Human perception is less trustworthy than you think. It isn’t worth a moment of time thinking about the lag in the setup unless the setup had an HDTV or some grimey controller converter.
Is it possible that there is lag differences between ps3s? Yeah, sure, I guess, but the likelihood that you are simply mistaken is like 99.99% so just don’t bother
I did not say that 2 frames is irrelevant.
What I was getting at is that a human cannot accurately determine differences of 2 frames of lag between one setup or another. If you are playing on a CRT or a known lagless monitor that is properly hooked up and you are missing links (that you do based on visuals) it is FAR more likely that you either learned the combo on a lagged monitor or you are just having an off day.
You shouldn’t be visually linking hardly any combos anyways. (uh, chickenwing combos?)
I know exactly how trustworthy human perception is. Which is why I’m asking if there’s been a test for what I was talking about and not just assuming anything.
These are BNC to phono adapters. They will allow you to use a set of component video cables to connect your PS3 directly to the monitor.
Now, if your ps3 is set to any progressive resolution and you attempt to connect to the FW900 or any RGB monitor in this way you will see a clear image on the screen but you will run into one issue very quickly. The screen will be tinted a very heavy green color. The is due to differences in the color encoding between YpBpR aka component video and RGB video. Notably, the green is so overpowered because the other two color channels are missing much information.
HOWEVER. That’s irrelevant for the purposes of this test. The image is perfectly playable and recognizable and could be used to measure the speed of a PS3 over a component video cable in 480p, 720p, or even 1080p.
Additionally, it is possible to go into the PS3’s display options and select the AV-multi/SCART option. From here you can choose RGB mode and actually get real correct RGB colors in 480p. But if I remember right you have to do this blindly because it defaults to 480i first, and the FW900 can’t display that. So it’s something like:
select AV-multi/SCART
select RGB and push right
the screen will go dark
press down to select 480p
then press right again to get correct video
and choose ok
I’m not really sure what the real closest thing to native res would be for 3rd strike online. There’s a lot of X factors to take into consideration beyond the original resolution of the CPS3 game. However, RGB 480p from a PS3 to a CRT computer monitor *should *have the lowest input delay possible excluding outside factors.
Now, I know the HDFury is a fine product, but if you wanted to test and be absolutely sure it wasn’t adding anything you could consider the above options.
Thanks for this test! It’s somewhat relieving as a PS3 3SO owner after all the talk the last few days, lol.
I somewhat disagree with filters/scanlines are irrelevant for the testing here. It’s part of the 3SO package as a whole (which we were told would match the arcade) and I think that makes it somewhat relevant if we are not able to run the features they gave us at arcade speeds.
Also, on the topic of the comparison between consoles, filters/scanlines may have difference performance on each console, which would be an important piece of data for us to be aware of.
Needless to say, I hope you do get the time to try a comparison with the filters/scanlines on =).
Warpticon I get what you’re saying (also sup BTX), but my understanding is that inputs are processed on a console’s video output’s VBLANK interval, which is at the end of every frame. Basically the console takes a snapshot of your inputs every 16.7ms / 60 times a second. You can input a command exactly when a frame starts, and wait the 16.7ms for the frame to end and have the game process it, or exactly before the end, and wait 0ms for the frame to end and have the game process it, but the fact of the matter is your command comes out at the end of that frame regardless. There’s no intra-frame command reading. You can only have lag in 1 frame increments.
If I am wrong about VBLANK on the 360 and PS3, let me know. But it’s my understanding that reading on VBLANK is how almost every console (or at least older console) processed inputs.
Also, the earlier comment about a supergun itself lagging shouldn’t be correct. Unless this was going through an XRGB converter or something known to lag, the RGB->NTSC conversion is done with analog circuitry (such as an AD725), which does some math to the RGB signal on-the-fly to turn it into NTSC signals. Again, correct me if I’m wrong (I’m not as much of a nitty-gritty-hardware guy as Toodles and the like), but my understanding is there’s nothing digital happening, so none of the caveats of reading 1+ frames into a framebuffer, processing them, and releasing a delayed stream of the converted signal, should apply.
