QED- renomowany producent kabli, którego historia sięga 1973 roku, opublikował wyniki swoich wieloletnich badań nad wpływem okablowania na brzmienie pełnego systemu audio w artykule „The Sound of Science”. Poniżej przedstawiam tę publikację, w wersji bardziej przyjaznej niż oryginał, dla ewentualnego poddania go procesowi tłumaczenia. Zdaję sobie sprawę z tego, że obszerność publikacji może być zniechęcająca do uważnej lektury, więc same wnioski wysnute przez firmę QED (tu oznaczone kolorem niebieskim), wraz z moimi komentarzami przeniosłem, do strony następnej- „QED- o wpływie kabli (2)„.
We have always believed that the most sophisticated sound receiving device of all time is the truly incredible human ear, and we both acknowledge and are challenged by how this wonderful instrument can detect the most minute of sonic differences that we currently struggle to measure. Exhaustive listening tests therefore remain an essential element in turning a very good cable in to an exceptional one. But to create a very good cable in the first place, requires an understanding of the science of cable design. It was for this reason that in 1995 we published our very first ‘white paper’ on speaker cable design titled ’The Genesis Report’. This established design principles that still hold true today, some 22 years after its original publication Over the years, we published additional white papers covering HDMI & digital audio and this new report ’The Sound of Science’ consolidates all of this research in to one publication. I believe it makes fascinating reading.
Bob Abraham Co-founder QED July 2017
Around the world, debate continues among both audiophiles and videophiles as to which cable designs offer the best performance. QED, an established market leader in this field, has won many awards for its high-performance audio and A/V cable products. Key to this success has been the application of strong engineering principles, coupled with extensive test and measurement of both our own and competitors’ products. Listening tests are vital: QED engineers are all too aware that measurements alone don’t tell the whole story. But if any cable introduces measurable errors and distortions it obviously cannot convey the signal accurately. QED believes that cables should be as accurate, transparent and neutral as possible, and it is with this credo that our cable development is undertaken, based on measurement and guided by exhaustive listening evaluations. Here we re-visit some of the areas covered in previous QED cable reports and add three new sections, one about biwiring loudspeakers, one about HDMI cables, and one about digital audio optical cables.
Ideally every cable should transfer a signal between two items of equipment with zero loss and distortion. In the real world this is not possible because subtle changes occur in the signal and these may result in readily perceived changes to sound or video quality. The degree of signal degradation is determined directly by the design of the cable. Maximising real world cable performance requires an understanding of the signal transmission process and the engineering tools available, to ensure that the signal arrives in the best possible condition.
Some theory and interesting facts
Cables appear at first sight to be very simple components in both their construction and operation. Once you look at them in detail, though, they become more complex and a lot more interesting.
Conductors: Metal conductors such as copper work very well. These provide free electrons which are not locked into the metal’s atomic structure and can therefore move. There are an incredibly high number of free electrons available: in just one cubiccentimetre of copper there are approximately 8.5×1022 – almost 100,000 million million million. The movement of free electrons is random, normally with the net effect of zero current flow. But when a potential difference – a voltage – exists across the conductor, an electrical current flows in response. Copper is an excellent conductor because its unbound electrons have a large mean free path of about 100 atomic spacings between collisions. The electrical resistivity of a conductor is inversely related to this mean free path, so copper’s resistivity (its resistance to current flow) is low. Silver has lower resistivity still – the lowest of all the metals – but copper is far more abundant, and therefore cheaper, and almost as good, so it is used more extensively. Unfortunately all metal conductors possess finite resistance which opposes the movement of electrons and causes some of the electrical energy to be dissipated as heat. Only superconductors have zero resistance, but a superconductor that works at room temperature has yet to be discovered.
How does the signal travel and how fast?
Scottish mathematician and theoretical physicist James Clerk Maxwell (1831-1879) provided the answer. Signals in fact travel down cables as an electromagnetic wave, and this wave travels at a very high velocity – close to the speed of light. If the metal conductor was perfect virtually all the energy would be transferred from one end of the cable to the other by this means.
The electromagnetic wave consists of electric and magnetic fields that are spatially at right angles to each other and to the energy flux or power flow along the axis of the cable (Figure 1). Electromagnetic wave theory further shows that when conductor pairs are used, the majority of the energy propagates in the space between the conductors (Figure 2)
These fundamental principles apply to all types of signal transmission, whether low frequency audio or high frequency video, digital or analogue. Free electrons are required for the electrical conduction of signals but signal energy travels principally in the fields in and around the conductor.
The importance of dielectrics: Since the space around the conductor carries signal energy, it’s obvious that the material in that space is an important component of cable design. The dielectric space around the conductor has a direct effect on the propagation of the electromagnetic wave along the cable, particularly in terms of its speed, energy storage and energy loss. Dielectric loss increases linearly with frequency. In the non-conducting dielectric, the electrons are bound to the atoms but when an electric field is present, the orbit of the electrons becomes distorted as the negatively charged electrons become attracted to the positive conductor. Displacement of these electrons requires energy, which is drawn from the signal and ultimately dissipated as heat within the dielectric. The better the dielectric, the lower the loss.
