Jump to content

Photo

Really racing the beam with PC-based emulators.


7 replies to this topic

#1 Keatah OFFLINE  

Keatah

    Missile Commander

  • 21,035 posts

Posted Thu May 10, 2018 11:03 PM

Found this article and haven't even read it yet. But it looks interesting!

https://www.blurbust...tor-developers/

 

..and WinUAE is starting to implement it!

http://eab.abime.net...?t=88777&page=8

 

TRIVIA: I was quite surprised to read that a modern graphics card can page flip its frame buffer 10,000 times per second on a low-res image. And do computations on each frame, too! Didn't mention what low-res means, but I would guess around 640x480.


Edited by Keatah, Thu May 10, 2018 11:16 PM.


#2 Mr SQL ONLINE  

Mr SQL

    Stargunner

  • 1,946 posts

Posted Thu May 10, 2018 11:34 PM

Found this article and haven't even read it yet. But it looks interesting!

https://www.blurbust...tor-developers/

 

..and WinUAE is starting to implement it!

http://eab.abime.net...?t=88777&page=8

 

TRIVIA: I was quite surprised to read that a modern graphics card can page flip its frame buffer 10,000 times per second on a low-res image. And do computations on each frame, too! Didn't mention what low-res means, but I would guess around 640x480.

 

Interesting article! No monitor has a refresh rate that can handle that though.

 

Their experiment inserting blank frames to make the HZ match the FPS for clearer animation matches my research with STARBLITZ:

 

"In the past, Blur Busters has made a few contributions to the emulator scene, such black frame insertion (www.testufo.com/blackframes) which allows 120Hz monitors to gain better 60Hz motion clarity for emulators."



#3 DirtyHairy OFFLINE  

DirtyHairy

    Moonsweeper

  • 450 posts
  • Location:Germany

Posted Fri May 11, 2018 9:17 AM

Found this article and haven't even read it yet. But it looks interesting!

https://www.blurbust...tor-developers/

 

..and WinUAE is starting to implement it!

http://eab.abime.net...?t=88777&page=8

 

TRIVIA: I was quite surprised to read that a modern graphics card can page flip its frame buffer 10,000 times per second on a low-res image. And do computations on each frame, too! Didn't mention what low-res means, but I would guess around 640x480.

 

This is interesting from a technical point of view, but I don't think it maps well to VCS emulation: the frame structure and frame rate output by the TIA are generated by software and usually are not exactly 50Hz or 60Hz, so locking frame rate to the host will cause deviations to the emulated CPU speed (and cause issues with audio emulation). Also, I must admit that I find it hard to believe that a lag of one or two frames between input and video is actually noticeable --- that's <  50msec. Afaik, human reaction speed is above 100msec, and there is also a lag from OS input processing and scheduling that has to be taken into account. I think this is an area where placebo effect and measurable facts are hard to tell apart.


Edited by DirtyHairy, Fri May 11, 2018 9:33 AM.


#4 Mr SQL ONLINE  

Mr SQL

    Stargunner

  • 1,946 posts

Posted Fri May 11, 2018 1:28 PM

 

....Also, I must admit that I find it hard to believe that a lag of one or two frames between input and video is actually noticeable --- that's <  50msec. Afaik, human reaction speed is above 100msec, and there is also a lag from OS input processing and scheduling that has to be taken into account. I think this is an area where placebo effect and measurable facts are hard to tell apart.

 

If that were true no one could beat WARPDRIVE in 30 FPS mode where it requires a 30th of a second reaction time (33 msec)!

 

Fair enough in my research only individuals with extremely fast reflexes from meditative practices like yoga or martial arts have been able to beat WARPDRIVE on the first try, though many have been able to master it over time. 

 

Now a 17 ms reaction time really is pushing it for humans - I haven't seen anyone beat warpdrive at 60 FPS where it requires a 60th of a second reaction time.

 

Anyone can try this experiment for themselves, after starting the game the BW switch will switch between 30 and 60 FPS:

 

http://javatari.org/...RPDRIVE_AFP.bin

 

Regarding the other experiment the researchers taked about inserting a black frame to make the HZ match the FPS, you can see the improved clarity of animation in the contest version which inserts a blank frame to achieve the effect:

 

http://javatari.org/...LINEBLITZII.bin

 

imo Stellarator should not merge three frames by default because it prevents these effects from being seen or researched; the Atari is a great platform for such research due to it's nonstandard video output.