Let me start with the observation that ‘1 frame’ isn’t actually a fixed time interval. When, for example, I run SSFIV on my Xbox, plugged into ye olde fashioned CRT television with a composite NTSC signal, then the screen is displaying at about 29.97 frames per second. Meanwhile - as long as buttons are down - the game runs at about IIRC 60.15 frames per second. (SSFIV’s frame rate is different if no buttons are pressed.)
Other current generation games like MvC3 also do odd things with frame rates due to rendering issues. Earlier games probably had similar issues when run on period hardware.
Street Fighter 2 HDR on the XBOX apparently implements ‘tie breaking’ by sampling one player, and then the other. This, and possibly other factors, affect when the system will sample the input so it’s possible for the game to drop inputs that are there for a duration of a perfect 1 frame time.
There is a wide variety of sources for lag. For example, the MADCatz SE stick runs about 1ms slower than the PS360 does. (Probably also slower than the HORI.) It’s not enough to affect play, but it’s definitely measurable.
There’s a lot of noise on the scale of 1 frame, though so making good measurements gets a bit harder.
I helped run a SSF2T tourney on superguns at RevaLAtions. For the recorded games that I have up on Youtube, there was a splitter/converter that apparently introduced lag on one of the heads of one of the set-ups. It was taken out when one of the players complained - rather emphatically. He previously (and later) played on other stations without issue. A number other players had used that set up (with the lag) without complaint. There were no complaints about any of the other setups having lag.
When I discussed the lag issue with some of the players, one mentioned that every ST set up feels a bit different. I have no idea how much of that is psychosomatic. The use of various converters, displays, and set-ups in general is an issue of some contention with the ST crowd.
If I get a chance this weekend I’ll see if the filters cause any lag. Is anyone else doing any testing that anyone knows of?
Thanks for explaining it more succinctly than I did.
I’ve heard a lot of speculation about the PS3’s USB ports, but never seen any real evidence. PS3 SF4 series lags a frame more (some say two) than 360, and after that info came out, I started hearing about PS3’s USB ports being slow. Or that PS3 takes 1 frame longer to draw to the screen. Or that ports to PS3 have input lag. Or that it just depends on the game. There is more speculation than truth.
From my perspective that was just SF4; other games have different results.
Thanks for posting that, I’d almost forgotten about the BNC inputs. I have them hooked up to something else. I do want to stress that I’ve tested the 360’s native VGA vs HDFury on several occasions and there have never been skewed results between them, at the level of resolution we’re talking about. I’ll use the HDFury for any other tests relating directly to this, but I’ll check out the BNC method for future tests.
It shouldn’t be, I was just posting what Laugh had said (about either the arcade hardware itself or the supergun causing it). I would bet on the CPS2 hardware.
Do you have more info on lag tests of retail sticks and enthusiast PCBs? I’ve only seen a little about this. I know Toodles’ code in the Cthulhu (I would guess this extends to the TE Kitty) does 1ms updates.
Right, but we’re talking about two different things here… :lol: At some point I’m going to have to break this down in a detailed writeup because apparently I’m having a very hard time getting the point across.
Play high level DDR/ITG and tell me you can’t feel three frames of lag. You’ll get a “Perfect!” (or worse) instead of a “Marvelous!” every single time on your first step.
I think he means that by feel, players that think they can tell how many frames of lag something has (“accurately determine”) don’t have reliable results. Sometimes they even disagree with themselves; last year at Evo, Tokido complained about lag on one of the stream setups. They had him blind-test a directly-connected setup vs. the stream setup, then he said the direct one lagged and the stream one didn’t.
