SDI is the method that is used to move pictures (and sound but-let's-not-get-ahead-of-ourselves) around a digital TV plant. Let's attack the SDI acronym from the rear:
First, since this is an electrical signal, it requires a source, and a return path for the electricity, so when we say "one wire" we really mean two, and when we say "eight wires" we really mean nine (the signals can share a common "ground" return."Ten wires" means.... (did everyone get eleven?).
|
|
In our familiar co-axial cable, the shield functions as the ground return for the electrical signal. |
Above is a second look at digitizing an analog signal. The black-box is an analog to digital converter, known as an "A-to-D", and used in sentences like "Maybe the A-to-D is bad". Out of the A-to-D marches groups of ones and zeros (actually electrical pulses that are or are-not there). This converter is using eight-bit bytes (numbers 0 to 255) that are highlighted in alternate cyan and white strips for clarity. The eight bits exist simultaneously as a parallel "byte".
Parallel data are efficient, because all of the bits are available at once, but it takes at least ten wires to carry an eight bit parallel signal. "Ten" isn't a typo... one's the common return, and an extra one is required for the sender to signal when the bits are valid... there is a time while the bits change from one byte to the next when the bits aren't necessarily correct.
Those big ribbon cables that you see inside a computer carry parallel data.
It is desirable to keep our new digital video signal all on one wire (imagine nine times the wires in the racks! Not to mention the expense). The solution is to send the bits serially. There isn't anything really tough about this, but to move the same amount of data, the serial stream has to run about ten times faster than the parallel connection. Why ten times? Well, it's not exactly ten, but that's a good ballpark number. If you look at the output of the serial to parallel converter above, you'll notice that it's an indecipherable string of ones and zeros. Where does each byte begin? The correct answer is "I don't know". Serial streams need gaps, start-bits, and stop bits to define the individual groups of bits. This will be familiar if you have ever set up a modem for a dial-up computer connection.
|
|
|
Here's the same string of bits, but below it is a series of short pulses that show where each bit resides. This is known as a "clock". Uh-oh, we're back to two wires. |
|
This is how SDI solves the problem on only one wire. The pulse train above is greatly expanded in time (horizontally). Zeros are equal to the clock time, and ones have an extra transition during the clock time. As long as we know what the clock time is supposed to be, we can tell a zero from a one. If you're happy with this explanation, don't read the next paragraph. Notice that the two zeros at the right side of the graphic have no transition between clock periods, which is correct, but one of the zeros is an electrical "high" level, and the other is "low". Wow. Following the rule: "Zeros are equal to the clock time, and ones have an extra transition during the clock time", means that the actual polarity of the signal doesn't matter. This is pretty elegant. Inquiring minds might wonder how we can be sure that we can recover the clock signal that we need to decode this. The dashed lines in the graphic show the clock period. Our clock recovery circuit locks to any transitions that are the correct distance apart. "AHA!" you say... what if we lock to the wrong edge, say like the transition in the first one? We will know that we have the wrong transition edge when we hit the first zero in the string, and the data in SDI is shuffled to make sure that there are lots of zeros and ones. |
|
|
This is the SDI signal as seen on a waveform monitor. Oscilloscopes
don't freeze a moment in time, they scan constantly, so the SDI pulse
stream is all overlapped in this display.
This display is called the "Eye" display on a SDI waveform Monitor. |
|
|
The red arrows highlight a double-speed "one" transition. |
|
|
Here is a longer "zero" transition. |
|
|
The transitions take some time to get from an electrical high to low, and vice-versa. This time gets longer (more "slope-y") in longer runs of cable. In this example, the up transition and the down transition cross midway, so a clock recovery circuit will probably recover a good, accurate clock. |
|
This is the eye after about 100 meters of cable... still recoverable. (It's a different scale than the above) |
The same signal after 200 meters of cable... getting close to the limit, but still useable. |
|
This eye diagram shows lots of "JITTER". As the straight edges of the pulses get to be more of a slope, the actual moment in time when the transition takes place becomes uncertain, and the recieving equipment recreates pulses with time jitter. |
Since a perfect picture can be displayed as long as the equipment can tell the difference between an electrical high and low, the actual SDI signal can get worse and worse during its travel down long cables and through pieces of equipment, and still display a perfect picture... until that last patch cord is inserted in the path and the picture goes away completely! Thus the cliff analogy... all is well until you suddenly fall off the edge.
