I’m at a better computer and a faster Internet connection than yesterday, and I just tried all 5 tests again. I’m not sure what we are supposed to be looking for. There are two horizontal lines that flash when the moving horizontal bar is near the middle, and there is the descending/ascending bar at the right that hits the bottom about the same time. So there are 3 visual actions plus the click track. That’s four things happening. Which ones are we supposed to be comparing?
Maybe it doesn’t matter, as they all appear to be synced within a half-number (what unit do those numbers represent? Frames?) of the bottom scale, at least in a quick run-thru. I just don’t see a greater variation, and if there is supposed to be one, I guess I’m deaf or blind.
I’m also not sure if my video playout is playing all frames. Many players drop frames here and there in order to keep from getting behind, and I have no way of determining if this is happening. The CPU, etc. should be fast enough, but I still can’t tell. If a single frame is dropped, that will distort the sync test by at least 1/30 of a second, or 33ms.
I know exactly what you’re talking about, but it’s a totally different experience. The test is syncing audio and visual stimuli, which our brains receives from two different organs. It has also been suggested (in the GQ thread I linked to in my OP) that our brains will do a little bit of gymnastics to make audio and visual signals appear to be synced up even if they aren’t.
When I was first making the test examples I screwed up and added the 40 ms delayed audio (or was it advanced 40 ms… one or the other but I don’t remember which) to the master video without removing the master audio track first. The two tracks playing simultaneously sounded like two metronomes just a little off from one another. It was painfully obvious. As soon as I removed the master audio track, it was much harder to discern the audio/visual delay even though I knew it was there.
Each notch on that ruler measure a single frame, with 25 frames for every second (weird, I know). So that’s about 0.4167 seconds per tic on the ruler.
No, you’re just contributing meaningful data to this experiment.
I’ll be posting the results later today, and I think everyone who participated will find that our data has been very informative -especially comments in this thread like yours.
Alright, the poll has reached the magically round number of 50, so I’m going to post the results:
[SPOILER]plus = audio delay, minus = audio advanced
VIDEO 1 -40 ms
VIDEO 2 +/- 0 ms
VIDEO 3 +80 ms
VIDEO 4 -80 ms
VIDEO 5 +40 ms[/SPOILER]
and my reaction to them:
First of all, congrats to brazil84 and muldoonthief for being spot on. You guys have some fantastic perception. The results indicate that about half of the Dopers all correctly heard #2 as being in sync. About 3/4 of you thought it was either #2 or #5, which has the audio delayed by 40 ms. That’s totally in line with what the links I provided indicate -a slight delay is less noticeable than any advancement since that is a more natural occurrence. It appears that it does in fact depend on the individual, since some of you could not tell the difference between any of the videos. That being said, this experiment seems to support brazil84’s hypothesis that 40 ms of delay is noticeable.
I said that I would also post the Youtube counts, but you can see for yourselves that they’re all close enough to indicate that our little experiment went according to plan.
And interestingly enough, VIDEO 2 has the most views. I guess a bunch of you were double-checking!
Thank you to everyone who participated! To those of you who didn’t… why are you reading this thread anyway?
Wasn’t Brazil’s view that 40ms is very noticeable? Because the results of this would seem to indicate that if there can be an objective idea of “very noticeable” 40ms isn’t very noticeable, since only about half the people noticed it.
So if I see the top two match, but not match the bottom, what does that mean? If I see the right match the bottom but not the left? If I see…well, my point is you are comparing 4 things and I don’t know what you expect to line up. If all 4 don’t line up, what shall we report is wrong?
Sounds like that was created in a PAL environment, not NTSC. Dunno how that would affect your experiment, but it shouldn’t be ignored.
TO SUM UP:
You got some people who lived up to your expectations. You found at least 2 that said they saw no difference. You found some that observed differently from what you expected.
Then you praised those that matched your expectations. Maybe they were just lucky? Give 10 random answers and some will appear to be right.
And you haven’t eliminated the technical problems with this test, so the random answers might be just as good as the supposedly careful ones. Toss some dice and see if your results change.
Sorry to rain on your parade, but I just don’t feel this is a valid scientific test until all variables have been addressed. However, I agree with you strongly that audio/video sync is important and should be better than we usually see it.
Any internet poll is going to be unscientific by nature. The results were extraordinarily similar to the controlled University experiment I linked to. They also match the information from the other two linked articles. If the results did not match, I would agree that this was a highly invalid test. However, that just isn’t the case.
You can feel free to interpret the data as you see fit.
It’s not the Internet factor that makes it unscientific or invalid, it’s that the technology isn’t appropriate to detect what you want it to.
If you are testing how even and clean an image can be from a particular camera, you don’t use low-res JPGs to compare, because they introduce artifacts that distort your test. I think your sync test may have a similar flaw.
It might just be your honesty. I don’t see any difference, either, at least not consistently and without doubt.
That’s a fair point of view, but I chose the video purposely because it is designed to reveal any delays in audio. I am sorry that you were unable to detect the delays, but that doesn’t make the test invalid. It simply means that your individual data does not reflect the sum of the data. Approximately 50% of those tested accurately chose #2. Approximately 75% of those tested chose either the correctly synced video or the next closest approximation. Only about 15% chose the outliers. Just because your individual perception did not reflect the majority does not mean that the whole experiment was fundamentally flawed, and there is no need to be defensive over it.
I got it, but I’ve worked in audio video industry and am aware of the issue. For anyone who didn’t get it, don’t feel bad. Like many other perceptions, a little bit of training or even exposure goes a long ways. Most people can start to pick up these sort of things quickly.
They may have chosen #2, but that doesn’t mean they were accurate. If I made 5 guesses, and one of them was #2, that doesn’t mean I was perceptive, just lucky. You are assigning a value based on nothing more than random chance. You are fooling yourself.
Your test is fundamentally flawed, no matter how I perceived it. You have not demonstrated that a valid test can be done within the highly variable parameters. You have no controls. You are looking for confirmation and you found it. It may make you feel good, but this test is worthless.
No, really, she(?) isn’t. I complained about the test, too, but if it worked for the majority of people, then I have to admit that I was wrong. Claiming that it’s just a fluke that so many people got the right answer beyond the margin of error is just ridiculous.
Now, could the test have been better? Sure. But that doesn’t mean the data is invalid. Trying to point out that it could random chance when the data shows a marked deviation from random chance is just grasping at straws.
Musicat, like me, made a mistake in criticizing the test. For some reason, Musicat doesn’t want to admit it. You are not entitled to the opinion that you were right when you were wrong.
Your #2 video is processed somehow and not just a direct copy of the BBC test pattern video, right? I think you may have done something to introduce a frame or two of desync when making the videos, because I can identify the BBC reference video as being in sync, but your #2 seems slightly delayed to me.
FWIW, my guesses were that 1 was slightly early, 2 slightly late, 3 very late, 4 very early, 5 later than 1 but earlier than 3. Which correctly guesses the relationship between the videos. However, I suspect that something you did in treating the video introduced a frame or two of lag to every one of the tests, because they’re all off.
Test 1 becomes closer to being accurate simply because the video seems delayed and that one is the slightly early one. I knew the real time fell somewhere between #1 and #2, but I don’t feel that either is spot on.
I felt this way as well and ended up picking #1. On re-watch after looking at the “correct” answer, I now feel that #2 is probably more correct, but there’s something about a delay that jumps out at me, whereas slightly early seems OK. It’s crisper, maybe more acceptable. The ones where the audio were delayed just seemed sluggish and obvious. But I do still feel that there’s a slight delay with #2.
I’m surprised that way more people picked #5 than #1.
