[James Ross]

Genma Saotome, on 23 October 2014 - 09:20 AM, said:

Ok... so this problem could have been avoided by either raising the higher texture a tiny bit or merging the two textures together, right?

How close together is too close?

Yes, it's caused by the two textures being layered so close and the particular solution OR takes to applying z-bias.

If I am reading the code right, it moves items forwards (towards the camera) by up to 1mm (technically very, very slightly more due to near-plane scaling). Take this as a minimum separation needed, however, since the accuracy of the depth buffer depends on the viewing distance so it's not always obvious what the closest something can be without error.

[copperpen]

James, if you look at my post #13 on page 2, you will see that I have tried the same location with single textures, by renaming so one is not used. Each texture on its own shows the same behaviour.

[James Ross]

copperpen, on 24 October 2014 - 01:48 AM, said:

James, if you look at my post #13 on page 2, you will see that I have tried the same location with single textures, by renaming so one is not used. Each texture on its own shows the same behaviour.

Editing posts later causes people to miss information, FWIW.

I looked at the track myself and it definitely has two layers very close together supporting my original theory. I cannot guess what you're seeing - I could not make sense of your description - so please provide some pictures.

[copperpen]

Four pictures to show the problem, three from OR and an MSTS shot for comparison.

Attached thumbnail(s)

• 
• 
• 
• 

[Genma Saotome]

What James said.

Perhaps changing the offset to be more than 1mm will help.

What's with that guy on top of the locomotive?

[copperpen]

The guy on the locomotive is an ongoing Freightanim problem yet to be solved. As for moving anything anywhere, these are not my shapes, and do display normally in MSTS.

[James Ross]

copperpen, on 25 October 2014 - 06:59 AM, said:

Four pictures to show the problem, three from OR and an MSTS shot for comparison.

This is still showing the same problem I described because you're not actually removing either of the layers, just its texture - so it still gets rendered but in "I don't have a texture grey". The options are to fix UK Finescale track to not have two layers so close together (either move them apart or use just one layer) or we can attempt to reduce the distance we offset for z-bias (but that will definitely have a limit based upon depth buffer accuracy).

[copperpen]

Therefore my original statement is correct, OR cannot properly run routes that use the UKFS track system.

[James Ross]

copperpen, on 27 October 2014 - 01:53 AM, said:

Therefore my original statement is correct, OR cannot properly run routes that use the UKFS track system.

In your opinion, perhaps. I strongly disagree with you here and find this statement distinctly unhelpful.

[dennisat]

copperpen, on 27 October 2014 - 01:53 AM, said:

Therefore my original statement is correct, OR cannot properly run routes that use the UKFS track system.

I think this is a bit sweeping - other routes that use Finescale, Dorset Coast, Somerset and Dorset, Great Eastern, to name a few, don't display the effect to such an extent. There is a slight flicker that you notice but you have to fiddle around with the cameras to find the effect that is displayed so obviously in SCP. It is also the case that not all track sections show this effect in SCP.

Having said that, London and South East displays the track just like SCP. The track must have been built slightly differently in some routes. I believe I've read somewhere that some routes make use of customised versions of Finescale. Certainly, Dorset Coast and Greater Eastern use concrete sleeper versions and I think the effect is hardly there at all whereas the wooden sleeper versions show the effect to a much greater extent.


Dennis

[copperpen]

No need to fiddle with the cameras in SCP, you get the flicker effect at any angle' I only chose that particular one to show the effect clearly. I accept that not all UKFS shapes have that problem, unfortunately it is present on several, if not more.

Dorset Coast was built using v3.1.6, it would therefore look like both SCP and the S&D were built using a different version. I am therefore going to remove the SCP Global Shapes folder and drop in the Dorset coast one and see what happens.

[copperpen]

Following on, it is definitely the wooden sleeper set of shapes that have the problem, whatever version of the shapes is used. Therefore one cannot apply a blanket solution to the problem. As it looks like OR will not be able to run these UK finescale wooden sleeper routes without flicker present, unlike MSTS, there is no other choice for them than to continue with MSTS.

[dforrest]

Modified wooden sleeper track textures are available at UKTS (File ID 25768). They appear to partly resolve the SCP problem.

[James Ross]

For anyone who's affected by this bug, they should click "Does this bug affect you?" on the bug report. The bug reports are the only way we can reliably gauge interest in a problem and prioritise fixing it.

[copperpen]

dforrest, on 27 October 2014 - 06:15 AM, said:

Modified wooden sleeper track textures are available at UKTS (File ID 25768). They appear to partly resolve the SCP problem.

Thanks for that. Just dumped them into SCP and texture "cutting" shown in my earlier screen shots is gone, and the flicker effect is much reduced. Still there, but mainly confined to the edges of the shapes now, with a hint of moire type artefact well ahead between the rails. I can live with that. I don't know how I managed to miss seeing that file.