Shielding and filtering
While both the coaxial and twisted-pair configurations can be considered “self-shielding” to some degree, an additional conductor is often added to cabling to provide further protection from external noise sources, and/or to reduce radiated emissions from the cable. This is more commonly required in the case of the twisted-pair type, whose performance in practice tends to be further from the theoretical ideal than is commonly the case for a coaxial cable. An added shield will affect the characteristics of the cable to some extent; the degree to which these may change will depend on the precise configuration of the shield and conductors, as well as whether or not the “shield” is truly used simply as a shield, or if there is the possibility of signal or return current being carried via this path.
Other factors affecting the effectiveness of such shielding include the type of material and construction of the shield itself, and the quality of its connection at either end. Ideally, the shield would be perfectly conductive and completely cover the inner conductors of the cable, neither of which is, of course, achievable in practice. Low-resistance shielding with the required flexibility is most often provided through the use of a braided copper layer, but such braids cannot generally provide 100% coverage. Conductive foils, often a Mylar or similar plastic-film layer with a conductive layer on one side, can provide better coverage than a braided shield, but may also represent a significant inductance (depending on the construction of the foil shield) and will typically exhibit a higher series resistance than the heavier braided conductors. The most effective shielding which still retains sufficient flexibility is a combination of the two – a film/foil layer providing 100% coverage, in intimate contact with a braided shield to create a low-resistance/inductance path. As a lower-cost (and somewhat less effective) alternative, a “foil” shield may also be provided with a “drain wire”, which is simply an exposed conductor running the length of the cable assembly, in contact with the conductive portion of the film or foil. (Note also that, per the earlier discussion regarding safety standards, any conductor which has the possibility of carrying fault currents will likely be required to demonstrate that a specified minimum current can be carried for at least a certain minimum time. This requirement can in many applications constrain the type and design of the shield layer and it connections at either end.)
Protection from external noise sources may also come in the form of filtering, which is often designed into the cable assembly and/or connectors. Many connector types intended for video applications are offered in “filtered” version, generally meaning that the signal connections pass through a ferrite material that adds inductance into the signal path. This of course limits the bandwidth of the connection by increasing the impedance at higher frequencies. While this can often be effective in removing unwanted high frequencies – those which do not affect the image quality, but which may cause problems at the receiver or be radiated as unwanted EMI – some care must be taken to ensure that the use of such measures truly does not impact the quality of the displayed image at all desired timings.
Another form of “filtering” in the cable involves the addition of a ferrite “core” (a toroid) over the entire cable bundle, or at least over individual signal/return pairs. In theory, since both the “outbound” and “inbound” currents for any given signal pass through the ferrite, but in opposite directions, this adds no inductance to the signal path itself. However, noise coupling in to the lines from an external source is presumed to be induced on both conductors in the same sense; i.e., it is “common mode” noise. In that case, the added ferrite material places significantly greater impedance in the path of the noise, and so preferentially reduces its magnitude vs. that of the signal. Ferrites are also often used in this manner as a countermeasure against electromagnetic interference, or EMI. The theoretical basis here is the same as for the reduction of common-mode noise; since both the “outbound” and “inbound” currents pass through the ferrite, the signal path sees a low impedance only if both the currents are matched. This effectively places a higher impedance in any other possible return path, and so helps to maintain the match between these two currents. With equal and opposite currents on the cable, radiated emissions are minimized.
Cable losses
It is not possible, of course, to produce any practical physical cable with zero losses. Even if the conductors themselves were lossless, there would be the unavoidable effects of the capacitive coupling between the conductors, etc. Not surprisingly, cable loss generally increases with frequency (owing both to these capacitive effects, plus the “skin effect” increase in the series resistance of the conductors), and are lower for low-capacitance types. A “twinlead” or twisted-pair configuration typically provides lower loss than coaxial; larger coax cables will provide lower loss than smaller types of the same characteristic impedance for the same reasons. The material and construction of the insulation also is a major factor in determining both the cable loss characteristics and the bulk capacitance of the cable. Low-loss designs often use a foamed-plastic dielectric, which of course replaces much of what would have been solid plastic with air. The extreme case of this approach, in coaxial cables, is hardline, in which the space between the center conductor and a solid metal shield is mostly air (or a dry, relatively inert gas such as nitrogen); plastic spacers are used only to support the center conductor within the outer, pipe-like shield. (This is used only in very critical applications, and never to my knowledge in any common display connection; it is mentioned here only in passing.)