That’s not to downplay the effects of lag. When it comes to 1f differences it can be frustrating, because you can feel that something is wrong, but may not be able to A/B it well. That’s why it’s better to do this sort of testing.
these are 10 consecutive photograms from a video I originally recorded at 60 fps
CRT television, xbox 360, RGB microsoft cable, madcatz TE
look at the shoulder to see when he’s moving (open widescreen to see it better)
[media=youtube]V_HSIAWWdfA[/media]
this video contains photograms from just one of the many attempts with same results
Besides the fact i like the old graphics better and thats what i plan to use… “smooth” filter and scanlines don’t seem to alter anything in terms of input lag… ist still 4 frames… I tried both
i just wanted to confirm RoboKrikit’s results, after having looked at the 720p source files. there does not appear to be any lag difference between the ps3 and xbox360. most of the time, the delay was 4 frames after the LED lights up. it never went above 4. sometimes it was 3 frames (again on both ps3 and xbox360), but RoboKrikit already accounted for that earlier, i.e., the captured frame captured “the middle” of the LED’s lightup, vs capturing “the start” of the LED’s lightup.
some of these frame-boundary issues might be allayed if an even faster camera were used, recording at 120fps or something.
not completely relevant to this thread, but, like others, i would like to know what accounts for the 4 frame delay. i don’t know anything about the HDFury converter but it would be nice to remove it somehow, just because i feel like any sort of conversion is going to introduce some concrete amount of delay.
actually i thought of another possible test, one that pretty much anyone could do since it wouldn’t require expensive video equipment: recording sound. this may provide for even finer resolution, considering sampling rates are 48+ kHz, and you can isolate the “start of the frame” down to one sample simply by looking at the wavestream in sound forge or something. simply mark where the button hit starts and where the resulting sound starts, and calculate how many frames that corresponds to. this may require a microphone for the stick and a microphone for the TV, then producing a sound loud enough for both mics to pick up to facilitate syncing of the two recordings. dude i might even do this test myself. it might not be super practical information, since there’s pretty much always a delay between audio and video, but still, it provides another data point.
There’s a thread somewhere about Hori vs MadCats sticks. My testing was only on the Xbox360 side of things. (I’m not going to bother spending time or money testing something that nobody really cares about. I really only checked stuff because I could do so easily and was curious after the Hori v MadCatz thread.)
More people should run tests. I see a few people re-analyzing the data, but Robo’s a smart guy. His data should be analyzed correctly. Other people should do independent tests to verify these results under varying conditions.
I think it’s unfair to have him blind-test things AFTER that Evo since he’d already be adjusted to the lag on the Evo monitors, biasing his opinion. That may have affected the results of that test.
It’s worth pointing out that the equipment isn’t one frame precise. Consider, for a moment, SSFIV on the Xbox 360 which has a frame rate independent of the display. So a bad case scenario on input is that you push the button just after the game has polled instead of before, so that’s 1/60 sec variation before it processes the input, and then the game video update could be just in, or just out of sync with the ‘scan line’ which is another 1/60 of variation. So it’s basically inevitable to have almost 2 frames of variation in lag.
The nature of human vision (and of cameras) will mask a lot of that.
i recorded at 60 fps and watched frame-by-frame, which is not the best… i did many different attempts, around 15-20 i’d say… its still kinda approximated, i know, but it never resulted in more than 4 frames…
You guys are awesome for running all these tests. I’m new to Third Strike so wouldn’t know if anything was off. But it’s cool there are people running these tests and sharing the info with the rest of us. Thanks again for the hard work. :tup:
Thanks for your test. Sounds like good news. I’ll still run one if I get the time.
Thanks for checking those out, I appreciate it. With some of the issues I had with VirtualDub I wanted to get my results double-checked. I’ll see if I can do a quick run using the 360 VGA cable this weekend just to be sure.
Yeah, being put on the spot like that with things like 1f differences is tough. I’ve done my own blind A/B tests before and it gets hard when you’re talking about 1f. If you’re very used to tight timing on a certain setup, a 1f difference is enough to throw you off and think something is wrong. But trying to just A/B them side by side and guess which is which is not easy. 2 or 3 frames is easier to notice. With 2+ frames lag it’s easy to whiff simple stuff like jump-in 2-hit SF2/SFA combos.
He did claim that the stream system was laggy (because of the stream equipment) though, so they called him on it. I only brought that up as an example of one-off player opinions being hard to take at face value, not trying to pick on Tokido.
Thanks everyone for the thank-yous. I didn’t respond to all of them to avoid cluttering up the thread, but it’s appreciated.