Technical question about video framerates

[Bleeding Jawa]

Not sure if this is the best forum, but perhaps someone can help me out.

I am looking for information about framerates, using standard consumer video cameras.

I understand that a normal camera will operate at 30 fps.

The application for which I am inquiring is track and field timing. There are manufacturers of timing software that uses standard video cameras, in conjuction with other hardware, that produces a video of the finish line with a time-stamp. The way it works is that the clock is started electronically when the starter shoots the gun. Then the video camera records the finish of the races and is run through a analog-to-digital converter and into a computer where a time-stamp is overlaid on the race finish. An operator watches the finish, frame-by-frame and notes the finish-time for each competitor as they cross the finish line.

The company's the promote these systems claim that they can produce 60 frames per second. I came across a note on one of them claiming "An Eagle Eye digital timing system uses specialized deinterlaced software to capture and replay finish line video at 60 frames per second"

Here is some (sales) information about the technology in question (probably not necessary to read it):
http://www.eagleeyetrack.com/TimingCompare.htm

I do not completely understand the difference between interlaced and non-interlaced video and how that applies to this sort of technology.

What I need to know is, if software is able to split up the frames from 30 to show 60 frames, does time actually elapse between frame 1 and frame 2, or would frame 1 and 2 both show different "parts" of the same exact moment in time?

For example, "Timmy" crosses the finish line on frame XXX and his time is displayed as 10.000 seconds. Then, "Jack" crosses the finish line on frame XXX+1...is his time actually 10.017 (1/60th of a second later), or did he actually finish in 10.033 (1/30th of a second later) but had the image cut up and split between two frames?

I am comparing these timing systems for a major presentation (several hundred people), and I want to make sure I understand the limitations of the video-based systems. The opposing technology CAN accurately produce 100-2000 fps, but I still need to know if the video can produce 60 or only 30 (ie how far off is the margin of error). These companies boast the ability to time to 1/1000th of a second, rounded up the next 1/100th of second. Clearly, neither of these claims is exactly true, but the difference between 0.017 accuracy vs. 0.033 accuracy is pretty significant.

Any thoughts? THANKS!

 

[fuzzybabybunny]

Someone correct me if I'm wrong, for I am not a video guru.

Your answer depends on the video source and the method used to go from 30fps to 60fps.

If you're using a camera shooting at 30fps progressive, then each frame is a different moment in time, 1/30s, 2/30s, 3/30s, etc. Progressive means that the camera is recording the entire picture per unit of time. When you double the framerate you're going to be generating "fictional frames."

Frame 1: 1/30s, runner 2 feet from finish line
Frame 2: 2/30s, runner 1 foot from finish line

When you convert to 60fps, the conversion system could do something called interpolating, basically guessing at what it thinks should be between frame 1 and 2, and frame 2 and 3, and so on. So if in frame 1 the runner was 2ft from the finish line and in frame 2 the runner was 1ft from the finish line, it would add in an extra frame that shows the runner somewhere in between, somewhere between 1 and 2 feet from the finish line. It would "make up" this frame.

Frame 1: 1/60s, the original Frame 1 from your 30fps source, runner 2 feet from finish line
Frame 2: 2/60s, the interpolated, made up, guessed frame, showing the runner 1.5 feet from finish line or something
Frame 3: 3/60s, the original Frame 2 from your 30fps source, runner 1 feet from finish line

So frames 1 and 2 on 30fps progressive will always show different moments in time. When upconverted to 60fps using interpolation from a 30fps progressive source, frame 1 will show a specific moment in time but frame 2 will show a "made up" moment in time, a blend of frames 1 and 3.

Now there's interlaced, the alternative to progressive. If the camera records interlaced video, it records HALF the entire frame per unit of time. Half of a frame, half of an entire image, is called a field. 2 fields make a frame, an entire image. Say that the vertical resolution is 500 pixels and you've got a camera shooting at 30fps interlaced. Field 1 will record lines 1, 3, 5... 499 at t = 1/60s. Field 2 will record lines 2, 4, 6... 500 at t = 2/60s. Then it combines these two fields into a single frame, so frame 1 will have been completed at t = 2/60s, or 1/30s.

In this case, Field 1 of Frame 1 and Field 2 of Frame 1 are different instances in time. Frame 1 is a combination of these different instances in time. When you try to upconvert this stuff to 60fps, the system could try to interpolate the lines of the fields, not the frames as in progressive scan. It would take Field 1 from Frame 1 with all the odd lines and try to fill in all the missing even lines using its logic. Once all the odd lines have been filled, the Field 1 becomes Frame 1, an entire frame, an entire picture. Then it would take Field 2 with all the even lines and try to fill in the missing odd lines, after which Field 2 would now become Frame 2.

Field 1 of Frame 2 would become Frame 3, Field 2 of Frame 2 would become Frame 4, and so on.

So in the case of 30fps interlaced being converted to 60fps, Frame 1 and Frame 2 of the 60fps video will be separate instances of time, but half of each frame will have been "made up" with logic. Only the odd lines of Frame 1 will be "truthful and only the even lines of Frame 2 will be "truthful."

 

[Bleeding Jawa]

I'm 99% sure this is the analog to digital converter that is resold as part of their "package" http://www.firewire-1394.com/canopus-advc-110.htm

The timing software that starts the clock and adds a time stamp is proprietary.

