During the Olympics, all video feeds were sent between venues at 270 MBit/Sec uncompressed. Using this as a baseline for the technology of the Superbowl and multiplying by the 33 cameras, so we are dealing with 8.91 GBit/sec, which is a lot of data in anyones language.
Of course the argument is that you do not need to store the 8.91 GBit/Sec, but you do, since you want to be able to go back to chose a frame to start many seconds later.
The computing power required for the storage is amazing. Of course if you compress each frame things become more sensable. Then we come down to about 100-200 KByte/frame, or 2.5 - 5 MBytes/sec which is more realistic in terms of computing power.
I read that all the cameras are focused and zoomed onto a particular point on the field, but shoot from different angles. There are two ways to get the effects here...
* Firstly grab the same frame from each of the image streams and play them back in real time in order. This will generate the effect of stopped motion rotate [Frame x from camera 1, then frame x from camera 2, etc]
* Change frame and camera each time [Frame x from camera 1, frame x+1 from camera 2... etc]. This will create the illusion of the slow motion rotate.
I do not think you would need to do any morphs to get this to work. It would appear as if the camera was moving rather than many cameras being used. All the zooms are synchronised, but I think the depth of field would vary from shot to shot. This would not matter since it would just add to the effect.
Another example of this sort of technology is the 2000 film staring Bjork [I cannot remember the name] which used 100 cameras simultaniously to film the action.
Slashdot Mirror
During the Olympics, all video feeds were sent between venues at 270 MBit/Sec uncompressed. Using this as a baseline for the technology of the Superbowl and multiplying by the 33 cameras, so we are dealing with 8.91 GBit/sec, which is a lot of data in anyones language.
Of course the argument is that you do not need to store the 8.91 GBit/Sec, but you do, since you want to be able to go back to chose a frame to start many seconds later.
The computing power required for the storage is amazing. Of course if you compress each frame things become more sensable. Then we come down to about 100-200 KByte/frame, or 2.5 - 5 MBytes/sec which is more realistic in terms of computing power.
I read that all the cameras are focused and zoomed onto a particular point on the field, but shoot from different angles. There are two ways to get the effects here...
* Firstly grab the same frame from each of the image streams and play them back in real time in order. This will generate the effect of stopped motion rotate [Frame x from camera 1, then frame x from camera 2, etc]
* Change frame and camera each time [Frame x from camera 1, frame x+1 from camera 2... etc]. This will create the illusion of the slow motion rotate.
I do not think you would need to do any morphs to get this to work. It would appear as if the camera was moving rather than many cameras being used. All the zooms are synchronised, but I think the depth of field would vary from shot to shot. This would not matter since it would just add to the effect.
Another example of this sort of technology is the 2000 film staring Bjork [I cannot remember the name] which used 100 cameras simultaniously to film the action.
Darryl VK2TDS
darryl@radio-active.net.au