method.

Load up an A-spec race, one where you have a rolling start and wide surface, and a race that starts you out at a decent speed.

If you're on a wheel, turn ffb off and keep the wheel at full lock before the start and until you hit something/leave the racing surface. Complete the lap/race and look at the data.

If .15 and .16 show different telemetry during the full lock at start sequence, then there has been a physics change.

I really like this idea, simple and eliminates driver input. 👍 Should have saved some replays from the previous versions, but better to start now than never.

 

[Ridox2JZGTE]

@Ridox2JZGTE it is just me or the FF cars has been improved closer to real I've raced all your four Focus STs. They all are just a blast to drive. Very enjoyable. Happy!

He's a regular driver in GT6, mostly online racing and A spec races. He has close to 70k miles under his belt, reviewing many tunes and replicas for variety of tuners here. He has not been made aware of this thread or any discussion about the physics changes. I quoted his post and gave a link to this thread about possible cause for his finding.

I also found the 25th Anniversary Countach has somehow react differently on racing hard tires ( without body rigidity ) I have built replica of the car in variety of version ( stock, race car etc ) and before 1.16, they were not particularly good on racing tire. A few days ago, I decided to drive it again, and it was so much better I tried the Ken Wolf/Rain X Countach replica on RH tire, and immediately have a great time at Suzuka, Motegi Road and Midfield.

Midfield is the track that made this car shines, the racing tires felt like making the chassis flexed/overworked, but the way the wheels react to my steering input was a lot better than it used to be. I can feel when riding the slip angle ( the car has good amount of camber ), and when making corrections with throttle I can feel the rear heavy tail squirm and gain traction like a real MR on steroids The braking part also great, where I can better trail brake ( no ABS 7/8 BB ), particularly Suzuka 1st turn and Motegi 1st turn and the downhill straight leading to finish line.

 

[Griffith500]

As much as i appreciate discussing these matters with data backup, the diagrams doesn't convince me that there was a physics change for the following reasons:

• The steering angle FFTs of both laps have differences of ~10dB, just as much as the yaw rate FFTs. So if the steering angle is the input and yaw rate is the output why need a physics change for explanation when input differences suffice?
• The FFT can give results that can be misinterpreted.

Just to give you an example for the second point, here is another diagram:
View attachment 322828

As you can see, i added the throttle input to the other stuff. As you know the throttle input is mostly constant, but there are spikes when shifting. The FFT of the throttle input has therefore components in all frequencies (the frequency is plotted linear here, and cut off at 10Hz). As you see, there is a ~10dB magnitude difference of throttle at 6.9Hz between the two laps -- easily misinterpreted as meaningful when it's just random differences in the spikes.

I'm open for any hard facts showing a physics change -- but i stay in the 'don't know'-camp for now.

...

The throttle blips account for a small proportion of the total input, and being broadband all they really do is raise the noise floor. The FFT doesn't say whether the input is truly constant, because it doesn't go down to DC. All it says is how much of the changes in the signal can be binned into certain frequency range, for a full cycle of adding-removing-adding (or, equivalently, removing-adding-removing) throttle. The high frequency stuff is adjustments, the low frequency stuff effectively "set and forget", the same as for any other control.

There is a difference between a 10 dB peak difference over a band 0.2 Hz wide centred at 6.9 Hz (representing a 0.04 octave range), and the 3+ dB approximate average difference in yaw rate starting at 0.6 Hz all the way up to 4 Hz (well over 2 octaves) in the yaw rate between laps 3 and 1. Especially when that difference is consistent between the two 1.16 laps. -3 dB power is a reduction in amplitude of about 30% - significant, especially over a broad band.

The reduction in throttle inputs at 6 - 9 Hz is evident in the plot you provide - only half an octave, but far more significant than the more impressive 10 dB gap you isolated. I know you were illustrating a point, but it's just not a pitfall I fell into. This is why frequency-domain plots are usually logarithmic in the frequency scale - think you can tweak your input by 1 Hz equally well from 3 Hz and from 10 Hz?

Additionally, the throttle inputs, like the brake inputs already discussed, will be subject to changes if the car is harder or easier to control. The input differences are all that are being discussed; but why would someone's driving style change for the same car, same setup at the same track, when aiming for the same thing?

I would argue that lap 3 shows a significant response from the car in its "inertial" range, i.e. the literal limits of the car's response.

 

[tarnheld]

Let me try to formulate your point, as i get it, before we talk at cross-purposes.

First the facts:

In your post the first diagrams show the FFT of yaw-rate and steering-angle of hasslemoffs 1.15/1.16 laps.

When looking at the Lap3 vs. Lap1 and Lap2 vs. Lap1 yaw-rate, i can see there is a difference between 1.15 and 1.16 in the 1-3 Hz range peaking at 2Hz with a magnitude of ~10dB. The steering-input in the Lap3 vs. Lap1 comparison shows roughly the same difference at the same range, in the Lap2 vs. Lap1 the situation is more complicated.