#5 Trebor OFFLINE  

Trebor

    River Patroller

  • 4,525 posts

Posted Sat May 12, 2018 7:40 AM

Afaik, human reaction speed is above 100msec, and there is also a lag from OS input processing and scheduling that has to be taken into account. I think this is an area where placebo effect and measurable facts are hard to tell apart.

 

Agreed.  For example, the aforementioned link provides a quick test. Even under an extreme accounting of hardware latency (150 ms) was subtracted from the results, the statistics still leaves the average (284 ms), from well over 64,000,000 samples, at well over 100 ms.



#6 Keatah OFFLINE  

Keatah

    Missile Commander

  • Topic Starter
  • 21,035 posts

Posted Sat May 12, 2018 8:00 AM

That would be for a single-instance test. When there is a rhythm going with many sequential "tests" the reaction time is much lower. The single test requires conscious thought, which typically consumes 250ms. Once the activity is learned, and the required action anticipated in a regular or semi-regular cadence, like in a videogame (!), the reaction time is significantly decreased.



#7 Trebor OFFLINE  

Trebor

    River Patroller

  • 4,525 posts

Posted Sat May 12, 2018 8:30 AM

That would be for a single-instance test. When there is a rhythm going with many sequential "tests" the reaction time is much lower. The single test requires conscious thought, which typically consumes 250ms. Once the activity is learned, and the required action anticipated in a regular or semi-regular cadence, like in a videogame (!), the reaction time is significantly decreased.

 

Certainly the above 'learned activity' may be applicable for a few video games like Pac-Man, under the condition of a player following a maze pattern and knows their move in advance, or something along the lines of a tool assisted game as seen in TASVideos; however, for many (most?) games being engaged, the player is waiting for an action to take place on the screen (I.E. An enemy moves in a direction, a door opens, a platform motion/appearance, the correct position of a target or passing shot), prior to responding, and reacts with a corresponding directional movement or button press.

 

Similar for the test linked to earlier.  In that case, the action on screen of a color change, from red to green, is met with a reaction click of the mouse from the individual.  This can very much parallel what happens in video games, such as waiting for an enemy to show a certain action or vulnerability prior to being able to launch a successful (counter)attack, or determine in what direction to move next.

 

If anything, the argument can be made that at least the person knows what their reaction response should be in advance respecting the above test.  In the case of a video game where a player is 'standing-by' (for whatever length of time), needing to react to a certain action, and the specific reaction requires a response that differs, depending on what transpires on the screen, there can be a longer response delay from the person: First, recognize when the time is right to (re)act; second, process what appropriate input needs to be accomplished; and then finally, match it accordingly.



#8 Mr SQL ONLINE  

Mr SQL

    Stargunner

  • 1,946 posts

Posted Sat May 12, 2018 10:48 AM

That would be for a single-instance test. When there is a rhythm going with many sequential "tests" the reaction time is much lower. The single test requires conscious thought, which typically consumes 250ms. Once the activity is learned, and the required action anticipated in a regular or semi-regular cadence, like in a videogame (!), the reaction time is significantly decreased.

 

Excellent point, the fact that people become focused on playing a game, drawn into it, creates a state of heightened concentration that is probably key to achieving the reduced reaction time of 33 msec illustrated by gamers who can beat WARPDRIVE.

 

The MotionBlur site has a lot of interesting research, found some more findings that mirror my own research:

 

Many of our readers crave perfect motion clarity. Some are former CRT users who prefer a similar motion clarity experience. It benefits certain kinds of games (more than others). In addition, some of our readers get eyestrain/headaches from gaming motion blur rather than PWM/flicker.

 

https://www.blurbust...blur-reduction/

 

This is certainly true for me, I find STARBLITZ more pleasant at 30 HZ and 30 FPS than when switching to 60 HZ and 30 FPS; the impedance mismatch creates double vision that is particularly unpleasant in games like Thrust where the scrolling objects double up.

 

Also interesting that gaming monitors are including the ability to insert black frames as part of their feature-set now. imo that is something a software emulator should do on it's own so as to utilize the black frame to leverage real phosphor emulation - with 120 HZ, an emulator could use a phosphor trail frame in place of the black frames to show fluid 60 FPS games. 

 

Their analogy with the Atari racing the beam was a good one - we get 60 HZ at 60 FPS with the nonstandard signal and it's up to us to manage both flicker and motion blur. Atari programmers are generally familiar with flicker but could benefit from researching motion blur when designing high frame rate scrolling games.






0 user(s) are browsing this forum

0 members, 0 guests, 0 anonymous users