Electronic Design

For Better Analog Video, Try Differential Signaling

Compared to single-ended signaling, differential signaling offers many benefits: less electromagnetic interference (EMI), less distortion, lower supply voltagei, and lower costii. These advantages have prompted the adoption of differential signaling in many applications, including digital (low-voltage differential signaling, or LVDS) and analog (audio). Similar benefits should accrue to analog videoiii. But for reasons that may no longer be valid, most video applications continue to use a single-ended 75- coax connection.

By leveraging differential signaling and low-cost CAT5 networking cabling, analog video can be distributed over significant distances, reducing the cost of video interconnects and distribution. CAT5 cabling could be cost-effective in common applications like security systems, closed-circuit television (CCTV), and automotive systems.

If a system configuration has a lot of sources like cameras, especially for security or CCTV, video over Internet Protocol probably will be a better choice. That’s because the total cost of the system can amortize the cost of network hardware, video compression hardware/software, and software necessary to control and maintain all systems. Yet in systems with fewer sources and when the total cost of the system is important, analog video interconnects deliver the best cost/performance tradeoff since such systems don’t require compression or any software control.

There hasn’t been a lot of momentum toward the use of differential signaling. One reason for this lack of inertia may be a lack of understanding of how differential circuits operate and are specified. To help remedy that deficiency, the concept of balance can provide an intuitive understanding of differential circuits and help compare them to more familiar single-ended types.

The concept of balance has various names, depending on the part of a differential circuit in which it’s used. It isn’t specified at all in some cases, like most analog differential drivers. But the term is necessary, and other parts of the differential circuit can modify its importance.

An impediment to the implementation of differential analog video has been the absence of a low-cost transmitter-receiver IC that converts, transmits, and receives video over a differential cable while also providing an interface to existing 75- single-ended video circuits. This capability requires a lot of circuitry, including (at minimum) an active balun circuit with amplitude equalization to compensate for the cable loss, as well as clampingiv or biasing to stabilize the video black level. This IC also would need I/O protection, single-supply operation, fault diagnostics, and low cost.

Such an IC is now available, so next, we’ll look at a typical application. To illustrate what can be accomplished, we’ll consider several practical circuits for sending video over several hundred meters of CAT5 cable while also considering the enhancements necessary for handling specific application problems. No formal design standard specifies the design and performance of a differential analog video system, but you can specify such systems using a standard 75- interface at each end. You also can test them using existing standards and test equipment.

Basics of differential signaling
One of the best ways to understand differential circuits is by an analogy with the more familiar single-ended (SE) form (Fig. 1). Although the differential circuit looks complex, it’s simply two SE circuits configured to produce equal and opposite signals with respect to a common pointv. The common point is designated as ground in Figure 1, but it could just as well be a dc bias voltage. In any case it’s a virtual ground, and it joins the two SE circuits that make up a differential one. Why should we use an obviously more complicated circuit?

The answer is subtle, and it hinges on the one or more lines that connect the various parts of a differential circuit to the common point shown as ground in Figure 1. We assume these lines have no electrical properties, but nothing could be further from the truth. At a minimum, they exhibit resistance. At higher frequencies, their distributed inductance and capacitance also become important. These parasitic elements are one of the reasons an actual circuit acts differently from a simulation based on its Spice model. Differential and SE circuits address this problem in different ways. Differential circuits rely on the property of balance, and SE circuits use the property of structure.

Structure is the mechanical construction of a component, as is evident in a coaxial cable. Balance, as the word implies, is a state of equilibrium between two quantities. In Figure 1b, currents IA and IB are equal and opposite and are therefore in equilibrium. They will cancel each other if they flow in the same wire, as will the ones shown connected to the ground symbol. The consequences of this cancellation are profound!

Differential circuits are relatively immune to the effect of parasitic elements, because any currents with a tendency to flow in those elements are cancelled. In turn, that effect reduces the need for structure (shielding) because it is extremely difficultvi for signals to couple into or out of a balanced system. As a result, the signal lines designated for CAT5/6 cables are low-cost, unshielded twisted pairs (UTPs).