When looking at the Lap2 vs. Lap3 (both 1.16 laps) yaw-rate, the difference magnitude is only ~5dB and in the 0.7-1Hz range. The steering-angle difference is ~10dB between 1-2.5 Hz.

So if you are saying the yaw-rate difference points to a physics change, i would argue that the steering-input differences can explain the yaw-rate differences just fine, being of the same magnitude. I see the yaw-rate as being caused by the steering-input and the GT6 physics, so any change in the output without corresponding change in the input would be evidence of change in the physics.

But if you are saying the yaw-rate difference and steering-input difference have to be taken together to come to the physics change conclusion, i can not follow this argumentation, because i think the steering-input differences can be attributed to the driver alone. When trying to get evidence for physics change, i think one has to eliminate the effect of the driver.


The throttle input should not be taken into consideration here, i think the frequency analysis of throttle input is meaningless, because it is just a random almost square signal. Nonetheless it has a Fourier Transform, but to give any meaning to the frequency magnitudes is misleading. I just wanted to show that with my diagram.

 

[Saidur_Ali]

My gut feeling is that nothing has changed regarding physics but I haven't driven any combinations I drove on 1.15 with wheel. Only back-to-backing it I think I would have a much better chance of spotting a difference if there was one but I'm always amazed the changes people can notice.

 

[Johnnypenso]

One of the problems comparing single laps is that there is far too much variance in a single lap even with the same overall lap time. I know myself when hotlapping and trying to climb up a leaderboard, I can run laps back to back with similar lap times, but composed of what seem to be significant errors on certain corners, requiring larger than normal corrections but in reality I'm only a few feet off the "proper" line. Is there a way to take the data and compile multilap averages?

 

[tarnheld]

Is there a way to take the data and compile multilap averages?

Would be cool, i didn't found that in Motec i2. The averaging needs to be done sensibly, could easily hide the changes searched for.

Part of me thinks about reinstalling GT6 from 1.00 up to 1.16 and trying to do @Bhowe83's procedure, maybe on Motegi Oval. I think the reinstall procedure is described somewhere on GTPlanet, could take several hours, and you would have to be careful not to mess up your savegame. The procedure could even be done with a DS3 if i understood it correctly, just start an arcade race (rolling start), press the left/right button until hitting a wall, then drive one lap and save the replay.

But the other part shies away from this herculean task. EDIT: and it could be all in vain, as the procedure is not guaranteed to produce the data that shows physics changes.

 

[hasslemoff]

Uploaded replays from Version 0 GT6.

• Custom DS3 settings for GT6 used (ds3 not wheel)
• Courtesy cars used only FF, FR, 4WD and MR
• Alfa Mito, Scion , Range Rover and KTM
• All Aids off including ABS (driving line and pass indicator was on couldnt be bothered turning them off)
• Comfort Soft tyres on each car
• Track settings real
• Time change for silverstone stowe standard
• Arcade Mode single race
• Tracks used Silverstone Stowe and Motegi Oval

When starting in the rolling start I had the stick of the DS3 fully to the left and (X) fully pressed for each car and track, Stowe went grass cutting, motegi Oval went into the wall.
Im might get time to upload from version 1.16 later on tonight.

There is also two extra replays using the same input with SRF on and Track settings Low just for kicks.
Not uploaded in Motec form sorry running out of time.

1 to 4 = Stowe FF, FR, 4WD then MR
5 to 8 = Motegi Oval FF, FR, 4WD then MR
8 to 10 = GT86 with Low track setting then SRF on

 

[hasslemoff]

Motec replays from Silverstone Stowe have been uploaded for Version 0 and Version 1.16.
For Silverstone Stowe in the courtesy cars, Alfa Mito, Scion, Range Rover and KTM I fortgot to rename them but when you view them in Motec you will notice a name difference T-Test is Version 0 replays.


Also the 4 replays of the Stowe test in Version 1.16 are uploaded as well starting BCJ

 

[Johnnypenso]

Would be cool, i didn't found that in Motec i2. The averaging needs to be done sensibly, could easily hide the changes searched for.

Part of me thinks about reinstalling GT6 from 1.00 up to 1.16 and trying to do @Bhowe83's procedure, maybe on Motegi Oval. I think the reinstall procedure is described somewhere on GTPlanet, could take several hours, and you would have to be careful not to mess up your savegame. The procedure could even be done with a DS3 if i understood it correctly, just start an arcade race (rolling start), press the left/right button until hitting a wall, then drive one lap and save the replay.

But the other part shies away from this herculean task. EDIT: and it could be all in vain, as the procedure is not guaranteed to produce the data that shows physics changes.

I wouldn't so far as starting over at 1.00. 1.15 vs 1.16 is good enough.

 

[tarnheld]

I wouldn't so far as starting over at 1.00. 1.15 vs 1.16 is good enough.

Is there a way to go back a version?

 

[Johnnypenso]

Is there a way to go back a version?