Cables have electrical components called capacitance and inductance. Capacitance resists change, and inductance causes abrupt changes to "ring" The above string of pulses shows the effects of both. The transitions from low to high take a significant amount of time, and the peaks of the pulses are wavy.
Various blocks in the signal path like distribution amplifiers, routing switchers, etc. can pass the signal in different ways. The equipment can PASS the signal, making it strong enough to drive more cable, but containing the defects that it arrived with (to be made worse by further lengths of cable). The equipment can SHAPE the signal, stretching it and clipping the dirt off of the top and bottom. The shaping circuit will make a decision about where the actual up/down transition is, resulting in a good-looking pulse train that has time-jitter in the transitions.
Re-Clocking devices recover the clock component with a Phase Locked Loop circuit that acts as a flywheel to extract a stable, average clock signal from a jittery train. It then uses the new, stable clock to regenerate the pulses with accurate edges-in-time. The flywheel analogy is a good one, as the flywheel on a lawnmower engine keeps the blades circling evenly as the engine putts, often unevenly (at least the author's does!). A re-clocked signal is "as good as new".
There is a lot more to say about SDI, but at this point, it's just a way to get video from A to B. With analog video, the picture was always worse after a long cable run, a recording, or even a pass through a switcher or effects unit. SDI carries perfect pictures from device to device, as long as we keep the pulse shape and jitter under control.
There are different flavours of SDI... the most common codes Y, R-Y, and B-Y components and calls them Y, Pr and Pb. One form of SDI actually codes the composite signal, subcarrier and all!
The component form (Y, Pr, Pb) samples the Y channel at the 13.5 Mhz rate, and the colour components at half of that... 6.75Mhz. Adding 13.5, and 6.75 twice results in samples occurring 27 million bits per second, which needs to be ten times as fast for serial transmission, so component SDI runs at a serial data rate of 270 megabits. (Yikes!)
Component SDI uses 10 bit samples for the colour components. 10 bits results in 1,024 analog electrical levels (2 to the 10th. power)
SDI video is carried on video cables, using the familiar BNC connectors.
Component SDI uses the very core of the NTSC system matrix and Y, R-Y, B-Y system standardized in 1954. The two colour components (R-Y and B-Y are re-named Pr and Pb for "phase-red and phase-blue", even though there is no "phase" thing any more... there is no subarrier or colour-burst.
SDI is a series of digital pulses that can be decoded into pictures and more, so it might be natural to think of it as being like say, an MPEG movie that gets sent over the internet or even a Firewire™ transfer from a camcorder to a computer. There is an important difference. The preceding examples are file transfers. In a file transfer, all of the bits that make up a movie or clip are sent as fast as they can be sent, for assembly later. I some cases, they can start to play when enough of the file has been received to buffer the speeding up and slowing down of the data path as the internet or local computer handles tasks as quickly as it can. Even home DVD players freeze up occasionally as they encounter patches of errors. File transfers rely on start-of-file and end-of-file markers in the data, and usually take place at any speed.
SDI is a continuous stream of video. The data flows continuously with no beginning or end. The streams can be switched like analog video. As well, the data proceeds at one pace (although "pace" doesn't do 270 million bits per second justice).
Component SDI is the simple digitization of analog component video. Much of the analog video system was "internally digital" for a long time.... DVE's, TBC's, and Framesyncs all used their own A to D converters, processed the signal digitally, and output analog through their own D to A converters. SDI adds the final needed link to connect the units together without repeated conversions to and from analog.
HOME | BACK | HOW IT WORKS | TIMING | RECORDING | EDITING & TIMECODE | AUDIO | COMPUTERS
w
Of course, there's more. Above is a bunch of serial bits. The low spots are zeros, and the high spots are ones. How many zeros are there in
the section with the red arrow? Again, the correct answer is "I don't
know". We need a signal that marks off the position of each bit.