Cable loss characteristics are most commonly stated in terms of decibels (dB) of loss per a given distance, most often as dB/100 m or db/100 feet. The resistance and capacitance of the cable are also stated in similar units, such as Ω/km or pF/100 feet. However, the loss numbers will generally be stated for multiple frequencies covering the range of typical uses expected for that cable; the specifications of coaxial cables for PC video interconnect use, for example, will typically provide figures for loss within at least the 1-1000 MHz range.
Table 5-1 Comparison of characteristics for various typical video cable types.
|
Type/description |
Construction |
OD (mm) |
Velocity factora |
Capacity (pF/m) |
Loss MHz |
dB/100 m |
|
75 Ω minimum coaxial cable (Belden 9221) |
30 AWG stranded center, foamed HDPb dielectric, tinned copper braid shield (89% coverage), black PVC jacket |
2.46 |
0.78 |
56.8 |
1 |
2.3 |
|
5 |
5.2 |
|||||
|
10 |
7.2 |
|||||
|
50 |
16.7 |
|||||
|
100 |
23.9 |
|||||
|
200 |
34.4 |
|||||
|
400 |
50.9 |
|||||
|
1000 |
87.3 |
|||||
|
Standard 75 Ω coaxial cable (RG-59/U type; Belden 8241) |
23 AWG solid center, polyethylene dielectric, bare copper braid shield (95% coverage), PVC jacket |
6.15 |
0.66 |
67.3 |
1 |
2.0 |
|
10 |
3.8 |
|||||
|
50 |
7.9 |
|||||
|
100 |
11.2 |
|||||
|
200 |
16.1 |
|||||
|
400 |
23.0 |
|||||
|
1000 |
39.4 |
|||||
|
Precision 75 Ω video cable (RG-59/U type; Belden 1505A) |
20 AWG solid center, gas-injected foam polyethylene dielectric, 100% coverage foil plus 95% coverage tinned copper braid, PVC jacket |
5.97 |
0.83 |
53.1 |
1 |
0.95 |
|
10 |
2.85 |
|||||
|
71.5 |
6.89 |
|||||
|
135 |
8.86 |
|||||
|
270 |
12.5 |
|||||
|
360 |
14.5 |
|||||
|
720 |
21.3 |
|||||
|
1000 |
25.6 |
|||||
|
Standard 75 Ω coaxial cable (RG-6/U type; Belden 8215) |
21 AWG solid center, polyethylene dielectric, 2 bare copper braids (total 97% coverage), polyethylene jacket |
8.43 |
0.66 |
67.2 |
1 |
1.3 |
|
10 |
2.6 |
|||||
|
50 |
6.2 |
|||||
|
100 |
8.9 |
|||||
|
200 |
13.4 |
|||||
|
400 |
19.4 |
|||||
|
1000 |
32.1 |
|||||
|
Standard 75 Ω coaxial cable (RG-11/U type; Belden 9292) |
14 AWG solid center, foamed polyethylene dielectric, 100% coverage foil plus 61% coverage tinned copper braid shield, PVC jacket |
10.29 |
0.84 |
52.8 |
1 |
0.6 |
|
10 |
1.6 |
|||||
|
50 |
3.0 |
|||||
|
100 |
4.3 |
|||||
|
200 |
5.3 |
|||||
|
400 |
7.6 |
|||||
|
1000 |
14.1 |
|||||
|
300 Ω twinlead cable (Belden 8230) |
2 bare copper-covered steel conductors, parallel; brown polyethylene insulation |
1.83x 10.16 |
0.80 |
11.8 |
100 |
3.6 |
|
200 |
5.6 |
|||||
|
300 |
7.2 |
|||||
|
500 |
10.2 |
|||||
|
900 |
14.8 |
a Velocity of signal propagation along cable as a fraction of c.
b HDP, high-density polyethylene.