Not that I know of

 

[tarnheld]

Not that I know of

Dang!

But @hasslemoff's replays from 1.00 and 1.16 should show differences if the method works at all. 👍 Will check tomorrow!

 

[DolHaus]

Is there a way to go back a version?

According to this-
https://www.gtplanet.net/forum/threads/1-16-update-download-onto-memory-stick.324798/
It might be...

Just depends on what lengths you are willing to go to

 

[Polsixe]

I don't think Polyphony could change anything about the physics of the game without having to block the use of replays before the update.

GT replays only record the acceleration, braking and steering, everything else is reconstructed by those entries every time you see them.

If you change the parameters of physics engine, replays recorded would result in errors. I.e. that rushed near the wall now would be a direct crash.

Excuse the engrish.

If the replay is just a rendering of the the car in x y z co-ordinates then the physics engine doesn't come into play. The physics only matter in creating the path the car follows which is then saved. Things like rolling, pitching, yawning, drifting are all factored in only during the initial run.

 

[Griffith500]

There are likely many differences between 1.00 and 1.16, but it'll be interesting anyway. I might have a poke at them when I get chance.
I have a PS3 I can possibly take the time to bump up to 1.15 and repeat these tests with. That may be several weeks, as I have grander plans for it.

Let me try to formulate your point, as i get it, before we talk at cross-purposes.

First the facts:

In your post the first diagrams show the FFT of yaw-rate and steering-angle of hasslemoffs 1.15/1.16 laps.

When looking at the Lap3 vs. Lap1 and Lap2 vs. Lap1 yaw-rate, i can see there is a difference between 1.15 and 1.16 in the 1-3 Hz range peaking at 2Hz with a magnitude of ~10dB. The steering-input in the Lap3 vs. Lap1 comparison shows roughly the same difference at the same range, in the Lap2 vs. Lap1 the situation is more complicated.

No, it doesn't. Looking at the FFTs, it's clear that laps 1 and 2 are similar in steering, but laps 2 and 3 are similar in yaw rate. That about sums it up, for me, but here's another way of looking at it:

The green trace is the difference between laps 2 and 1; the blue trace is the difference between laps 2 and 3.

In the first image we have the raw difference in yaw rate at a given track position, then the cumulative sum of that difference, then the cumulative sum of each yaw rate trace separately (effectively the car's heading).
In the second image we have the raw difference in steering angle at a given track position, then the cumulative sum of those differences.

As I said, the steering is similar between 1 and 2 (green), but the yaw rate is similar between 2 and 3 (blue). This contradicts the idea that a change in steering results in a comparable change in yaw across the 1.15 -> 1.16 transition, just as @Ridox2JZGTE 's laps did.

When looking at the Lap2 vs. Lap3 (both 1.16 laps) yaw-rate, the difference magnitude is only ~5dB and in the 0.7-1Hz range. The steering-angle difference is ~10dB between 1-2.5 Hz.

Why are you using hand-picked peak values? You should be focusing on trends. What about the inertial band?

So if you are saying the yaw-rate difference points to a physics change, i would argue that the steering-input differences can explain the yaw-rate differences just fine, being of the same magnitude. I see the yaw-rate as being caused by the steering-input and the GT6 physics, so any change in the output without corresponding change in the input would be evidence of change in the physics.

I'm discussing a change in the player / game feedback loop. The yaw rate is influenced by things other than steering, and steering is not even a direct yaw control - it's a physics input that twists the tyre contact patch to create extra drag and effect a yaw through all the mechanical connections.

If you divide the FFT of the yaw rate by that of the steering input (effectively yielding the hypothetical "transfer function" for the steering -> yaw path in the physics engine), you'd see there is far from a 1:1 correspondence. That'd be even more true of their derivatives.

But if you are saying the yaw-rate difference and steering-input difference have to be taken together to come to the physics change conclusion, i can not follow this argumentation, because i think the steering-input differences can be attributed to the driver alone. When trying to get evidence for physics change, i think one has to eliminate the effect of the driver.

Once again, I'm not arguing a physics change. I'm arguing a point about driving feel, in which the player is the only thing that matters. About the only thing uncertain is whether the differences are part of a natural variation in driver input (which given the statistical nature of the change, I'd doubt) or a (even sub-conscious) response to a change in feedback.

It is unlikely we will ever truly ascertain what caused the change in feedback. To me, it feels like a change in rotational inertia, to others it's "weight transfer", to some others, it's turn-in and braking, others feel more "response" from the car in general. All of these point to some change in the feedback of the dynamic (as opposed to steady-state) character of the car.

EDIT: the feedback is obviously the visual recognition of yaw rate (which is primarily how I drive), but what I mean is we don't know what's affected that process either in the game, or what extra signals are creating a change in perception in our brains.

The throttle input should not be taken into consideration here, i think the frequency analysis of throttle input is meaningless, because it is just a random almost square signal. Nonetheless it has a Fourier Transform, but to give any meaning to the frequency magnitudes is misleading. I just wanted to show that with my diagram.

