SDI does NOT stand for "Short Distance Interface." But, a serial digital interface run can revert to that status if some basic rules are not followed.
The Serial Digital Interface-SDI (SMPTE 259M) grew out of the need for longer distance connection of component digital television equipment, the result being the viability of a truly digital broadcast station. SDI is capable of running hundreds of feet and can run thousands of feet if properly distributed. For additional information on SDI, see my articles in the January - April 2000 and the September - October 2000 ExtroNews.
Taking It One Bit At A Time
Digital component recording began in 1987 with the creation of the D1 format (SMPTE 125M). The D1 interface is an 8/10 bit parallel system intended for close-in connection between digital tape recorders (19 mm tape). Its interface cabling is short due to the difficulty in maintaining proper bit timing over a byte-wide data channel. Somewhat like DVI, D1 requires management of differential signals over 8 or 10 twisted pairs. Bit skew, crosstalk, and attenuation are adversarial to the task of transmitting parallel D1 for long distances. The interface uses a 25-pin D-sub miniature connector. As a result, termination is not really easy, and the thought of managing that many parallel bits through a router is good for a migraine, not to mention the hardware cost of 10 switching planes. Therefore, parallel D1 connections are easily managed over only a few meters.
Reformatting the byte-wide D1 data via a serializer yields a very high-speed serial data stream. Serializing a 10-bit data word results in a data rate ten times faster. The 27 MHz D1 data becomes serial data at 270 megabits per second for standard component NTSC. See Figure 1 (below) showing the basic conversion methodology. Benchmark signal performance is captured in Figure 2.
Why SDI?
Although SDI bit rates are very high, distribution of serial data as a single cable connection presents significant advantages. First, it's much easier (read cheaper) to route and switch one cable than a parallel system of cables. Having all data bits organized as one stream means there will be no issues with clock and data synchronization. Managing bit timing and cable equalization is easier. Data skew problems encountered with multi-conductor cables do not exist.
As seen in the operational diagram, Figure 1, the SDI format utilizes a differential signaling technique and NRZI (non-return to zero inverted) coding. Although SDI is transmitted as an unbalanced signal on 75-ohm coax, transmission and reception involves differential amplifiers that format and detect, respectively, both data phases. Utilizing differential reception creates additional headroom and robustness in signal-to-noise performance. Pseudo-randomizing the data bits and use of NRZI coding increases channel transmission reliability. NRZI coding is desirable because its operation is independent of signal polarity. In this coding scheme, high and low levels do not communicate data 1s or 0s. High and low states are detected simply by the change from one level to another. A zero means that the transmission level stays the same, while a one is transmitted each time the level transitions from one level to the other.
SDI is more immune to extraneous noise and low frequency components (hum) because the receiver takes one phase of the data transmission, inverts it, and then adds it to the in-phase portion. Like a regular analog differential amplifier, common mode noise induced into the signal is cancelled out during this inversion and addition operation.
So, what problems do exist? As in life, all modes of travel have distinct advantages and disadvantages. One must weigh the relative difference. Key factors affecting SDI are cable attenuation, signal jitter, signal wander, error detection/handling (EDH), and receiver sensitivity. See Table 1 for a list of the SDI rates supported within SMPTE 259M.
Cable Quality is Job 1
The single largest effect on SDI transmission rests with the quality of cable used relative to the transmission distance required. Any 75-ohm coaxial cable may be used for SDI. The big question is always: "How far can I go?"
SMPTE 259M guides us in determining cable transmission length. It states that, for a class A receiver (the best type to have), the maximum transmission distance is given by a coaxial cable length having 30dB attenuation at one-half the SDI clock rate. For example, at the 270 Mbps rate for component NTSC, one half the rate is 135 MHz. Many cable specification tables show attenuation in dB at 135 MHz since this is a popular rate. Taking the attenuation value for a 100-foot cable at 135 MHz and scaling it the 30dB limit (attenuation is linear) gives us the maximum cable length. If we have a specified loss of 10dB at 135 MHz at 100 feet, then the maximum usable length will be at 30dB divided by 10dB times 100 feet, which is 300 feet for that cable.