As far as the camera goes, they do not recommend anything in particular, and leave it to the buyer to purchase their own off-the-shelf camera, suggesting that they only cost $300-400 (thus suggesting basic and cheap).

When you talk about "interpolation," "30 fps progressive source," "interlaced and non-interlaced"...is a standard video camera capable of doing any/all of these things, or would something have to happen during the conversion process or later?

Averaging out the pictures and inserting a "made up" frame would at least be a meaningful comprimise, even if it isn't 100% "accurate." However, if the hardware is limited repeating frames and adding intermediate time-stamps to those frames serves no real purpose and is entirely inaccurate. I guess I am just trying to find out what the hardware limitations are so that I can better distiguish between the actual technical "features" of the system and the deceptive marketing gimmicks...both are clearly present but hard to seperate.

In this case, Field 1 of Frame 1 and Field 2 of Frame 1 are different instances in time. Frame 1 is a combination of these different instances in time. When you try to upconvert this stuff to 60fps, the system could try to interpolate the lines of the fields, not the frames as in progressive scan. It would take Field 1 from Frame 1 with all the odd lines and try to fill in all the missing even lines using its logic. Once all the odd lines have been filled, the Field 1 becomes Frame 1, an entire frame, an entire picture. Then it would take Field 2 with all the even lines and try to fill in the missing odd lines, after which Field 2 would now become Frame 2.

Now this option makes sense. But, is this how a typical camera would take in video to allow for this sort of seperation later down the line?

 

[fuzzybabybunny]

I'm going to guess that when you're buying the camera you should pay attention to whether it shoots progressive, interlaced, or both. Older basic cameras probably would just shoot interlaced. Newer basic cameras would probably shoot progressive. And just because I illustrated two ways to go from 30fps to 60fps doesn't mean that this company actually uses any of these methods.

I don't know of any cameras that do their own interpolation, as this is often done in post processing. I'm pretty sure that there ARE cameras out there that shoot natively in 60fps though.

 

[FP]

I believe most consumer camcorders capture at 30 frames per second however that frame is made up of 2 fields. The fields are interlaced so Field 1 is recorded a split second before Field 2 and put together into a single frame.

Now, how could you make this 30 frames per second video into 60 frames per second? A couple of ways... Double the fields or double the frames. Either way there is no "new" data being introduced. They are simply doubling something that already exists.

To answer your question... the 30fps video will show the same images as the 60fps video. Whether there is a time difference between each frame of the 60fps video depends on how they change the framerate internally.

More info can be found here

 

[laptoppowerbattery]

acer 91.45B28.001 battery for acer 91.45B28.001 laptop

 

[Fardringle]

Originally posted by: laptoppowerbattery
acer 91.45B28.001 battery for acer 91.45B28.001 laptop

Thanks for sharing. Now please go away.

p.s. If you are going to anonymously spam message boards with overpriced laptop batteries, you might want to try a forum that actually talks about laptops next time..

 

[Bleeding Jawa]

Thanks for the link FP. It helped...I think. (*Note: It did drive me nuts that they were talking about 25 / 50 fps instead of 30/60 all the way through until I got to the very end and read the noting saying they were discussing PAL instead of NTSC).


From what I am gathering, the 2 separate fields or the 2 separate sets of scan lines ARE recorded at different times and thus "should" represent a consistently measurable time change between the interlaced frames at 60fps. Make sense?

So maybe what I need to learn now is exactly what methods are being used by the analog-to-digital converter (or the computer software) in order to de-interlace the pictures. Again...make sense?

 

[FP]

Well that depends on your goal... Are you simply trying to determine the time difference between two frames/fields?

The ADC shouldn't matter... it will convert what it is given.

The important question like you said is how their software is converting the framerate.

If you are able to view the footage frame by frame it would be easy to determine.

 

[Bleeding Jawa]

I guess the ultimate question is this =

Once the video is converted to 60 fps/ips, does each "frame" actually display a picture that represents a distinct moment of elapsed time? I am gathering that it does.

ie
Frame One = Displays an accurate picture of what was actually happening at 0.01668335 seconds.
Frame Two = Displays an accurate picture of what was actually happening at 0.0333667 seconds.
Frame Three = Displays an accurate picture of what was actually happening at 0.05005005 seconds.

What (I am presuming) is NOT happening = Frame 2 displays a TIME of 0.033 seconds, but is actually displaying part of a picture scanned at the 0.166 mark in time.

 

[Fardringle]

Originally posted by: Bleeding Jawa
I guess the ultimate question is this =

Once the video is converted to 60 fps/ips, does each "frame" actually display a picture that represents a distinct moment of elapsed time? I am gathering that it does.

ie
Frame One = Displays an accurate picture of what was actually happening at 0.01668335 seconds.
Frame Two = Displays an accurate picture of what was actually happening at 0.0333667 seconds.
Frame Three = Displays an accurate picture of what was actually happening at 0.05005005 seconds.

What (I am presuming) is NOT happening = Frame 2 displays a TIME of 0.033 seconds, but is actually displaying part of a picture scanned at the 0.166 mark in time.

As others have said, it's technically possible if the original video source is interlaced (i.e. 1/2 frame recorded at a time at 60 half frames per second, or 30 full frames per second). The question is whether the software you are looking at actually interpolates full frames from each half frame in order to show a real difference in time between each frame, or if they use the full frames from the 30fps source and "create" new frames between each full frame by averaging the two frames. Only the creater of the software (or someone that has reviewed it properly) can tell you that for sure.