I just want to echo
@HugoTwoWheels sentiments about GT7’s FFB implementation. There is without a doubt
software clipping occurring within the physics FFB calculations. The fact that v1.15 improved the dynamic range feedback seems to have occurred purely by accident.
It appears PD use a combination of physics based FFB (eg tyre cornering loads) and ‘canned’ effects (eg understeer judder) that are calculated based on simpler conditional parameters (like speed vs steering angle vs vehicle trajectory etc) and combined with the physics FFB data to form the final output signal that’s sent to the wheel via the single “Force” output channel.
The following is a simplified example for illustration purposes only. The actual values and scaling is not based on any recorded data from GT7 itself, but rather a visual representation to assist with describing the software clipping theory.
View attachment 1161169
Above is a basic example of physics based FFB steering loads, combined with other FFB canned effects to calculate the final full FFB signal. This signal can be scaled to achieve varying levels of maximum torque or weight to the users preference. All of the information received from the physics based tyre load is used and accounted for within the final full FFB output.
View attachment 1161170
This example illustrates where I suspect the heart of the issue lies. For whatever reason, physics-based loads are seemingly subjected to some sort of soft limit, with any force values exceeding this limit being capped at this maximum.
This soft limit seems to be achieved
remarkably quickly in GT7. It’s reasonably easy to perform a practical demonstration using a stiffly sprung car. Take Grp.4 GTR out on a track with long corners like Yamagiwa (Kyoto Driving Park). Gather some speed and gently weave the car left and right to feel the tyres and steering start to load up. You should notice the load increase with more aggressive weaving, and even feel some bumps and texture of the road surface through the wheel. But at some point, (and very early on) you will generate enough lateral load to hit a soft limit for the FFB to stop increasing in load. Winding in more lock or load won’t make any difference to the amount of feedback. You’ll just feel a single constant force that doesn’t change. The road bumps and texture you could previously feel at lower lateral loads now disappear completely, as you’re already exceeding the maximum soft limit load permitted by the FFB code.
However, canned effects such as the understeer effect do still activate, suggesting they’re not subjected to the same soft limit.
View attachment 1161171
I suspect that the temporary measure introduced by PD scaled the physics based FFB calculation to prevent the aggressive oscillation seen on a handful of extreme cars which inadvertently (at least somewhat) dropped the signal below the soft limit clipping bug. This resulted in a final FFB output signal that utilised all (or at least more) of the dynamic range available from the physics based FFB signal resulting in a more complete (albeit significantly weaker overall) FFB signal.
View attachment 1161172
The final complete FFB signal is scaled using the game’s max. torque setting (1 – 10) at the user’s preference. Due to the temporary scaling ine V1.15 of the physics based FFB, the end result was much lower overall FFB values across the board. The weaker signal could be scaled up using max. torque 10 in game, but this was still only achieving similar loads to max. setting 3 in v1.13 and v1.16.
It also seemed that canned effects such as the understeer judder became more aggressive too, probably due to the fact they were effectively scaled up to 10, while the physics based were still operating at level 3 strength. I suspect individual vehicle parameters such as damper values are also included as “other” effects and therefore also unaffected by the measures introduced by PD, meaning they too were effectively scaled up to 10. This would’ve increased the weight of the wheel relatively to the feedback generated by lateral physics-based loads, which would explain why many people found reducing the additional Fanatec filters such as damper and natural friction to be favourable with v1.15.
I could be completely off the mark with the methodology of how PD have implemented FFB but I’m not sure how else to explain the FFB saga. The soft limit may even be deliberate and simply a poorly conceived fail-safe against excessive FFB.
At any rate, there’s no doubt soft clipping is happening and we’re completely at the mercy of PD to fix it.
