As the cost of digital video processing decreases it becomes possible
to integrate devices such as video frame synchronizers, noise reducers,
color correctors, and digital effects to produce high performance and
cost effective production switchers and systems.
Since many switchers integrate video frame synchronizers in their inputs,
a variable video delay which ranges from almost none to one frame is added
to the video. In addition, many DVE circuits add an additional delay of
one frame, and this delay is changed as the DVE is switched in and out
of operation. It is not uncommon for these production systems to add variable
delays to the video which range from a few hundred microseconds to over
4 frames, depending on the path the video signal takes through the switcher.
Obviously, a delay of 4 frames will create a lip sync problem. Even delays
of 1 frame can be noticeable and will detract from the program quality.
In order to prevent this problem, the audio program which is associated
with the video signal being output from the production system must be
delayed to match the video delay, but how is this to be accomplished when
the delay is constantly changing? When a typical video production installation
is analyzed, it will often be the case that there is a separate audio
source associated with each video input. For example, Video1 might be
a network feed from a Satellite and Video2 and Video3 might be from two
VTRs. It is important that as each video feed is passed through the production
switcher that it’s corresponding audio be delayed to match the video
delay. This can be accomplished by using a DD2100
to measure the delay and AD3100 to delay the
audio.
Each video input is connected to the undelayed input of a DD2100
delay detector, with the output of the production switcher coupled to
the delayed inputs of all of the DD2100s. The
delay detector can then measure the delay of each video signal and provide
a steering signal to a companion AD3100.
For example, when the network video signal is selected by the video switcher,
the DD2100 will compare the undelayed network
video which is input to the switcher to the delayed network video which
is output from the switcher. The DD2100 compares
the two video signals with a powerful correlation technique to measure
the delay between them.
The delay steering pulse from the delay detector then causes the companion
AD3100 to delay the network audio by a matching
amount. The delayed network audio is then passed on to the program switcher
where is selected and sent on to the transmitter or recording tape machine.
Similarly, when VTR1 or VTR2 is selected, its corresponding audio from
that source is delayed by the proper amount. The operation of the DD2100
is specifically designed for applications such as these, and has several
design enhancements which make it highly desirable for such use.
Some of the questions which frequently come to
mind are answered below:
Q. What happens when the video delay undergoes
an instant change, such as when the DVE is switched in, or when the input
video synchronizer undergoes a pointer crossing?
A. Instant delay changes, especially pointer
crossings in four field synchronizers, can cause delay changes up to 66ms.
This large change places a difficult burden on the audio synchronizer
since it must acquire the new delay value without losing audio, causing
pops and clicks or pitch artifacts. Older types of audio synchronizers
handled the problem by simply jumping to the new delay value in small
1ms or so steps, but this created pops and clicks. Fortunately these artifacts
only lasted for a few seconds, at best and up to a minute in extreme instances.
Most audio synchronizers control pitch artifacts by combinations of limiting
the rate of delay change and/or making changes during silence periods.
As a consequence of this action, when rapid or instant video delay changes
occurred, the audio synchronizer would take several seconds to chase the
new delay. If the audio program contained music or other continuous audio,
this slowed the rate of change so that it could take a few minutes for
the audio synchronizer to catch up to the new delay. During that period
the lip sync would be out by a noticeable amount.
The AD3100 has a pitch correction circuit which
makes it capable of making delay changes of over 2 frames in less then
1 second without any missing audio, pops, clicks, pitch change or other
artifacts. This capability allows extremely fast response to instant changes,
keeping audio in sync at all times.
Q. What if there are video feeds from character
generators or other sources when no delay is needed?
A. There is a two wire remote control dump
which forces the delay to zero. This can be wired to the tally system,
or other switch contacts to provide this capability.
Q. What happens to the DD2100
when it does not have the same video present at the delayed and undelayed
inputs?
A. In this instance, the DD2100
instantly recognizes that there is no correlation at all between the two
video signals, and the delay output is held at the last valid value. This
value is held until the DD2100 senses matching
delayed and undelayed video signals again at which time it again calculates
the proper delay.
Q. What happens if a color corrector, noise
reducer or proc amp is switched into the production system to make video
corrections?
A. As long as the delayed video signal is
reasonably similar to the input video signal, the DD2100
correlation is capable of determining the delay. Changes to the delayed
video such as noise reduction, color correction, hue, saturation, luma
gain and luma offset frequently have no negative effect on the correlation
capability of the DD2100.
Q. What happens if a key or matt is created
in the delayed video?
A. Again, as long as the delayed video signal
is similar to the input video signal, the DD2100
correlation is capable of determining the delay. Normal keying & matting
of station logos and titles frequently have no negative effect on the
correlation capability of the DD2100.
Q. What happens when the video signal is
altered by rotation or other special effects?
A. Effects which spatially alter the video,
such as rotations, flips and compression are detected as gross changes
to the video and the last valid delay value before the effect is held
by the DD2100 until the video returns to its
normal state. It might be noted here that such gross changes rarely last
for more than a few seconds and the change of video delay during that
time is usually quite small. In addition, it is difficult for viewers
to detect any lip sync error on video images which have experienced these
effects.
Q. Will the DD2100
operate with live camera feeds that are chroma keyed, such as with weather
caster which is chroma keyed over a weather map?
A. With effects such as chroma keys the
DD2100 is capable of detecting delays providing
a significant portion of the undelayed signal is keyed into the delayed
signal. If it is desired to guarantee operation of the DD2100
under all key conditions, such as when only a small amount of the undelayed
video is keyed into the delayed video, it is suggested to use a switcher
ME output or a preview output which is upstream of the chroma keyer and
the delayed video input to the DD2100. If necessary,
a fixed offset can be added to the AD3100 to
compensate for any delay which the chroma key adds to the video, however
chroma key delays are usually very small.
Q. Why can't one delay detector be used
for all the video signals?
A. Each video signal is asynchronous, and
undergoes a unique delay in its frame synchronizer. Consequently, a delay
detector is needed for each source. It is however possible to lock a single
AD3100 to a tally driven delay pulse generator
DG1200 to compensate for the average delay of
each video signal through the production system. Contact Pixel's sales
personnel for further information on this system.
It is possible to reduce the number of audio synchronizers which are used
in the system by detecting which of the delay detectors is active from
a status signal provided thereby, and switching the output of the active
delay detector to the input of a single audio synchronizer. This system
of course requires that the current audio signal be coupled through the
audio synchronizer.