In the case of analog video interconnects, losses in the cable represent a loss of dynamic range of the signal; the image will generally remain usable, and amplification within the display can compensate to some degree for this loss, at the expense of increasing the noise as well. The insertion of buffer or distribution amplifiers into the path will often be required to maintain a usable signal over long distances (generally, greater than a few tens of meters). A more serious problem may result from the loss of amplitude in the synchronization signals, as these may be degraded to the point at which they are unusable by the display or are confused with noise. In either case, the display will not be able to provide a usable image at all. The sync signals are in many cases particularly vulnerable, as they are often produced by relatively simple output circuits with limited drive capability. Often, the sync outputs are simply standard TTL drivers, not really intended to drive lengthy cable runs.
A sampling of cable specifications for coaxial types intended for video use is given in Table 5-1. Note the differences in capacitance, loss, and velocity factors for the various dielectrics and cable diameters.
Cable termination
As in any transmission-line situation, it is important in analog video applications not only to use cable of the proper impedance, but also to ensure that the cable is properly terminated. Again, the norm for analog video systems has almost always been 75 Ω, and so the display inputs ideally provide a purely resistive 75 Ω load across the full range of video frequencies of interest. If the display inputs were, in fact, to provide this ideal termination, the impedance of the source (the video output) would be irrelevant (as long as the source could drive the line with the proper amplitude signal), as no signal would be reflected by the load. However, as we will see, real-world inputs are rarely even close to the ideal, and so source termination is also an important consideration in preventing reflections on the cable and the resultant “ghosting” in the displayed image.
The easiest method of terminating the cable at the display input is to simply place a resistance of the proper value across the input (Figure 5-7a). The display takes its input across this resistance, through a buffer amplifier stage (which is presumed to have a very high input impedance compared to the value of the terminating resistor). This method, though, will not provide a constant impedance over any but the lowest frequencies. It is can be used successfully for standard baseband television, as such signals do not exceed a few MHz, but may not be satisfactory for higher-frequency use (such as PC monitors). The problem is that the next stage – the amplifier – typically presents a load which is significantly capacitive, and in addition the terminating resistance itself can present a significant parasitic capacitance. As a result, the impedance seen at the display input shows the characteristics of a parallel R-C combination, at least up to a certain frequency. At some point, parasitic inductances – including those in the terminating resistor, as well as the typical coupling capacitor between that termination and the first stage of the video amplifier – will begin to dominate, and the input impedance will again increase, often well beyond the intended nominal value.
Figure 5-7(a) Simple resistive shunt termination. This will terminate the video transmission line in the proper impedance at low frequencies. However, parasitic capacitances, including those across the termination resistor itself (Cpr) as well as the expected capacitive portion of the video amplifier’s input impedance, will significantly reduce the total effective impedance seen by the input signal at high frequencies. This form of termination can often result in significant reflections and “ghosting” of the video signal. (b) Splitting the shunt termination. Dividing the single terminating resistor into two in series (e.g., if the original Rterm was 75 Ω, using two 39-Ω resistors in series) improves the situation by breaking up the parasitic capacitance Cpr. (c) The addition of a series impedance to the terminating network, especially an inductive reactance, will serve to compensate for the increasing capacitive effects at high frequencies and maintain the proper termination of the signal, but at the cost of reduced signal amplitude at the input to the video amplifier itself.
Some slightly more elaborate termination schemes can provide significantly better results. First, a common method of dealing with the parasitic capacitance of the termination resistor is to break this into two resistors of half (or slightly higher) the desired total value (Figure 5-7b). This places the parasitic capacitances of the two in series, reducing the total capacitance of the termination. It may also be desirable to introduce a small series resistance, or even a small inductance, between the input connector and the shunt termination, to help maintain the impedance as seen by the cable over those frequencies at which it would normally decline (Figure 5-7c). This does result in a slight increase in the impedance at low frequencies, and a loss of signal across the termination. However, these effects may be negligible – and in terms of minimizing reflections on the cable, it is always preferable to be somewhat over the desired terminating impedance than under it by the same amount. Obviously, much more elaborate termination networks can be designed, which would present an even better (more constant) load impedance to the line, but these are generally beyond the limits of practicality in mass-market designs.
As mentioned, source (output) termination is not as great a concern as that at the load (display input) end of the cable, and in fact is generally not done as well in terms of providing a well-matched output over the entire frequency range. Depending on the characteristics of the signal driver, a simple resistive termination may be all that is provided (if that). However, many computer-graphics cards employ a somewhat more complex output network, which provides both resistive termination and filtering of the output. As noted earlier, some simple filtering may be provided by choosing a “filtered” connector, which generally refers to one in which the signal pins are surrounded by a ferrite material for extra inductance. This can be effective in reducing high-frequency noise on the line, but may attenuate desired high-frequency components in the video signal too much, especially in the case of a “high resolution,” high-refresh-rate output.