The throttle input is far from random, or else the car would not stay on the track. The throttle input affects yaw, also, otherwise power-oversteer would not be possible.

 

[SVPSkins]

race on London and Sierra alot and ive noticed my Mini can out corner faster cars now on Sierra, as before i struggled and ive noticed more in my camaro it likes to over steer more then it used to. So yes i think the
physics have changed

 

[Kryption]

I tested the new physics in the s14, skyline gtr r33 '97, and mr2. Mr2 is 450 pp Sh tires, s14 is 497 pp CH tires and is built for drift, Skyline gt r33 is 500pp with SH tires tuned for nurb, I haven't touched any of the tuning settings. I use DS3

First off,
I drive my mr2 on sierra only, sometimes nurb but not recently. The handling seems a lot different, it's a lot easier to keep control of it going from a downhill to uphill hairpin. In general the mr2 seems to have a lot better reaction for controller at least, I don't have a steering wheel. there's a corner on sierra my mr2 always wants to slide at, and since the patch it hasn't slid there at all, no changes in driving style or tuning.
also, it seems my mr2 can accelerate faster?

S14: I drift mainly on trial mountain and suzuka full circuit. Before patch I could keep it at 4th gear redlined and not spin out on the first corners of suzuka, now I spin out almost instantly as soon as it reaches redline on the apex. I have to control throttle more which I don't mind, makes it more realistic for ds3 at least. Trial mountain the first 3 corners I can keep it redlined at 4th gear and follows the line perfectly, since patch I have to downshift to 3rd at the apex on the last corner before that little tunnel at the beginning now. The weight shift feels a lot quicker than before.

R33: biggest difference I noticed is at karasoul on nurb, I always take it at 82 km/h, now I can take it at 84 km/h, the corners seem a lot easier to navigate while braking less, I can keep throttle down for longer, and brake later. Unless that's just me improving my line. I can slide into the entries of the corner now, and it straightens out without me doing anything except holding down the throttle. Before I would have to counter steer slightly to straighten out.

I can notice the changes, it's subtle for the most part, but it all plays a part. I hope this might have helped a bit, I still haven't done midfield yet haha.

 

[Ridox2JZGTE]

Another car that was found to be better than before in 1.16 :

Have you drove this wild horse after the 1.16 update?

I did and I was astonished. Just try it out. 👍

Won at Ascari 10-Laps.
​

No aids nor ABS.
Suspension: base
LSD: base
BB: 0/0

​

It was one of the wild GT4 replica I built last year ( 39/61 distribution - real life spec ). Can't wait to drive it again on 1.16


If anyone is still in pre 1.16 ( any version before 1.16 ), please try this Evora GT4, and compare it to 1.16

LOTUS Motorsport/Cosworth Evora '09 GT4 Replica
Tuned to replicate Lotus Evora GT4
Sports Soft​

View attachment 177244​

CAR : LOTUS Evora '09
Tire : Sports Soft

Specs
Horsepower: 355 HP / 360 PS at 6800 RPM -max engine rev 7200rpm as in real life.
Torque : 309.1 ft-lb at 4800 RPM
Power Limiter at : 100%
Weight: 1200 kg
Ballast : 53 kg
Ballast Position : 47
Weight Distribution : 39 / 61 as in real life
Performance Points: 505


GT AUTO
OIL change
Improve Body Rigidity (INSTALLED IN THIS BUILD ) - MANDATORY - as GT4 cars are seam welded and has weld in cage.
Aero Kits Type A
Custom Rear Wing :
Wing Mount Standard Type A
Wing Large Type A
Winglets Large Type C
Height +7 and Width +2
Wheels : Standard Size - GT Ab12 in Black or Bronze
Car Paint : New Sherwood Toning / Dark Green / White Factory Delivered Color


Tuning Parts Installed :
Racing Exhaust
Isometric Exhaust Manifold
Intake Tuning
Twin Plate Clutch
Fully Customizable Dog Clutch Transmission
Adjustable LSD
Fully Customizable Suspension
Racing Brakes Kit
Weight Reduction Stage 2
Window Weight Reduction



Suspension - EIBACH ERS Springs (650/900) and OHLINS TTX 4 Way Damper Kit
Front, Rear

Ride Height: 75 100 ( similar rake to real life Evora GT4 )
Spring Rate: 11.60 16.07
Dampers (Compression): 6 7
Dampers (Extension): 8 8
Anti-Roll Bars: 3 4
Camber Angle: 1.2 2.5
Toe Angle: -0.08 0.16


Suspension - EIBACH ERS Springs (650/850) and OHLINS TTX 4 Way Damper Kit - Stability Enhanced Update
Front, Rear

Ride Height: 75 100 ( similar rake to real life Evora GT4 )
Spring Rate: 11.60 15.18
Dampers (Compression): 6 7
Dampers (Extension): 8 6
Anti-Roll Bars: 6 5
Camber Angle: 1.2 2.5
Toe Angle: -0.08 0.16


