investigating Skhval performance degraded starting with DCS 2.8

[okopanja]

The 2.8 version did bring the new green color of the screen. along with it I have noticed the performance degradation with SKHVAL which results in much more difficult spotting. I have noticed this while playing at BS server, on the first session where I tried to use shkval.

I have made screenshots to demonstrate in following way:

1. I selected hot Ka-50 on ground (for stability and reproducibility, if needed you may take of or use the one that is already in flight)
2. I made no changes to the view, used just what the game displays by default
3. Located the same area
4. Used Brightness & Contrast to get the best possible picture
5. I captured full resolution pictures (3440x1440) and saved them as PNG (lossless format), but could not upload them (please tell me how to deliver them 6,28MB vs 7,71MB), so I cropped the image.
6. Compare the images, the difference is quite obvious.
7. Notice the subtle but important difference: follow the left vertical line of the caging box slightly up. In old Shkval you will notice a Stinger manpad, it is not much visible but it is there! In new Shkval this is nowhere to be seen no matter what I do with Brightness & Contrast.

Used versions for comparison:

- 2.7.18.30865

- 2.8.0.32235.1

Attached are:

- mission file

- trk files

Ka-50 - Skhval test.miz 2.7.18.30865.trk 2.8.0.32235.1.trk

[okopanja]

Adding back-reference link containing all differences people notices in free text:

 

[Hobel]

Hmm. Interesting I'll have to take a closer look at that too. 

I played a mission on Syria yesterday and had no trouble at all with the screen.  It felt the same as always only this time the image was green. 

 

But your pictures again show a difference.  

[AeriaGloria]

I feel like during the day it is the same. But as things get close to dusk or nighttime the new green tint makes it much harder. I have been able to have an easier time seeing things with my eyeballs at night 

[NightMan]

I just wanted to add a small detail:

- Each time a target is destroyed the Shkval jumps, usually a bit down, but I've seen it jump to the side also.

 

I also wanted to ask something:

- Does the increase in the delay before launch to 2 seconds reflect what it is in real life? It makes it much more difficult to achive what I call "fast firing" without having to remain in the same position for a long time.

[AeriaGloria]

2 hours ago, NightMan said:

I just wanted to add a small detail:

- Each time a target is destroyed the Shkval jumps, usually a bit down, but I've seen it jump to the side also.

 

I also wanted to ask something:

- Does the increase in the delay before launch to 2 seconds reflect what it is in real life? It makes it much more difficult to achive what I call "fast firing" without having to remain in the same position for a long time.

ED said the 2 second delay was in pilot notes. I do wonder if perhaps it’s really 1 second and pilot notes say 2 for a margin of error. But Considering how hard they make it to fire in the first place with launch angle constraints I wouldn’t be surprised if Kamov really though “2 seconds to make sure pilot is REALLY sure”

[okopanja]

@NightManand @AeriaGloriathis bug report is about the quality of picture we are getting, let try to keep it sharp on subject.

My second comment contains the link toward general commenting session for ka-50. You will find some answers there.

[NineLine]

22 hours ago, Hobel said:

Hmm. Interesting I'll have to take a closer look at that too. 

I played a mission on Syria yesterday and had no trouble at all with the screen.  It felt the same as always only this time the image was green. 

 

But your pictures again show a difference.  

Any results @Hobel?

[Hobel]

vor 21 Minuten schrieb NineLine:

Any results @Hobel?

so this is what it feels like

 

 

have looked at his mission and the track, for me it looks the same from the sharpness as before only in green

 

what i see in his track:

@okopanja does mine look better than yours?
Your pictures are unfortunately quite small that you have sent here
and therefore I do not recognize anything.

Edited November 3, 2022 by Hobel

[okopanja]

@Hobel Thank you for trying. I noticed you have lower graphical settings than me and are actually using zoom, this is normal, but in this case can be hardly used as a baseline for comparison. Zooming in more details will reveal a bit more details for sure, however, I left it initially at default zoom level so we have reproducible comparison.

Furthermore I did try to replace the skhval mask with the one found in 2.7 (IC check warning!). This improved slightly, but it appears the problem is deeper than just replacing this texture.