The need for more sophisticated output filtering comes from requirements to minimize radio-frequency interference (RFI) emissions from the display and/or the cable. Very often, the video outputs of computer graphics cards are capable of significantly faster signal edge rates (rise/fall times) than is actually required by the display. In other words, the limitations of the display device are such that faster signal transitions, beyond a certain point, do not result in any visible improvement in the displayed image. In such cases, very fast edge rates become a liability; they do no contribute to the image quality, but the high frequencies they represent are potential sources of RFI. (Very high frequencies in the signal are particularly troublesome, as these are often the frequencies most effectively radiated by the display or cable assembly.) Therefore, edge rate control becomes an important tool in reducing unwanted emissions.
Connectors
Various connector standards are discussed in later topics, but a few words are appropriate at this point regarding the effect of the connector choice on the video signal path. While practically any connector can be made to work in a display interconnect design (and often it seems that practically all types have been used!), the connector choice can have a significant impact on the performance of the interface.
The role of the connector is basically the same as that of the cable itself: to convey the signal with minimum loss and distortion, while protecting it from outside influences. To this basic task is added the requirement that connectors provide the ability to break the connection – otherwise, there would be no need for a connector, or more properly a connector pair, in the first place. This function brings with it the need for mechanical ruggedness – the ability to withstand repeated connection and disconnection while maintaining the electrical performance characteristics – and generally a need for some degree of mechanical security while in the connected state (i.e., you do not want to the connector to be too easy to disconnect, to prevent unwanted failure of the interface).
Electrically, the connector system is governed by the same theory as the cable, in terms of needing to provide a stable and constant impedance, its behavior in terms of shielding performance, and so forth. The importance of the connector in this regard is, however, admittedly far less than that of the cable in all but the most critical applications, due to the much shorter electrical length of the signal path through the connector. Still, any discontinuity in the characteristic impedance of the path or a break in the shielding or return paths can have a very significant impact on the signal quality. Quite often, a given connector type will work quite well if everything is “just right”; one hallmark of a good connector choice is that its design is robust enough to ensure proper performance without undue effort on the part of the manufacturer or user.
A good example of this is the ubiquitous “VGA” connector of the PC industry.At first glance, one would not expect this connector to be a good choice for high-frequency video signal applications, and in fact it is not by any objective measure. However, it has benefited from a truly enormous installed base, and so has been pressed far beyond its original performance requirements by the need for “backward” compatibility. But this connector – basically, a higher-density, 15-pin version of the standard 9-pin D-subminiature type used often in other computer applications (such as the relatively low-speed “serial port” common on PCs) – has several factors working against it. First, none of the contacts are well suited to the connection of coaxial cables; when using the “VGA” connector with video cables, the most common termination method is to attach short lengths of wire to the two conductors of the coax, and then solder those (usually by hand)onto the connector’s contacts. This can result in a significant impedance discontinuity within the connector, unless care is taken to make sure that the wires remain in close physical proximity (Figure 5-8). (Numerous examples can be found where this is not the case.) Next, the contacts themselves – including the contact between the outer shells of the plug and receptacle, commonly used as a connection for the overall cable shield – are not particularly robust, and often fail to make a solid, low-impedance connection for high frequencies. Finally, the connector design itself was not intended to provide a constant impedance, and (depending on the lead configuration used) can result in a fairly large impedance discontinuity even if all else works correctly. Contrasting this connector with one specifically designed for high-frequency video, such as the “13W3” or “DVI” types is an interesting exercise.
Figure 5-8 Termination problems using the “VGA” 15-pin connector. As this connector family was not original intended for use with coaxial cables, there can be significant problems resulting from the method of connecting such cables. In a typical VGA cable assembly, three miniature coaxial cables (supplied in a single cable bundle) are connected to the pins of the connector using short lengths of wire tack-soldered to the shield and center conductor. However, if care if not taken in routing these wires prior to the addition of the overmolded shell (dotted outside line), this may result in an impedance discontinuity due to the large loop area thus formed (a). Simply twisting the wires together (b), to ensure that they remain in close proximity, will result in a significantly improved connection from the standpoint of maintaining the characteristic impedance of the line, although still not as good as a true coaxial connection. (Additional wiring and pins not shown for clarity.)