Suspension - EIBACH ERS Springs (650/850) and OHLINS TTX 4 Way Damper Kit - Stability Enhanced Update + No Rake / Equal Ride Height
Front, Rear

Ride Height: 75 75
Spring Rate: 11.60 15.18
Dampers (Compression): 6 7
Dampers (Extension): 7 5
Anti-Roll Bars: 5 4
Camber Angle: 1.2 2.5
Toe Angle: -0.08 0.16


Suspension - EIBACH ERS Springs (500/650) and OHLINS TTX 4 Way Damper Kit - Stability Enhanced Update + No Rake / Equal Ride Height - Specialized for Bumpy Tracks
Front, Rear

Ride Height: 85 85
Spring Rate: 8.93 11.60
Dampers (Compression): 6 7
Dampers (Extension): 7 5
Anti-Roll Bars: 6 4
Camber Angle: 1.2 2.5
Toe Angle: -0.08 0.16

View attachment 178170​

LSD - 2 Way High Preload and Lock LSD
Initial Torque : 27
Acceleration Sensitivity: 48
Braking Sensitivity: 48

LSD - 2 Way High Preload and Medium Lock LSD - Alternate Setup
Initial Torque : 27
Acceleration Sensitivity: 27
Braking Sensitivity: 48


DOG CLUTCH TRANSMISSION - Xtrac 6 Speed Sequential 426 Gearbox
Install all power parts
Set Default
Set Final to 5.100
Set Auto Max Speed at 180kmh / 112mph
Adjust each gear :
1st 3.000
2nd 2.067
3rd 1.625
4th 1.316
5th 1.111
6th 0.960
Set Final : 3.286 or shorter : 3.769 for better acceleration/less top speed.


AERO:
REAR : 20


Brake Balance:
4/4 ( personal BB) or for ABS 0 wheel : 4/4, for ABS 1 - feel free to use your preferred brake balance. I recommend to run 5/5 for ABS 1 or 1 click higher at the rear.

Recommended setting for DS3 user :

Steering sensitivity at +1 or +2, all aids off, except ABS 1 ( if not comfortable with ABS 0 ) with 4/4 brake balance as starting point.

​

Notes :
Lotus Evora has great track record in the GT4 class, the car was built in collaboration with Cosworth, who provide the powerplant, tuned to output 360PS. The car weighs 1200kg with 39/61 weight distribution. The MR layout provides excellent traction and good handling balance.

Suspension uses Eibach race springs and Ohlins TTX 4 way damper kit ( optional ), the base model comes with 2 way damper instead. I tuned the springs based on the real car motion ratio/whee rate, settled for 650 lb/in / 900 lb/in spring rate. I also made the damper stiffer as the car has considerable rake to mimic real life ride height setup. ARB, camber and toe are set to balance the springs and the stiff damper.

Gearing uses Xtrac 426 6 speed sequential with sintered dual plate clutch, so I fitted twin plate clutch and set the gearing using ratios from Xtrac 426 catalogue. The gearing provides good acceleration and top speed on the 3.286 final ( there are 2 final ratio option from Xtrac 426 that can be used ). A higher final 3.769 is used when the car is driven on the track with lower top speed required.

LSD in this replica is quite unique, I used 2 way setup with high initial + high locking rate, and it works well for the Evora GT4. This still allows for freedom when trail braking with the rear rotates nicely.

I tuned and tested the car mainly at Silverstone GP, aiming to replicate real life lap time. Target was 2:14-2:15s. During testing, the Evora managed 2:13s on 1.08 ( camber still broken ).

Please leave feedback as I haven't tested the car on 1.09


UPDATE 1.09 : Credits to @cbarbosa, added another LSD setup with lower accel value at 27 for more manageable corner exit.

UPDATE 1.11 : Added additional suspension setup built for enhanced stability, credits to @danbojte for the rear spring rate test ( 15.18 ) and @ColinH96 for feedback on my tweak suggestions. This updated suspension offers easier handling and more refined on the edge behavior.
Tested at Silverstone GP, managed to beat the old time at 2:11.4xxx Crazy fast for 355HP/1200kg car. The Evora also tested at Nordschleife, handles okay there, should be great for 500PP racing.

UPDATE 1.11 V2 : Added additional suspension setup using no rake / equal ride height. This will add eagerness to rotate due to the reversed ride height effect bug.
I have tweaked the damper extension front and rear to accommodate the changes, as well as readjusted the ARB to control the body roll and the added rotation.

I also have added softer spring rate setup with higher ride height ( by 10mm ) for bumpy tracks like Bathurst and Nordschleife. The car will still need smooth steering, but the added freedom to rotate mid corner will help the car easier to negotiate tight corners. The added body roll will be helpful as cushion / early warning before traction is lost, it is also highly recommended to use the original LSD setup ( 20/48/48 ) to maintain optimum traction. Avoid curbs and grass is paramount

ENJOY

View attachment 178173​

 

[tarnheld]

Mmmh, where to start...