The observed behavior is consistent with probably the following rendering sequence:

1. first model background (if you open EDM, you can see screen is dark, dark gray)

2. overlayed picture from shkval

3. overlayed the graphic elements (caging box etc)

4. overlayed the green mask with some opacity level.

This is not proper way to simulate the picture of the black&white CRT display. They do not work like this, and in fact these kind of displays usually had way better contrast compared to their color brothers for very obvious technical reasons (e.g. I am old enough to have witnessed this difference).

As a further proof, please observe the caging box on one and another picture, you will notice it is barely visible.

One contributing factor could the fact that lighting conditions got changed with 2.8, and as a result many of the models are affected. E.g. in Volk's video you can see the stick is way too bright, but also I noticed this on J-11A externally, which used to have a very sexy dark gray skin, which looks now bleached out.

BTW: I do not wear glasses, and do not exchange artificially contrast on my own monitor.

Edited November 3, 2022 by okopanja

[Hobel]

vor 11 Minuten schrieb okopanja:

@Hobel Thank you for trying. I noticed you have lower graphical settings than me and are actually using zoom, this is normal, but in this case can be hardly used as a baseline for comparison. Zooming in more details will reveal a bit more details for sure, however, I left it initially at default zoom level so we have reproducible comparison.

Furthermore I did try to replace the skhval mask with the one found in 2.7 (IC check warning!). This improved slightly, but it appears the problem is deeper than just replacing this texture.

The observed behavior is consistent with probably the following rendering sequence:

1. first model background (if you open EDM, you can see screen is dark, dark gray)

2. overlayed picture from shkval

3. overlayed the graphic elements (caging box etc)

4. overlayed the green mask with some opacity level.

This is not proper way to simulate the picture of the black&white CRT display. They do not work like this, and in fact these kind of displays usually had way better contrast compared to their color brothers for very obvious technical reasons (e.g. I am old enough to have witnessed this difference).

As a further proof, please observe the caging box on one and another picture, you will notice it is barely visible.

One contributing factor could the fact that lighting conditions got changed with 2.8, and as a result many of the models are affected. E.g. in Volk's video you can see the stick is way too bright, but also I noticed this on J-11A externally, which used to have a very sexy dark gray skin, which looks now bleached out.

BTW: I do not wear glasses, and do not exchange artificially contrast on my own monitor.

 

so I tried it without zoom and kept your view

 

 

Mine:

your


what are your graphics settings?

[okopanja]

I found also some more differences. The dark one is the 2.7 and light one is 2.8. I believe that above rendering sequence results in reduced dynamic range of the resulting shkval display, thus severely impacting the contrast levels. My graphic settings will follow shortly.

Added settings. On tiny details 2.7 image was captured with more rendering distance and with default shadows. I made quick check and could not see that this change improves the 2.7 shkval.

GPU is NVidia RTX 2070 Super.

 

Edited November 3, 2022 by okopanja

[Hobel]

okay another one.
My settings

your

[River]

On 11/2/2022 at 9:05 PM, okopanja said:

The 2.8 version did bring the new green color of the screen. along with it I have noticed the performance degradation with SKHVAL which results in much more difficult spotting. I have noticed this while playing at BS server, on the first session where I tried to use shkval.

I have made screenshots to demonstrate in following way:

1. I selected hot Ka-50 on ground (for stability and reproducibility, if needed you may take of or use the one that is already in flight)
2. I made no changes to the view, used just what the game displays by default
3. Located the same area
4. Used Brightness & Contrast to get the best possible picture
5. I captured full resolution pictures (3440x1440) and saved them as PNG (lossless format), but could not upload them (please tell me how to deliver them 6,28MB vs 7,71MB), so I cropped the image.
6. Compare the images, the difference is quite obvious.
7. Notice the subtle but important difference: follow the left vertical line of the caging box slightly up. In old Shkval you will notice a Stinger manpad, it is not much visible but it is there! In new Shkval this is nowhere to be seen no matter what I do with Brightness & Contrast.

Used versions for comparison:

- 2.7.18.30865

- 2.8.0.32235.1

Attached are:

- mission file

- trk files

 

I can see the Stinger on both pictures, if its the dude running left just above the caging box. 