In contrast, an SE circuit passes signal current through the lines that connect to the ground symbol. Additionally, it’s very susceptible to the presence of parasitic elements in those lines. To address this problem, SE circuits include a heavy ground structure made of wire braid to connect source and load, ensuring that the ground resistance is much lower than the signal-conductor resistance. Wrapping the braid around the signal conductor produces a form that later evolved into coaxial cable, which combines the requirements of ground and shielding. Compared with UTP, this arrangement requires higher precision and more material. It’s also more costly.

Although SE and differential circuits have similar circuitry, they rely on different methods to achieve performance. The performance of differential circuits depends on balance, but what determines balance?

The well-balanced connection
To answer that question, we need to resolve a differential circuit into its constituent parts: transmitter, source and load impedances, cable, and receiver. Each part contributes to the overall balance of a differential circuit, though it’s not always clear how, why, or to what degree.

For example, a receiver’s common-mode rejection ratio (CMRR) determines its balance. The matching between source and load resistors (their tolerance) determines the balance between them. The cable has multiple balance parameters, but longitudinal balance is usually the most importantvii. For transmitters, the balance specification is common-mode balance (CMB), a parameter that is well documentedviii. CMB is specified for LVDSix but seldom for analog parts.

To explain why, we first define CMB as the ratio of differential amplitude to common-mode amplitude, in decibelsx. In Figure 1b, CMB is the ratio of the two generator voltages (both Eg/2). The CMB specification is necessary to completely describe the circuit, but the cable acts to modify CMB error. Twisting the cable wires together tightly couples themxi, creating a 1:1 transformerxii.

Transformer action forces the currents IA and IB in Figure 1b to be equal and opposite, so the voltage across each load resistor RL/2 depends on the length of the cable used. Unfortunately, this causes current flow in those wires connected to ground. The result is that CMB errors unbalance the circuit, but not as much and not in the way you might think. (This error is seldom an issue, because it is generally less than the tolerance of the series resistors.) Not many parts specify CMB, so you may have to measure it yourself.

Consider why the Figure 1b circuit has two load resistors instead of one. Figure 2A has the single load resistor RL and no return to ground. That configuration eliminates reflections by matching the cable impedance. But it also counteracts the cable’s transformer action described above, thereby increasing the error due to CMB! To avoid that effect while not shorting out the dc bias from the source, we split the load and couple its midpoint to ground with a bypass capacitor (Fig. 2b). Adding a variable resistor (Fig. 2c) lets us compensate for any unbalance, regardless of the cause, while also reducing current flow in the ground wires.

Practical issues for distributing analog video over CAT5 cables
Setting up a CAT5-based distribution system for analog video involves many important issues. The CAT5 twisted-pair cable has many advantages over conventional approaches like coax cable. The first advantage is its cost. CAT5 cables are much cheaper than coax cables thanks to their wide use in computer networking.

In many cases, CAT5 cable may already be installed but not used any longer, perhaps due to an upgrade to CAT6 cables that support higher-speed digital data. In such a case, the older CAT5 cables can be used to distribute component analog video (three channels + sync) or composite video broadcast signal (CVBS) with some additional control signals.

While there are some known issues with CAT5 cabling, once the issues are understood, they’re easy to work around. For starters, the CAT5 cable has four twisted pairs. So if we need more then one channel to carry a signal at the same time, we have to take care of crosstalk between the channels. Differential driver and receiver pairs must have CMRRs of at least 30 dB in the video bandwidth (0-5 MHz). Most of today’s differential driver/receivers can meet this specification, and some of them are especially designed to drive CAT5 cable.

Another issue with CAT5 cable is that high-frequency loss (with cable length) is much higher than that of coax cable. This means that the designer must consider adding some type of cable equalization when the length exceeds approximately 10 feet (3 m). To some extent, this requirement really depends on the quality level of the video signal the system must maintain. Frequency loss isn’t equal among twisted pairs in the same CAT5 cable, resulting in different delay for different channels.

For long CAT5 cables, the designer must include an option at the receiving side to correct delay errors. This kind of problem doesn’t exist if you use four independent coax cables because coax cables with equal length have equal delay.

Basic interconnection
There are several basic connection choices using CAT5 cables (Fig. 3). Selecting the best configuration for your application depends on the conditions as well as the IC chosen for the differential driver/receiver. Every manufacturer usually provides information and examples regarding the best and most efficient configuration for a particular IC. The configuration issues discussed here should provide additional guidance to help determine which option to use to achieve the best termination.