In the first image we have the raw difference in yaw rate at a given track position, then the cumulative sum of that difference, then the cumulative sum of each yaw rate trace separately (effectively the car's heading).
In the second image we have the raw difference in steering angle at a given track position, then the cumulative sum of those differences.

Nice diagrams, how did you get the difference between two replays? Did you use the variance_dist function in maths? Seems to me the only way in i2 get a plot of differences between two logs.

Anyways, the cumulative yaw-rate (Int_Yaw) better be not too different between laps, as that would indicate the driver would not stay on track. As you stated correctly it's the heading of the car in degrees plotted over the track position. Maybe a drift lap would show differences here. As they stand, the diagrams show how much the steering input can differ between two laps that show the same lap time -- in support of @Johnnypenso's remark about lap variances above.

Why are you using hand-picked peak values? You should be focusing on trends. What about the inertial band?

I just took the parts that have the most difference. Can you describe what you mean with 'inertial band'?

I'm discussing a change in the player / game feedback loop

Once again, I'm not arguing a physics change. I'm arguing a point about driving feel, in which the player is the only thing that matters.

Here we part ways, as my intention is to find solid evidence for a physics change. That means i'm trying to eliminate player input as much as possible.

It is unlikely we will ever truly ascertain what caused the change in feedback.

That might be true, but we can find evidence that GT6 physics have changed somehow. What exactly was changed and what the consequences are comes later for me.

So just a quick glimpse on what i found in @hasslemoff's laps so far:

This is start of the X-Bow Street '12 @ Stowe race with the stick pressed to the left and full throttle, black is 1.00 color is 1.16. As you see the steering input is not exactly equal, there are minute differences at the start and it starts to diverge when the vehicle spins.

The lateral G Force is different in both versions, and the oscillations at the end could point to a difference in physics. But the difference is small and the oscillations at the end could be just a fluke. To rule out the small input differences, i would like to ask @hasslemoff to redo the X-Bow test on both versions 2 or 3 times, just to be sure how big the differences are for multiple runs for a single version?

Thanks for the testing @hasslemoff!

EDIT: corrected some misspellings

 

[super_gt]

The physics between 1.00 and 1.15 is different.
From where I know that?
Few months ago I deleted all updates and I tried 1.00 physics.

 

[Furinkazen]

The physics between 1.00 and 1.15 is different.
From where I know that?
Few months ago I deleted all updates and I tried 1.00 physics.

Show us your evidence then.

 

[super_gt]

Show us your evidence then.

This is the only evidence that I have.
https://www.gtplanet.net/forum/threads/gran-turismo-physics-poll.312945/page-7#post-10094142
I have no replays.
If you do not believe me delete all updates and see for yourself.

 

[Griffith500]

Mmmh, were to start...

Nice diagrams, how did you get the difference between two replays? Did you use the variance_dist function in maths? Seems to me the only way in i2 get a plot of differences between two logs.

Yes, I used the variance functions.

Anyways, the cumulative yaw-rate (Int_Yaw) better be not too different between laps, as that would indicate the driver would not stay on track. As you stated correctly it's the heading of the car in degrees plotted over the track position. Maybe a drift lap would show differences here. As they stand, the diagrams show how much the steering input can differ between two laps that show the same lap time -- in support of @Johnnypenso's remark about lap variances above.

They also show that steering and yaw rate are not intrinsically linked, as you suggest. Note, from previous posts, the absolute equivalency between the yaw rate and yaw acceleration, implying the yaw is a second-order system (with respect to time) i.e. inertial. What that says is there are no first-order inputs of considerable magnitude. So any steering input must be translated through the physics model.

I included the "heading" for that very reason, in case it seemed too "random".

I just took the parts that have the most difference. Can you describe what you mean with 'inertial band'?

The part where the steering and yaw curves diverge rapidly. About 3 Hz upwards. Note the massive sink on both cars, which are effectively kinetic energy dispersers. That makes operating the car in that region a tricky affair, and yet both drivers show they did so more often in 1.16 than in 1.15. It's only a couple of laps, but I'd still like to see someone consciously reproduce these changes in control whilst matching lap times.

Here we part ways, as my intention is to find solid evidence for a physics change. That means i'm trying to eliminate player input as much as possible.

That might be true, but we can find evidence that GT6 physics have changed somehow. What exactly was changed and what the consequences are comes later for me.

Do you think it's possible for there to be a difference in driving feel without a change in physics or FFB? (People felt a change with DS3, myself included).

So just a quick glimpse on what i found in @hasslemoff's laps so far:

View attachment 323465

This is start of the X-Bow Street '12 @ Stowe race with the stick pressed to the left and full throttle, black is 1.00 color is 1.16. As you see the steering input is not exactly equal, there are minute differences at the start and it starts to diverge when the vehicle spins.