We know that electromagnetic fields around the conductor radiate for some distance, rising to a maximum at the conductor-dielectric boundary (Figure 3). This means that objects and materials (other than the dielectric) which are present within the electric and magnetic fields will also have a direct effect on the propagation and power loss of the signal.
The importance of characteristic impedance: This area is often misunderstood. For audio signals (20Hz to 20kHz) travelling over a short distance (some metres), the characteristic impedance of the cable has virtually no impact on the energy transfer between the transmitter (amplifier) and receiver (loudspeaker), and therefore is not a consideration in speaker cable design. This changes once the wavelength of the electromagnetic wave traveling down the cable becomes shorter than the cable length. Designing a high performance cable that carries a radio frequency (RF) video signal consequently requires considerable care to ensure that signal energy is not lost. At low frequencies the characteristic impedance – which is determined by the physical dimensions of the cable’s conductors, their spacing and the relative permittivity of the dielectric – is a complex quantity but at high frequencies it becomes a resistance. Ideally the cable should transfer all the electromagnetic wave energy into the receiver, without any energy being reflected back. If the wave is partly reflected, the reflection will interfere with the incoming signal. To prevent this and ensure reflection-free signal transmission it is necessary for the transmitter output impedance and receiver input impedance to have the same value as the cable’s characteristic impedance. The effect of having mismatched impedances is readily observed using Time Domain Reflectometry (TDR) in which a very short pulse is sent down the cable. Any mismatch in terminating impedance causes some of the electromagnetic wave energy to be reflected back after a short delay. If the cable is unterminated then virtually all the energy is reflected back. Using TDR we can establish how well a cable is manufactured and how well-engineered (matched in impedance) its connectors are (see Figure 4).
When designing cables for carrying high frequency signals the connector impedance is critical. When we developed QED’s TTV range of aerial cables, TDR analysis showed that many so-called ‘high end’ connectors did not have the correct impedance, causing a significant fraction of the signal energy to be reflected back to the source. When signal levels are particularly low, as is the case with antenna feeds, the outcome is poor picture quality and fewer stations in the channel listing.
Conductor Chemistry: A conductor’s chemical composition is very important – the presence of trace elements such as silicon, magnesium and phosphorus decreases electrical conductivity. The most widely used copper for cables is electrolytic tough pitch (ETP) copper, which consists of extremely high purity metal plus oxygen in the range of 100-650ppm (parts per million) concentration. Oxygen is used as an alloying element and also as a scavenger as it reacts with most of the impurities in the copper. Adding around 0.02% oxygen to ETP copper increases its conductivity. Oxygen-free copper (OFC) is produced primarily for its ability to be heat treated without embrittlement, for ease of use when welding and brazing. The acronym for oxygen-free high thermal conductivity copper, OFHC, is a registered trademark of Phelps Dodge Specialty Copper Products. OFHC is a highly refined grade of copper that contains almost no oxygen or other impurities and this is the grade employed exclusively in all QED cable products. Certified oxygen-free high thermal conductivity copper contains a minimum of 99.99% copper making it the purest metal in common use. If the temperature and time constants are carefully controlled in an oxygenfree environment while the copper is manufactured the grain structures in the material can be reduced so that the purity of the copper increases to as much as 99.999% (termed ‘five nines’). The vast majority of QED cables, both in our speaker cable and interconnect ranges, utilise five nines copper. The IACS (International Annealed Copper Standard) gives a percentage ranking for conductivity of OFHC at around 102.4% whereas ETP coppers fall typically within the range 100-101.5% IACS. Although linear crystal coppers approaching a purity of six nines have been made in small quantities by various continuous casting or Czochralski processes, it remains prohibitively expensive for use in commercial products and, in the absence of any official IACS designation, at ambient temperatures has no conductivity advantage over its five nines counterpart. OFHC is particularly well suited to cryogenic treatment, in which the copper’s temperature is slowly reduced to below −190°C and then gradually returned to ambient, and QED takes advantage of this property by cryogenically pre-treating its high-end cable range. It is thought that the improvement in physical properties that occurs when copper is cryogenically treated results from the elimination of dislocations in the material’s microstructure which might otherwise impede the free movement of electrons throughout the crystal lattice.
In the digital world – jitter and crosstalk: Jitter is a term used to describe time variation in the arrival of a digital signal at the receiver. All metal cable links have a finite bandwidth which attenuates the high frequency components of the digital signal. A transition delay (that is, delay in the receiver detecting a 1 or 0) occurs which varies depending on the pattern of the digital signal. Digital equipment normally has engineered-in tolerance to jitter but it is not always effective, depending on the frequency of the incoming jitter. In such cases cable-induced jitter is transferred to the recovered clock signal, as is the case with an S/PDIF digital audio signal. For long cable lengths the digital signal ‘eye pattern’ closes and reduces in size, ultimately resulting in a failure of the link (Figure 5).
