20 replies to this topic

[SennaVsProst]

Well, here is possibly a simple question, and one I've wondered about for quite some time as well. Occasionally, I'll notice this strange optical illusion which makes the wheels on a car, or even more exagerated on an F1 car appear as though they are rotating clockwise when moving forward, typically under breaking there will be short moment where lets say the "Bridgestone" stamp on the wheel appears as though it is rotating counterclockwise, which is normal, then a second or so it appears stationary, then, bam it looks as if it is rotating clockwise... Why the hell does it do that?? I wish I could post video, because when i was watching Imola again I noticed a certain shot where this occurs absolutley clear as day... anyways thanks to anyone who can just tell me what the hell is going on??
thanks.
dave.

[blkirk]

The TV images are really a series of still pictures taken many times a second. In NTSC, the rate is 30 frames per second. I believe PAL uses a different frame rate, but I couldn't say for sure what it is. At any rate, for an NTSC feed, if the tire is turning at 30 revolutions per second (or 60, 90, 120, etc.) it will be turned the same way every time the camera takes a picture and will look like it isn't turning. If it's turning at 31, 61, 91, 121, etc. revs per second, it will look like it's rolling forward very slowly. If it's turning at 29, 59, 89, 119, etc. revs per second, it will look like it's rolling backwards very slowly.

Each time the tire's rotating speed passes through a multiple of the TV frame rate, you'll see this effect. Multiples of 1/2, 1/3, 1/4, etc also work, but cause other interesting effects like the logo appearing in several places on the tire.

If you can find a variable speed strobe, you can play with this effect at home.

[turin]

Actually, It doesn't happen only when you watch television.
Just watch at one helicopter propeller. At sometimes, you can even watch them staying still. It depends on the frecuency that one object passes for the same place. At specific frecuencies, you will be able to see it staying still, and then move counterclockwise.

Uff My english today....
Bye
Pato

[SennaVsProst]

Thanks for some info, but this also occurs, when just watching a car live in person, same with my ceiling fan, I can get the same effect...wonder why that is...
dave.

[Wolf]

SennavsProst- it's because Your eyes work on the same principle as cameras, except their sampling rate is 24Hz (24 pics in a second). As You may noticem it's close to sampling rates of all visual media (film, TV/video, &c).

[MattPete]

Originally posted by Wolf
SennavsProst- it's because Your eyes work on the same principle as cameras, except their sampling rate is 24Hz (24 pics in a second). As You may noticem it's close to sampling rates of all visual media (film, TV/video, &c).

Wolf: the critical flicker fusion frequency is 60-80Hz (higher in the magnocellular, er, peripheral vision, + some individual differences). That's why televisions are interlaced: to get rid of the flicker by effectively doubling to 60hz (and adding a lot of blur, too, for good measure).

All good monitors (do they make anything less these days, or am I just old?) have a refresh rate of 66 Hz or higher. For today's larger monitors, that's not even enough. If your monitor is big enough (or you can move closer so that it covers more of your visual field) and you stare at the corner, you might see flicker out of the corner of your eye (especially with a white background) if you set your refresh rate low enough.


Your brain certainly "samples" at greater than 24 Hz (I can see the flicker in a movie theater), but it depends what you mean by sampling. The closest thing to sampling (as you talk about it) would be some data from the early 1970's using the RSVP technique, in which items are rapidly flashed one after another to the same spatial location. I think the fastest a person can pick out a target word with decent acuracy is at a rate of 20 hz (50 msec per item). In an unpublished study with some undergraduates, we still got near perfect accuracy with a presentation rate of 45 msec/item.

Sorry to bore you.


Anyway, the tire thing has something to do with the frequency at which the environment is sampled for motion perception.

[Ross Stonefeld]

Did anyone notice sometimes when they had the old vertical camera pods on the top of the airintake, at high speed they appeared to wobble violently from side to side? Like a spring mounted on the wall behind a door so the knob doesnt damage the plaster

[DEVO]

Wolf,

Your eyes sample at more then 24 frames a second. It's more like 60 frames a second. It just so happens that 24fps is the point that your eyes will detect motion without flickering. That's why movies are filmed at that rate, why waste film when you can get away with 24fps.

[MattPete]

Originally posted by Ross Stonefeld
Did anyone notice sometimes when they had the old vertical camera pods on the top of the airintake, at high speed they appeared to wobble violently from side to side? Like a spring mounted on the wall behind a door so the knob doesnt damage the plaster

Yup. I don't know what that's all about -- I figured it probably had something weird to do with translating from PAL to NCwhatever.

