Coaxial cables for digital video
A topic discussed frequently by Canford’s technical support team is that of specifications for video cables, most particularly the suitability for reliable transmission of SDI signals. Dr. John Emmett here gives some background to the characteristics needed for high-performance digital video cables. John has enjoyed many stormy relationships with cables during his time managing audio and video standards work, starting with the AES/EBU interface in the early 1980’s.
Although a video signal may be digital in nature, the entire cable route through which that signal travels has analogue characteristics. The analogue distortions produced by the cable include the following, which are listed in rough order of importance to a digital signal:
- Frequency dependent cable attenuation.
- Signal reflections.
- Phase distortion.
- Noise introduction.
While a digital signal will retain some ability to communicate its data despite a certain degree of distortion, there is a point beyond which the data will not be recoverable. When automatic equalisation is used in the typical Serial Digital Video Interface (SDI or HD-SDI), this failure "cliff" will be approached very rapidly. Indeed, the difference in cable length between that producing a trivial error rate and one producing an unacceptable mess, can be as little as 15 metres in several hundred metres of total length.
Serial digital video (SDI) standards
The Society of Motion Picture and Television Engineers (SMPTE) developed several interrelated standards for the electrical specifications of serial digital transmissions (SDI and HD-SDI):
SMPTE 259M: Covers digital video transmissions of composite NTSC at 143 Mb/s (Level A) and PAL 177 Mb/s (Level B). It also covers 525/625 component transmissions of 270 Mb/s (Level C) and 360 Mb/s (Level D). In current European use, level C is the dominant form, sometimes described as "270Mbit SDI", or D1 component SDI, after the old D1 recorder that was one of the first users of this interface.
SMPTE 292M: Covers the newest format for HDTV transmissions at the single data rate of 1.458 Gb/s.
All these standards were specified to work with standard analogue video coaxial cables. The installation costs for any large video facility lie mainly in the cost of the cable system, so it made economic sense to use 75-ohm cable with BNC connectors as a recommendation, and this accelerated the introduction of SDI.
Coping with cables
Frequency dependent cable attenuation
There will be an automatic cable equalizer at the front end of most long-range SDI receivers, and typically it can be designed to equalise all serial digital data signals between 30Mbps and 622Mbps. The terminated input signal will pass through a variable-gain equalising stage whose frequency response closely matches an assumed inverse cable loss characteristic. This gain stage can sometimes provide up to 40dB of gain at 200MHz and this will equalise more than 350 metres of high-quality digital video cable used at 270 Mbit. A detector circuit produces an error signal corresponding to the difference between the desired edge energy and the actual edge energy of the equalised signal. This error signal is integrated by an external differential AGC filter capacitor providing a steady control voltage for the gain stage. As the frequency response of the gain stage is automatically varied by the application of negative feedback, the edge energy of the equalized signal is kept at a constant level. The equalised signal is then DC restored.
This then should give us an automatic method of correcting cable impairments. Well it should, but under two important conditions! Firstly, the transmitter end of the system must send signals with closely controlled amplitude and edge energy, and secondly, the cable attenuation is assumed to vary smoothly with frequency and match the values taken for the design of the equaliser. Whereas analogue signals can be equalised in relatively narrow-band chunks, even channel-by-channel, digital signals require a good cable performance across the entire frequency band at the same time.
Return-loss (RL) is a measurement of signal reflected from the cable or destination, compared to the forward signal level. Therefore, the higher the numeric values in decibels of RL specifications, the better the cable. But wait a moment; with a digital signal we only need to tell a digital "one" level from a digital "zero", don’t we? In that case anything more than 6dB should work? Well again yes, but jitter in the reception of the data edges will cause degradations, until after a few receptions and transmissions, the system will suddenly fail.
In the SMPTE SDI Recommendations, the return-loss for cables is expected to be better than 15dB from 5MHz to 1.5GHz, with the upper frequency limit raised to at least 3GHz for HD cabling. In order to ensure that this minimum level of 15dB RL is met easily after all the rigours of installation, the cables used must meet very much better specifications than this directly from the manufacturer. Bear in mind also that other components in the installed transmission chain can also degrade the RL, particularly bad terminations or improper patch-bay connections.