[okopanja]

11 hours ago, Hobel said:

okay another one.
My settings

your

Interesting, some questions:

1. your settings?

2. which control do you use to set brightness/contrasts (e.g. keyboard, mouse, mini-stick)

3. which GPU?

[Hobel]

vor 2 Stunden schrieb okopanja:

Interesting, some questions:

1. your settings?

2. which control do you use to set brightness/contrasts (e.g. keyboard, mouse, mini-stick)

3. which GPU?

1. turn off the "Cockpit Global Illumination", this makes the difference in the skhval.

2. I almost never adjust the contrast, is always default

3.2070s

Edited November 4, 2022 by Hobel

[bingbean]

Heat blur? Lens effects? Cocpit global illumination?

[Schmidtfire]

Looking at the little bit of real world footage and screenshots we do have... The monitor always has a very high contrast. Makes a lot of sense too since
the system tracks on object contrast.  Most likely vehicles and tanks would be very dark against the background, Much darker than even Hobel's
screenshots. 

Just look at the contrast on those trees and fence. Like a silhouette.

[bingbean]

Not big deal , however now on the green TV it is more noticeable the problem of the curved edge of monitor. There should not be any image, the edges should be black or dark green , no matter if the TV is on or off.

Current state:

 

 

 

 

How it should be:

Edited November 4, 2022 by bingbean

[Hobel]

vor 8 Stunden schrieb Hobel:

1. turn off the "Cockpit Global Illumination", this makes the difference in the skhval.

@okopanjais that a thing for you?

[okopanja]

14 hours ago, Hobel said:

@okopanjais that a thing for you?

Hi, I just repeated tests without cockpit illumination:

If "Cockpit Global Illumination" is disabled, there is no noticeable visual degradation.

If "Cockpit Global Illumination" is enabled, there is very noticeable visual degradation.

Disabling this options makes it a very good workaround, but unfortunately not a solution. IMHO: in order to retain prior contrast we should not loose the visual feature.

Question remains if the background of the model should be darker as in 2.7 (I attached the pictures above from model viewer)?

Present display still looks very washed out. True this can happen but only after many years of everyday usage in a household, which certainly outpaces the usage in the helicopter. Typically this was due to the degradation of CRT itself, where the background "green" would become more dominant over the phosphorus emitted photons over the time.  Furthermore I would like to underline that the monochrome CRT possessed the great contrast, which was in many cases better than LCDs for a given screen resolution (single monochrome dot always wins against 3 color dots when it comes to sharpness). The displays of this size were use for decades with oscilloscopes, where sharp high contrast picture was essential.

The contrast of this display should remain independent of the changes to the external lighting conditions.

Further changes to the lightning system could affect more drastically this important display in future. One of the main issues here is that majority of developers and players actually never used the B&W TV, or if they did they did witness the fairly old and barely functional devices. My late grandmother owned both old and new B&W TV set, so I do remember the "green" effect on the old one, where new showed true B&W. Back than I remember I got really disappointed when she bought second B&W TV set (mostly due to her being used to monochrome displays), since I carried my C64 during summer and winter holidays. I still remember the shock I had when I realized that the uncool B&W TV offers much better text readability due to the contrast than our shiny color TV set.

The advantages of having monochrome imaging for both sensors and display, meant that this kind of technology remained in use for many years in places such as surveillance, astronomy and avionics, even if display technology changed over the time.

If we are to get BS3 soon, I would expect the display to actually perform better than what we are offered with this relatively low effort visual uplift. IMHO: the rendering sequence I mentioned earlier likely needs improvements to avoid later side effects. On of the problems we are facing is that likely the majority of reference pictures used with cockpit used analog camera/camcorder devices, which further degraded the image and colors. This effect of media degradation should be taken into account.

I believe that related issue is likely caused by the same changes:

 

[Hobel]

Whether this is intended with the setting remains to be seen, but for now we know what causes the bad picture.

 

@NineLineI think we have found the cause.

[sasquatch98]

On 11/4/2022 at 12:43 AM, River said:

 

I can see the Stinger on both pictures, if its the dude running left just above the caging box. 

Same here