Ever notice the funky moire patterns that show up when guys wear pinstripe shirts on TV?

[Laphroaig]

Just for the totaly-useless-info dept., or the hemm-never-knew-that dept., whichever you want.

Interlaced is when one picture is split up in two, one with the odd lines, one with the even:
PAL is 25 fps interlaced, that means 50Hz frequency.
NTSC is 30 fps interlaced, thus 60 Hz.

These numbers aren't really amazing when you concider that the electrical powergrid in Europe runs at 50Hz, while in the US it runs at 60Hz.

Further differences are that PAL has a higher vertical resolution than NTSC, and NTSC uses a different tranmission technique.
According to some broadcast coorperations NTSC stands for 'Never Twice the Same Color';). Color is encoded differently for PAL and NTSC. NTSC broadcasts (esp. old transatlantic ones converted to PAL) regularly had colors switching suddenly. Green turned yellow, red turned blue, etc etc...
These days it's all transported digitally, so there is less quality loss.

[klipywitz]

Forget about the pinstripes...

Have you ever watched a movie where they have some computers in the background... it all flickers. that is way more interesting and visible

[Top Fuel F1]

The most exaggerated example of this was in the old Western Cowboy movies where the stage wheels are always turning in the opposite direction to what they normally should be.

[100cc]

thanks people, now that I tried looking at the monitor and saw it flickering, it won't stop, I never payed attention to it before!!!

[klipywitz]

Hey, 100cc, you are welcome!

That is what friends are for!

[Ali_G]

Senna: just to say I often also see this happening.

Under acceleration the wheel apears to be going forward.

But when braking it goes backwards.

I don't have any explanation for it.

Niall

[Arneal]

Congrats to Blkirk on a simple, easy to understand explanation as those are always the best. I'll try to add to it without mucking things up too much.

The incorrect spinning of wheels on TV is a particular instance of a problem class called Aliasing. Aliasing is when a high frequency signal appears to be a low frequency signal due to an insufficient sampling rate. IOW, because you are not sampling often enough your data will show a high frequency signal as a lower frequency signal. Thus a spinning wheel appears to be still, slowly moving, or even moving backwards.

To get a good grasp of this I would suggest that you draw a sine wave and simply examine what happens at different sampling rates. It is old fashioned, but effective.

On TV aliasing is easy to spot, in other applications it is not. Thus methods of defeating aliasing needed to be developed. Enter H. Nyquist who came up with what is known as Nyquist's Theorem. It says you must sample at a rate at least double the frequency of the highest frequency signal you want to accurately recreate.

Obviously it is not always practical to do that. In those cases you can use an anti-aliasing filter which is simply a low pass filter that attenuates frequencies above your target. With the higher frequencies eliminated you can then sample the data without undue worry about alaising errors. Yes, you lose data, but at least the data you have is accurate.

In the immortal words of MattPete, sorry to bore you.

[Ursus]

Originally posted by Arneal


On TV aliasing is easy to spot, in other applications it is not. Thus methods of defeating aliasing needed to be developed. A couple of dead white male's named Shannon and Nyquist came up with what is known as Shannon's Theorem. It says you must sample at a rate at least double the frequency of the highest frequency signal you want to accurately recreate.

IIRC that's the Nyquist Theorem. The Shannon theorem gives the maximum capacity of a comm's link as function of bandwidth and signal to noise ratio.

[SennaVsProst]

Originally posted by Ursus


IIRC that's the Nyquist Theorem. The Shannon theorem gives the maximum capacity of a comm's link as function of bandwidth and signal to noise ratio.

HAHAHA... didnt we all know that?? hhaaa.... damn I am always amazed at some of the stuff people know in this damn forum, the reason I keep coming back!!!

[Arneal]

Whoops, Ursus seems to recall better than I do. Shannon is indeed the bandwidth & S/N guy, Nyquist the sampling theory guy.

[Beej]

Originally posted by MattPete

Ever notice the funky moire patterns that show up when guys wear pinstripe shirts on TV?

The patterns are caused by the fact that colour signal is modulated onto a frequency within the black and white picture. For PAL this is 4.43Mhz (colour sub-carrier) but the black and white band width is 5.5 Mhz. This means that if you get a picture with fine detail i.e a pinstripe suit the frequency clashes with the colour sub-carrier causing the patterning.

Sorry for being a geek

[desmo]

Geeks are not only welcomed, but encouraged here!;)