The lateral G Force is different in both versions, and the oscillations at the end could point to a difference in physics. But the difference is small and the oscillations at the end could be just a fluke. To rule out the small input differences, i would like to ask @hasslemoff to redo the X-Bow test on both versions 2 or 3 of times, just to be sure how big the differences are for multiple runs for a version alone?

Thanks for the testing @hasslemoff!

The range 113 - 125 metres is very telling. It looks like the car continues to accelerate rotationally (lat. g and yaw rate are equivalent at a given speed during "grip" conditions) in 1.16 when it stays constant in 1.00. Either you could attribute that to the tyres slipping more at first, or to a change in the inertial model at some point between the two builds. The overshoot at 125 metres points to inertia again.

The use of the DS3 is a problem, because it clearly self-regulates its input ("steering assist" ) - I wonder if the steering sensitivity was the same in either case. Using a wheel "locked" at a sensible turning angle might be better.

 

[tarnheld]

They also show that steering and yaw rate are not intrinsically linked, as you suggest. Note, from previous posts, the absolute equivalency between the yaw rate and yaw acceleration, implying the yaw is a second-order system (with respect to time) i.e. inertial. What that says is there are no first-order inputs of considerable magnitude. So any steering input must be translated through the physics model.

I am trying to be totally ignorant of the nature of the link between steering input and yaw-rate/lateral G-force. I only see one as the input of a black box, and the other as the output. What is inside the box is unknown. When trying to find evidence that there was a change to the inside of the black box, i would fix the input and look for change in the output. If there is change in the output, there has to be a change in the box, as the input is fixed.

But you are right that the black box has to be a simulation of vehicle dynamics, and the output is some second derivative of position/orientation. But i try to ignore that for the discussion of a change in the physics, as it isn't necessary to know that to show a change.

Do you think it's possible for there to be a difference in driving feel without a change in physics or FFB? (People felt a change with DS3, myself included).

Sure, there could be many reasons for that: ranging from

changes in rendering causing more cycles available for physics computation, resulting in more accurate results​

to

driver has not slept enough and is slower with inputs, or slept well and is quicker, or has mastered trail breaking recently
​

By eliminating the driver input, only the black box remains.

The use of the DS3 is a problem, because it clearly self-regulates its input ("steering assist" ) - I wonder if the steering sensitivity was the same in either case. Using a wheel "locked" at a sensible turning angle might be better.

That's a good point, the mapping from DS3 stick to steering angle could be changed in versions too. 👍 Maybe the tests should be done again with a wheel at full lock. But as the car would spin very fast, the wheel angle at full lock should be restricted.

 

[Griffith500]

I am trying to be totally ignorant of the nature of the link between steering input and yaw-rate/lateral G-force. I only see one as the input of a black box, and the other as the output. What is inside the box is unknown. When trying to find evidence that there was a change to the inside of the black box, i would fix the input and look for change in the output. If there is change in the output, there has to be a change in the box, as the input is fixed.

But you are right that the black box has to be a simulation of vehicle dynamics, and the output is some second derivative of position/orientation. But i try to ignore that for the discussion of a change in the physics, as it isn't necessary to know that to show a change.

I agree absolutely. Where we seem to differ is in the boundary of the black box. I include the feedback, because in practice the system as a whole morphs the input / output space according to that feedback step. The physics engine is just one part of the chain: input -> physics -> output -> "magic" -> input.

I only do so as a point of comparison, and as a reality check: we don't interact with the physics, we interact with the whole system. Broad changes in control input can be taken to be indicative of a change within that system somewhere - with the same driver and same "attitude" or "approach", I don't think it's reasonable to expect such large changes in input, especially in the statistical manner it exhibits itself.


I was actually wrong in my reading of the FFTs of the derivatives; the equivalency arises from the mathematical definition of the Fourier Transform (sine and cosine representation of a signal), and the properties of those trig functions in respect of differentiation and integration: the sin -> cos -> -sin -> -cos -> sin cycle. So since the phase information isn't displayed, the magnitudes should be the same (scaled only by frequency).
However, the yaw is very noisy at "jounce" (third derivative of heading), and the steering is noisy at acceleration (second derivative of steered angle). So we reach the same conclusion, sort of.

The steering speed FFT is different because I manipulated the sign of the signal over a zero-crossing, doubling the apparent frequency of any input that crosses the steering centre. This is still a useful metric, because it highlights differences in the zero-crossings, which should be clear to be of interest for steering.

Sure, there could be many reasons for that: ranging from

changes in rendering causing more cycles available for physics computation, resulting in more accurate results​

to

driver has not slept enough and is slower with inputs, or slept well and is quicker, or has mastered trail breaking recently
​

By eliminating the driver input, only the black box remains.

Personally, I interpret an increase in physics rate to be a change in physics. Much more phase information is preserved, and so the dynamic behaviour is more defined up to a higher frequency. That would be evident in our "step change" tests @hasslemoff has started. You've also got to wonder just how much information can get through filtered at less than 60 Hz, but it's possible.