If we have a low-voltage single-supply (preferred) system, it’s generally necessary to add video clamp circuitry at the input to maintain the correct dc level when the video signal changes. In a dual-supply-voltage environment (±5 V or similar) where we have enough voltage swing space, the clamp circuits aren’t necessary.

Again, the choice of dc- or ac-coupled systems depends on the application. For instance, if we don’t expect a big difference in ground potential between source and destination and we have differential drivers/receivers from the same manufacturer, the dc-coupled system is the better choice. It delivers less low-frequency distortion, requires no big decoupling capacitors, has no vertical tilt, and does not need a dc restorer or clamp circuit on the receive side.

On the other hand, we may have a system in which we expect a high ground-potential difference between source and destination (5, 10, or more volts). Or, we may be forced to use driver and receiver chips from different manufacturers. Or, we may have to design only the driver side without knowing the receiver side. In all of these cases, ac-coupled connections will be a better choice. Such connections provide more flexibility.

Multichannel video issues
Since the CAT5 cable has multiple twisted pairs, it’s very convenient when the application needs more then one video channel. For example, when the four video components (R, G, B, and composite sync) or a YPbPr component signal must be transmitted, designers must take care to ensure the dc clamp levels are maintained (Fig. 4).

The dc clamp level is critical because Y, G, B, or R components have different dc clamp levels than the Pb and Pr components. If needed, though, the sync information could be extracted from the Y/G channel. The R and B channels may contain sync as well, but this isn’t mandatory, and the designer cannot assume that sync will be available through the R or B channels. Channels Pb and Pr do not contain sync information. Note that if we need to transfer sync information through the same CAT5 cable, composite sync must be filtered first.

No matter how good the driver/receiver IC pair is, high-frequency and high-energy sync edges will be visible in the video signal as noise because of crosstalk. This provides a strong incentive to limit the sync bandwidth to a maximum of about 1 MHz, depending on how good the chosen driver/receiver pair is.

The last task will be to take care of the delay difference between the channels. The four twisted pairs inside the CAT5 cable will have different signal delays due to cable length, cable position, and temperature. Since the human eye is very sensitive to phase errors between channels (differences in color), the cable driver/receiver circuits must compensate for the delay differences. All channels at the destination must have an error of less than 3 ns.

To handle the compensation, many designs use analog or charge-coupled device (CCD) adjustable delay lines that have adjustment ranges of between 0 and 50 ns. Delay correction could be manual or automatic. Auto delay correction is, of course, more convenient but more expensive.

Multichannel connection video + data
CAT5 cabling also is used in surveillance or security when data (control commands for the cameras) and video must be transmitted through the same cable (Fig. 5). In this combined system, the camera sends analog video to the host system, while the host sends command data to the camera (pan, tilt, zoom, etc).

In this example, assume that the CAT5 cable isn’t too long (10 to 20 feet) and the system doesn’t require cable EQ compensation. The circuit in Figure 5a uses one twisted pair for video and data channels, while the circuit in Figure 5b has the video and data channels in separate twisted pairs but in the same CAT5 cable. In either case, it’s important that the CMRR for the driver/receiver pair stays above 30 dB, and if possible, to set the timing so the control channel goes active only during the video vertical retrace period. This helps minimize possible noise that might become visible in the video channel.

If we need to communicate with a camera that’s farther away (30 to more than 1000 feet), we have to take a different approach since CAT5 cables will require cable EQ compensation. But we can’t apply the EQ compensation to the data channel because the EQ compensation assumes boosting the high frequency, which isn’t acceptable. One possible solution to this uses phase modulation for the control signals in the data channel (Fig. 6). The examples shown use a unidirectional channel. However, that’s not a restriction. With the addition of some hardware, the data channel can operate bidirectionally.

Cable equalization
We have to consider the cable length when we use the CAT5 cable. For long-distance interconnects, we have to select differential driver/receiver ICs that will permit cable EQ/gain correction. Sometimes we need adjustable cable EQ so designers will have more than one option to consider.