Utilizing the maximum calculated cable length in a primary distribution run for SDI is NOT a good idea. Suppose you have made the maximum length run. Now, you connect a 10-foot patch cable at the end to include some other device and, suddenly, there is no video image! You have just experienced the "cliff effect." When the loss parameters of the SDI signal exceed the receiver's ability to recapture the data, the system completely fails, ungracefully. For this reason, allow at least 10% margin in your cable length calculations to account for other connection changes, connector resistances, connector and termination reflections, etc. Most pre-calculated cable charts build in this allowance just for good practice. See Table 2 for calculated cable lengths for Extron cables.
There's Clocking, And Then There's Re-clocking
All digital data is derived and managed by a repetitive pulse train called a clock… the literal heartbeat of the machine. Without it, data transitions could not be identified in a coherent way. Either, digital data somehow contains the clock information embedded within it, or the clock signal accompanies the data separately. Since SDI is a singular wire transmission scheme, the clock is embedded. Therefore, not only does cable attenuation affect recovery of data, it seriously affects the receiver's ability to recover the clock signal such that the system can stay synchronized.
This is where basic cable attenuation comes in. The maximum cable distance is governed by the receiver's ability to recover clock and data reliably. As the digital cliff is approached, bit errors typically appear and escalate rapidly toward transmission failure. But, (you say) I need to transmit SDI over 1,000 feet and the best cable I have available is not guaranteed much beyond that. What will I do?
The solution is straightforward. Position an SDI receiving device in the line at a point where reliable communication is maintained. Make sure this device is a true SDI receiver that can equalize and re-clock the signal. In analog signal transmission systems there is no really good way to reform and transmit the signal while still maintaining good linearity. With digital data streams, however, the data can be captured and reconstructed by a "squaring" circuit that restores the original rise time of the signal. Because we aren't concerned about linearity in digital data, the data can be indefinitely reconditioned as long as good signal conditioning practices are followed. When SDI is reconditioned for retransmission, the data edges are sped up and the original timing accuracy restored. This operation is referred to as "re-clocking." Now, we can run the additional distance our cabling system will allow.
To Re-clock Or Not To Re-clock
Yes, that is a question. Properly re-clocking SDI calls for additional circuit complexity and cost. Good SDI matrix routers include re-clocking systems. Further, it's a good idea to consider the location of a router in a new installation such that cabling distances can work to your advantage toward minimizing the number of repeating stations, or re-clocking points, required. Routers are typically the focal point for re-clocking data. SDI distribution amplifiers can involve a receiver/re-clocker circuit as well. In major installations having large routers, it is common for the router to have re-clocking ability at the input and at the output as well. Why? The actual propagation distance and loss effects through some large routers represent a serious impairment to SDI signals. Therefore, re-clocking may be included at the output to ensure signal quality for the next long cable run.
Now, back to the question. Do I need to always re-clock SDI signal runs? No. Suppose you have a relatively small set of cable runs involving a small matrix router and good receivers at each destination. Good SDI receivers can recover the signal under some surprising signal degradation conditions. In some cases, adding the wrong equipment into the line (or something having a poor re-clocking system) may actually increase signal jitter, which makes recovery more difficult. As long as you carefully maintain signal quality and are not in danger of exceeding cabling distances, you do not necessarily need to re-clock. Re-clocking is primarily intended to clean up long run losses, allow easier decoding, and re-drive additional cable runs.
An analog matrix router under the right conditions can handle SDI nicely. The key issue is whether the router will introduce crosstalk or other noise that may affect the signal jitter performance. If the router is bandwidth limited, performance will be directly affected as high frequencies are attenuated rapidly. You must look at router bandwidth specs to see if the SDI rise time can be accommodated without significant effect. Since the SDI signal is about 800 millivolts peak-to-peak, it is not much different in level than standard video signals. SDI is uni-directional and has a good signal-to-noise recovery budget. So, with care, smaller, local routing systems can work on a budget without re-clocking.
Got The Jitters?