The third type of cable-induced distortion, sometimes called signal dispersion, is caused mainly by the frequency dependence of the propagation speed of the signal along the cable. This degradation will cause distortion of the digital signal edges and it will form an ultimate limit to the distance of signal transmission along any cable. Signal dispersion is dependent on the quality and performance of the dielectric material in the cable.
Why should this be so important for SDI signals? Well, think of an analogue cable TV signal with a channel spacing of 8MHz in a frequency band around 400MHz. Within that 8MHz range of a fiftieth of an octave, there will be little change in the cable parameters, so equalisation for gain loss is probably all that is required to obtain a good (but never as good as SDI) signal.
The scrambling system applied to the standard SDI signal, restricts the lower frequency that we must carry on any cable to around 5MHz. The upper limit should cover at least the third harmonic of the fundamental clock rate, that is: 270/2 x 3 or just over 400MHz. Now a stable phase response over those six and a half octaves is a tall order for any cable, and to illustrate the effects on data integrity, look at the Figure 1. This illustrates a simple data signal, and an associated bandwidth-restricted version.
If we now allow the phase shift of the higher harmonics to slip backwards in time to around a twentieth of the fundamental wavelength, the response at the receiver will look something like that shown in Figure 2. Increase the Phase shift to around a fifth of the basic wavelength (Figure 3) and you can see that the receiver will already be having serious problems in decoding the signal into true data.
In a coaxial cable, the quality of the screening affects not only the ingress of electrical noise, but also interference leaking out from the signal carried. Compare the solidity of the foil and braid screening on the special digital video SDV-L cable in Figure 4, with the loose wrap of the black-sheathed analogue television feeder cable.
Testing digital video cables
Currently there are is no simple standard to test digital video or HDTV cables. However, measuring and documenting the RL on every link after installation is the most reliable method of spotting any faults, and ensuring that the minimum level of 15 dB is easily maintained. Many apparently good installed cable links may actually be very near the point of failure, you simply cannot tell from a decoded digital video picture. In practice, the RL figure is much more influenced by the cable construction and handling, than by the paper specifications of the performance of one reel of cable in the laboratory. Newer coaxial constructions designed specifically for digital transmissions offer performance advantages over the old analogue designs, and it is worth going over the details why this is so:-
Digital transmissions contain low frequency elements that travel down the centre of the conductor and high frequency elements that travel on the outside of the conductor due to the skin effect. For these reasons, uncoated pure copper centre-conductors are best. The ideal dielectric material (insulation) would currently be high-density polyethylene foam. This needs to be more crush-resistant than standard foam polyethylene and is then less prone to conductor migration. Solid dielectrics generally have poorer electrical characteristics, and air-spaced dielectrics are very vulnerable to installation damage. Both crushing and conductor migration can cause a change in the cable impedance, and this in turn will inevitably cause an increase in RL.
While the nominal velocity of propagation of a solid dielectric is 66%, gas foaming can provide extremely consistent and high velocities of propagation from 82% to 84%. This velocity can be kept constant with frequency in order to minimise phase distortion. The braid shields of precision analogue cables are ideal for frequencies under 10 MHz while foil shields work best above that frequency. Since digital transmissions contain both low and high frequencies, foil plus closely-woven and optically opaque braid designs should therefore be used. Conductor adhesion is another important quality. If adhesion levels are too low, the conductor can move within the dielectric and actually migrate and appear to grow or lengthen in the cable. This can lead to the BNC connector inner pushed right out of the barrel from this effect.
Crushing the cable or other installation damage has possibly the most serious implications for the reliability of any installed cable run. This type of damage is easily inflicted and will quickly cancel out any electrical advantages in the cable specifications on paper, whilst the vulnerability of any particular cable is harder to specify. Figures 5 and 6 show a simple self-explanatory test for crush resistance. Notice the striking difference in cross-section squashing which has resulted from the same crushing force on these two cables that were first shown in Figure 4. It is not hard to guess how much better the SDV-L cable shown in Figure 6 would perform in any practical installation.
SDV-L is Canford's principal cable for SDI system installation, and is available in both normal PVC jacket and low fire hazard ("low smoke zero halogen") versions.
Canford gratefully acknowledges the assistance of Martin Stankovski in the mathematical analysis of the phase-shift effects.