Performance Concerns for Digital Connections
Attempting to draw a clear distinction between “analog” and “digital” signalling is often based on numerous unstated (and often unrealized) assumptions regarding these terms.It is much more difficult than is commonly assumed to point out meaningful inherent distinctions between “analog” and “digital” signals, and truly many of the concerns discussed above for the “analog” connection also apply to what are commonly said to be “digital” types. The signal must still be conveyed by the physical interconnect with minimum loss and distortion, and these applications generally require a very high bandwidth over which this must be achieved. However, “digital” in the context of a display interface generally means any of several possible binary-encoded systems, using a clock to latch discrete packets of information at the receiver. These may be of either the serial (single bits of information transmitted in sequence over a single physical connection) or parallel (multiple bits received simultaneously for each clock pulse, over multiple physical connections) types, but the basics are the same in either case. (Serial transmission will, however, obviously require higher rates on a per-line basis, if the same total amount of information is to be conveyed.)
The fact that these systems employ binary encoding (only two possible valid states for the signal, the simplest but least efficient form of “digital” transmission) implies that the sensitivity of the signal to noise is reduced. The receiver must only be capable of distinguishing between these states, rather than resolving the much smaller changes required in an analog transmission.However, the relative importance of the timing of signal increases dramatically. Consider the problem of transmitting a 1280 x 1024 image at a 60 Hz refresh rate, at 24 bits per pixel. If this were to be done using a serial transmission of binary data over a single line, the minimum bit rate on that line would be
1280 (pixels/line)xl024 (lines/frame)x24 (bits/pixel)x60 (frames/s) = 1.89 Gbits/s
This means that each bit transmitted is only a little over 500 ps in duration; during that time, the signal must unambiguously reach the desired state for that bit and be clocked into the receiver. This requires both an extremely fast rise and fall time for the signal, and also that this signal and the clock used to latch it into the receiver be properly aligned, to within a tolerance of absolutely no worse than ±250 ps! Given that a 250 ps change the signal position relative to the clock can be caused by a few centimeters’ difference in the effective cable lengths for each, it is very apparent that this is an extremely challenging task. (In fact, no digital interface system with the capability of achieving this rate on a single line has yet been brought to the display market.)
To achieve the levels of performance required for a digital display interface, much the same concerns apply to the cable and connector choices as in the analog video case previously discussed. Impedance control remains important, as does protection both from outside noise sources and from possible radiated emissions by the cable assembly itself. In addition, the cable material and design must be chosen so as to minimize differential delays, or skew, between the various signals carried and their clocks, and also to enable the fast transitions required. Further, to obtain the necessary data rates required in video applications, digital interfaces of these types will commonly require more physical conductors than comparable analog systems. System impedances are generally higher (commonly in the 90-150 Ω range). To meet all of these requirements within the constraints of a reasonably sized cable assembly, shielded-twisted-pair or shielded ribbon cable construction is typical.
A very useful method for quickly evaluating the performance of a digital transmission system is the eye diagram, as shown in Figure 5-9. This is created by observing the digital signal line in question on an oscilloscope, using the appropriate clock signal from the digital interface as the trigger, and transmitting a data signal which ideally alternates between the high and low states. (However, an acceptable “eye” display can often be obtained simply by monitoring an actual data transmission.) Care must obviously be taken in the selection and use of the oscilloscope and its probes, to minimize the influence of these on the measurement. The goal is that the “eye” appear to be as open as possible; vertical separation of the traces corresponds to noise margin, while lateral separation and the position (in time) of the data relative to the clock gives a visual indication of the effects of skew and jitter (short-term variations in the position of the data signal edges relative to each other or to the clock reference).
Figure 5-9 An “eye diagram”. Using an oscilloscope set up to display overlapping traces of a digital transmission, an “eye” is formed between the two traces (a). Noise, jitter, and amplitude instabilities will all have the effect of reducing the area of the “eye,” making it a quick visual check of the quality of a digital transmission. There must be sufficient margin between the high and low states, and a sufficiently long period during which the signal is unambiguously in one state or the other, to permit each transmitted state or “symbol” to be reliably identified in the system.
Details of many of the digital interface and transmission systems in current use.