The driver difference is important, I don't disagree. What I am skeptical of is whether the differences exhibited could reasonably be replicated consciously. It's also interesting that for both drivers, the differences are broadly very similar.

That's a good point, the mapping from DS3 stick to steering angle could be changed in versions too. 👍 Maybe the tests should be done again with a wheel at full lock. But as the car would spin very fast, the wheel angle at full lock should be restricted.

The output itself should give a rough idea of what angle to use: around 30 - 60 degrees minimum; 90 is easy to "see" on the wheel, but looks to be too much - maybe halve it to 45? Of course, that's assuming the in-game steering angle corresponds to the hardware angle, which can be checked easily enough.

 

[tarnheld]

Here is the big picture of the X-Bow Street '12 data from 1.00 (black) and 1.16 (color):

At ~100m the start line is crossed, at 174m something is hit in 1.16, at 188m something is hit in 1.00 (you see that the Vehicle Speed drops from 140km/h to 50km/h in an instant) and at 200m the 1.16 vehicle stops.

The Steered Angle first goes up to 75° (DS3 stick fully left) but then goes up even higher and diverges between versions. I don't know the reason for that, might be some DS3 stick to angle mapping. But it violates the 'fixed input' condition, so the diagrams should be interpreted with care.

The throttle input of the 1.16 lap has a drop between 155m and 159m. Could be (auto?)-shifting, Maybe @hasslemoff knows more.

At the bottom is Heading, that is the same as the Int_Yaw in @Griffith500's post above. It's not calibrated to zero, and starts at -184°, but you see the vehicle changes heading by 25° in the first 60m, and then diverges even more between versions. Maybe the car entered grass here and got slowed down more in 1.16 than in 1.00?

The lateral G force is considerably different, but i think it is too early to declare it caused by physics. We need the extra check of 2-3 extra laps with this procedure for each version.

The other replays show essentially the same picture, but have more input differences. So i leave it at that for the moment.

EDIT:

just one more diagram, almost all logged data, the first 50m:

This diagram contains almost all logged raw data, missing is throttle (100%) and brake (0%), gear (3rd) and engine RPM (minute differences).

You see that the vehicle speed is different between 1.00 and 1.16, and especially the Wheel speeds diverge in an interesting way: whereas the front wheel speeds stay the same from 115-140m in 1.00, they are slowed by 5km/h at 140m in 1.16. Different tire / suspension physics?

I think we all agree that the physics was changed from 1.00 to 1.16 (IIRC the big change was at 1.09), but it is a good test of the method.

 

[tarnheld]

Sorry for the double post, but this is interesting: seems in 1.00 they do no throttle auto-blip then shifting. You see the shifting throttle auto-blip while shifting into 4th gear at 155m in the diagram above, but only for 1.16. In 1.00 the shift is at the same position (is logged, but not shown), but there is no auto-blip.

 

[Kryption]

Here is the big picture of the X-Bow Street '12 data from 1.00 (black) and 1.16 (color):

View attachment 323536

At ~100m the start line is crossed, at 174m something is hit in 1.16, at 188m something is hit in 1.00 (you see that the Vehicle Speed drops from 140km/h to 50km/h in an instant) and at 200m the 1.16 vehicle stops.

The Steered Angle first goes up to 75° (DS3 stick fully left) but then goes up even higher and diverges between versions. I don't know the reason for that, might be some DS3 stick to angle mapping. But it violates the 'fixed input' condition, so the diagrams should be interpreted with care.

The throttle input of the 1.16 lap has a drop between 155m and 159m. Could be (auto?)-shifting, Maybe @hasslemoff knows more.

At the bottom is Heading, that is the same as the Int_Yaw in @Griffith500's post above. It's not calibrated to zero, and starts at -184°, but you see the vehicle changes heading by 25° in the first 60m, and then diverges even more between versions. Maybe the car entered grass here and got slowed down more in 1.16 than in 1.00?

The lateral G force is considerably different, but i think it is too early to declare it caused by physics. We need the extra check of 2-3 extra laps with this procedure for each version.

The other replays show essentially the same picture, but have more input differences. So i leave it at that for the moment.

EDIT:

just one more diagram, almost all logged data, the first 50m:
View attachment 323586
This diagram contains almost all logged raw data, missing is throttle (100%) and brake (0%), gear (3rd) and engine RPM (minute differences).

You see that the vehicle speed is different between 1.00 and 1.16, and especially the Wheel speeds diverge in an interesting way: whereas the front wheel speeds stay the same from 115-140m in 1.00, they are slowed by 5km/h at 140m in 1.16. Different tire / suspension physics?

I think we all agree that the physics was changed from 1.00 to 1.16 (IIRC the big change was at 1.09), but it is a good test of the method.

This is why I love this community, nothing goes overlooked , great diagnostics and explanations you have. Helps me understand this a lot better. +1