Tracking Delay

graphics/overview_tracking_delay.png
graphics/overview_tracking_delay_2.png

After reading the last section about Video Delay you may have the impression that tracking systems and cameras sample the position or record the image exact at the flank of the sync signal (blackburst). This is, however, not the case. Depending on the exposure time, iris and gain we do not know when the camera is recording the image. We only know for certain that the camera will send the image out at the next sync. But we don’t know at what time during the field the image was recorded. Still worse, if you have a long exposure time - lets say 2 milliseconds - where is the point in time, the image was recorded?

The same problem occurs with tracking systems. Mechanical tracking systems are alway near the sync flank - but they are never exact at the flank. They always need a few milliseconds to read all the encoders and calculate the position from them. Using optical tracking systems compounds these problems. At what point in time are the cameras shooting the targets? We do not know. The only thing we know for certain is that there is always a fraction of a field delay between the tracking data and the video data, even when the video delay is set as accurately as possible.

From the last chapter we know, that video delay can only be set in whole values of frames. How do we set the fraction of a field?

Note: When the delay differs 0.1 field from the video it is visible to the eye!

The only way to solve this problem is to interpolate using the tracking data received by the Tracking Hub. There is an Overall delay setting in the Studio Manager RIG control that can be used to interpolate the tracking data in fractions of fields. In addition, every axis can have its own delay bias. This is useful, when mixing different tracking systems. When the overall delay is set to 0.5, the tracking hub interpolates a half field back in time.

But what can you do if the delay must be set to -0.5 to achieve a good result? This is not possible. We cannot look into a crystal ball and attempt to extrapolate the data, because this will probably result in ugly jumps. In this case, the only solution is to increase the video delay.