Input Lag for Different Games

Tech Talk
41

Now, to give some explanation. This results may be out, however they should be inaccurate by a set difference which doesn’t depend upon the game or the system. If you look at the BBCF images. You will see on the first image that there is a purple line near the top of the screen (but with a red bar and then a little bit of black and white image above it). What has happened here is that the image coming from the HDMI out through the component converter has been watermarked at the time when the button was depressed.

After this, there is a full frame where nothing happens, and then on the following frame, the animation begins (subtle, but trust me, since BBCT Ragna’s animation involves him lifting his fist up on the first frame of animation) and the command history is updated. So, we know that at least 16.66msec has elapsed from the second frame where nothing happened.

In order to work out how far along the first frame it was when the watermark occurred, I (arbitrarily, and this is where the unknown issue occurs) chose to start counting from when the purple (and red) bar is at the top of the screen. From there I count how many frames it takes to get all the way to the top again. If we then work out what portion of time has elapsed. If we take the fraction of a frame and multiply by 16.66ms per frame, and add this to the whole frame, it will give us the total time before the animation change.

42

Now, we expected some variation in the results, as every previous test has shown that there is variation between different joysticks. I intentionally used the ufb because it is the fastest board that I’ve tested, and it also tends to give more results to analyse. In the end, I got 200-250 results for each game.

What cannot be explained by this is the massive variation in SFV’s input lag. I have provided examples of the quickest and slowest responses, but for them to vary by over two frames was something I wasn’t expecting at all. Also, SFV is the only game where the command history and the animations don’t seem to always occur consistently together. And further, it doesn’t seem to be particularly dependent on the location within the frame. For example, If you input a command too early in the frame for BBCPE or BBCF, it will be ignored, as it is only the last portion of the frame where commands will be accepted. SFV doesn’t behave like this, and bizzarely some of the quicker responses occurred at the end of the frame.

43