Signal jitter is another culprit in SDI systems. Maintenance of the timing relationship to a common timing reference provides auto-phasing recovery circuits in the receiver the ability to lock onto and decode the clock and video data. When an external factor, such as random noise, affects the absolute bit timing, the receiver encounters difficulty recovering clock and data. Cable loss affects the amplitude of the SDI signal while jitter affects the zero crossing point of the data edges. The data edges appear to dance back and forth with random uncertainty. There is a jitter budget allowance, but since noise and jitter effects can become generally random, bit error rate can creep up periodically and cause lost data. If the jitter budget is exceeded, data cannot be recovered at all.
As with analog signals, once you have noise in the signal, it is extremely difficult and costly to remove. Jitter caused by induced noise effects, unstable signal sources, or poor re-clocking systems is the demise of digital signals. Sometimes, basic signal attenuation effects are mistaken as signal jitter. SDI signals contain a range of low to high frequencies like analog signals. Cable attenuation still affects the high frequencies most. When looking at an eye pattern, the data zero crossing point (risetime/falltime area) appears wider than normal. The eye pattern is typically used to evaluate signal quality including jitter. This appears to smear the data edges and look as though large amounts of jitter are present, when, in fact, measurement with SDI measurement equipment may show the signal well within jitter specifications. Jitter measurements should be made with instrumentation capable of proper measurement. SMPTE Engineering Guideline, EG-33-1998, Jitter Characteristics and Measurements provides in-depth help for this task (www.smpte.org). See Figure 3.
Born To Wander
With the deployment of more digital video networks, the monitoring of video sync timing is more critical than before. In some applications, where time-base correctors or frame synchronizers are not used, problems with image shifts and hue errors may occur because of network induced wander of sync and color burst timing. This condition creates "video wander," which is defined as sync signal phase variations below 10 Hertz. When the video signal is converted to composite, this effect is not easy to remove.1 Specialized television test equipment, such as the Tektronix VM700T, can easily measure horizontal sync timing jitter and wander for serial digital systems.
Digital Safety Net
All of the aforementioned situations in addition to poor connections and improper terminations can cause data bit errors to occur. A bit error is defined as a change in one or more data values occurring between the source and destination. SDI includes an error detection and handling (EDH) system that can monitor data quality and provide some visibility of errors as well as location. Some bit errors may not affect picture quality directly but may signal impending failure. Groupings of bit errors may affect picture quality, sound, or both. SDI equipment will typically incorporate some level of EDH reporting or troubleshooting capability.2
Receivers - Some Are So Insensitive To Your Needs
SMPTE 259M mentions a typical range of expected SDI receiver sensitivity between 20dB and 30dB at one-half the data clock frequency. Further, proper cable equalization should be employed. What is cable equalization? It is a feature of the receiver's front-end amplifier that adjusts its gain to compensate for higher losses in the signal at the higher frequencies received, while maintaining lower gain settings for the lower frequencies. This is important for proper alignment and triggering with changing data edges so as to ensure consistent video recovery.
Looking back at our earlier example we used for calculating maximum cable length, our cable run was 300 feet for a 10dB loss cable at 135 MHz at 100 feet. Now, if we have a less sensitive receiver, say the 20dB type, our drive distance will decrease to 20dB divided by 10dB times 100 feet, or, only 200 feet. You can see there is a ratio of 2/3 here. This nominal 10dB performance spread in receivers severely limits SDI cable run lengths.
Remember, all things being equal, pay careful attention to receiver sensitivity and cable attenuation specs for realizing the most from SDI signal distribution. While good routers utilize re-clocking, the need for this feature primarily depends on the size and complexity of your system design. Be aware that good signal sources, routers, and proper cable routing techniques help reduce the invasion of signal jitter.
Footnotes: 1. Measuring Wander in Video Distribution Systems, by Tom Tucker, Tektronix
(http://www.Tektronix.com/Measurement/App_Notes/Published_Articles/measwander/index.html)
2. Monitoring in the Digital Environment, by Ted Gary, Broadcast Engineering, November 1999 (http://www.broadcastengineering.com/archives/1199/199911be40.html)