But not all differential driver/receivers have adjustable cable EQ. Figure 7a shows typical CAT5 frequency characteristics – the loss of the high frequencies versus cable length. Figures 7b, 7c, 7d, and 7e represent standard NTSC multiburst test signals at 0 feet (0 m), 500 feet (150 m), 1000 feet (300 m), and 1500 feet (450 m). The level of signal degradation that can be expected as cable length increases is clear from the pictures.

Systems can be implemented to compensate for high-frequency signal loss. Consider a system built using the Maxim MAX9546/47 differential CAT5 driver/receiver. For a different vendor’s driver/receiver pair, the particular EQ network would change, but the basic principles are the same.

The driver chip includes video clamping and error detection circuits as well as the ability to detect a short or no connection on the output. This eliminates the need to add extra support components and simplifies the overall system design. The interconnection between source and destination is dc-coupled, and with the MAX9546/47, EQ and gain compensation can be added at the receiving side.

The compensation can be added by replacing the Zt external resistor, which defines receiver gain, with a complex RC network. The compensation network consists of a dc-gain term set by R only and three ac terms set by R1*C1, R2*C2, and R3*C3. The parallel combination of these terms becomes the complex impedance, Zt, as it appears in the gain equation:

VOUT/VIN = K*(Rl/Zt)

where K is the current gain (internally set to 1) and Rl is the external output resistor (Fig. 8). This combination can efficiently compensate high frequency loss for up to 1000 feet (300 m). For some applications, it could be up to 1500 feet (450 m). Zt isn’t efficient above this length, and increases in gain and boosts in the high frequencies actually destroy the signal-to-noise ratio (which drops below 45 dB) and K factor (>3%).

We’ll need to use some kind of modulation for longer distances. In the example shown, the EQ network is a module that covers a certain frequency range (Fig. 9). The EQ network is based on three R*C components, but the user can achieve more accurate frequency compensation by increasing the number of R*C components. The figure shows typical values for the EQ network. The waveforms in Figure 10a through 10d show the NTSC multiburst signal after 0 feet, 500 feet (150 m), 1000 feet (300 m), and 1500 feet (450 m) with appropriate compensation.

EMI considerations
Designers who are familiar with coax cable know its performance and what to expect. The CAT5 cables in the wiring of a building, on the other hand, are unknown in terms of their ability to isolate between signals in adjacent twisted pairs. This uncertainty is reflected in the cable’s specifications for EMI emission and RF immunity. To determine isolation performance, you can run tests that compare traditional single-ended coax with CAT5 differential pairs (Fig. 11).

The typical results were generated by injecting a CVBS signal into a TV via a single-ended RG-59 coax, then a differential CAT5 unshielded twisted pair (UTP), then a standard UTP with single-ended amplifier. Next, the cables were irradiated with an RF signal, and they were examined for interference. When visible interference was seen, the signal level was reduced until the interference was no longer visible, and the interference level was recorded in volts per meter. As the figure shows, UTP cables delivered better performance.

One cannot compare coax and UTP directly. Yet performance comparisons show that differential signaling performs as well as coax, and for some applications, it may even be better while costing less. The cost advantage and performance benefits tip the scale in favor of differential signaling. It remains, however, for equipment manufacturers to incorporate differential signaling in their products.

i Amplifier Applications of Op Amps, J. Graeme, Chapter 6: Differential Output Amplifiers
iiUsing CAT5/5e/6 for Audio and Video Applications, S H Lampen, Belden Cable
iiiVideo and UTP, S.H.Lampen, SMPTE Journal, Feb. 1996
iv Video and UTP, S.H.Lampen, SMPTE Journal, Feb. 1996
vi Chapter 6, “High-Speed Signal propagation,” Graham & Johnson, Prentiss-Hall
vii “Video and UTP,” S. Lampen, SMPTE, Feb. 2, 1996
viii “High-Speed Signal Propagation,” Graham & Johnson, Prentiss-Hall
ix “Balanced LVD SCAI Drivers and Receivers,” Sept. 1997, SCSI Trade Association
x “High-Speed Signal propagation,” Graham & Johnson, Prentiss-Hall
xi “Transmission Line Design Handbook,” B.C. Wadell, Artech House, Chapter 4
xii “Transmission Line Transformers,” J. Sevick, Noble Publishing

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.