When more than one signal channel is transmitted down a single cable, as is the case with digital video, ‘crosstalk’ – the unwanted transfer of energy from one channel to another via inductive or capacitive coupling – must also be considered. Individual screening of signal pairs can improve crosstalk performance, but this can cause a reduction in the available bandwidth. The effect of this is a reduction in the pulse rate time or slew rate, which degrades signal transfer. Various design techniques can be used to improve performance, such as increasing the conductor pair twist rates, ensuring that conductor pairs have different twist rates to minimise coupling, and using better-performing dielectric materials.
Examining the geometry of each cable sample revealed that the majority of the multi-stranded cables tested were inherently inductive. The inductance of a cable depends on the area of the conductors, their relative spacing and the permeability of the surrounding media. (High permeability materials, such as iron and ferrite, are used to increase inductance, in wound inductors for instance.) In cables, the wider the conductor spacing, the greater the inductance. Many of the multi-strand loudspeaker cables on sale feature conductors that are widely spaced, some by more than three times the conductor diameter, resulting in higher values of inductance. Averaging the inductive effect across our sample cables gave an effective phase shift of 0.42 degrees per metre. So for a 10 metre cable length there would be 4.2 degrees of phase shift. At present, the audibility of nonlinear frequency-dependent phase shift is uncertain, although amplifiers that exhibit poor phase response are often criticised as being ‘grainy’.
Response peaking due to inductance and capacitance: Another effect of inductance is high frequency amplitude loss due to increased cable impedance (inductive reactance increases with frequency). High cable inductance can also be responsible for a rise in voltage at the loudspeaker terminals caused by interaction between inductive and capacitive reactances resulting in a damped resonance. This can be a problem with electrostatic loudspeakers, which present a higher capacitance load than conventional moving coil loudspeakers. An example of resonant peaking is shown in Figure 12, the two response traces having been measured at either end of the cable (amplifier output and loudspeaker input). Here the increased impedance of the cable at frequencies above the response peak results in considerable loss of signal level when combined with the amplifier’s own roll-off.
Capacitance is also governed by the spacing and diameter of the conductors. The greater the gap between any two conductors in a given dielectric, the lower the capacitance (the reverse being true of inductance). Designing a cable with low inductance and low capacitance is particularly difficult when poor quality dielectrics are used. The majority of lower-priced cables and many we tested, use PVC dielectrics, which cause inherently greater capacitance and dielectric losses. Whatever is done with the conductor spacing and diameter, such cables are at a distinct disadvantage, having greater capacitance or inductance or both.
Effects of Capacitance: In theory, the capacitance of loudspeaker cable should have little effect on system performance because the cable is driven by a very low source impedance, typically fractions of an ohm for most power amplifiers. Although the capacitance forms a low-pass filter when connected to this impedance, its effect on frequency response within the audible range is typically minuscule. More insidiously, unduly high cable capacitance in a speaker cable may indicate poor dielectric quality and high dielectric losses.
Some esoteric cables employ a number of separately insulated paralleled wires to form the two conductors. With certain geometries and lower-grade materials, this can cause capacitance to rise to a high level. One cable we tested had a capacitance of 1375 pF compared to the average sample capacitance of 500 pF for a 10 metre length. Another factor to be considered with speaker cables is amplifier stability. In some cases a little extra capacitance on the output of an amplifier can make it oscillate, overheat and even self-destruct. More usually the amplifier may oscillate momentarily at a frequency above the audible range during operation and show no obvious symptoms. Well-designed feedback amplifiers normally have a good gain/phase margin, which ensures that small extra phase shifts due to increased load capacitance do not cause such problems. But some commercial designs do not have sufficient gain/phase margin for unconditional stability and it is these
which can cause problems with long lengths of higher-capacitance cables. Additionally, inductance is likely to be low for high-capacitance cables, leading to a further reduction in the stability margin. Even if the amplifier is not outwardly unstable, sound quality can still suffer, with harsh and forward-sounding results as the amplifier operates on the edge of instability. Figure 14 shows an example of instability due to high-capacitance cable, visible as ‘ringing’ on a square wave signal.
Multi-strand vs single-core distortion: It has been claimed that multi-strand cables introduce ‘diode’ effects due to current jumping between strands within the cable and thereby crossing very many metal/oxide/metal junctions. This is sometimes said to be caused or worsened by skin effect as a result of it pushing the current towards the conductor surface at high frequencies. Making the assumption that the current does ‘jump’ as suggested, we put signal in via one strand of a multi-strand cable and measured the signal from a different strand. Even using the Audio Precision AP1 test set to its maximum capability, no increase in distortion compared to using all the strands could be found, as shown in Figure 22. This graph overlays curves for the two cases – they are so close that they could as easily be repeat tests of the same measurement. It seems that interstrand diodes do not exist, or if they do they are shorted out by the many good conductors pressed together over the cable’s length.
- DC resistance– Low cable resistance is of paramount importance if high sonic performance is to be attained, but it should not be achieved at the expense of other crucial parameters. High cable resistance results in several undesirable consequences: frequency response aberrations, impaired transient response, increased distortion at the loudspeaker terminals and reduced inter-channel separation. All cables exhibiting high resistance measure badly in these areas and subjectively their performance is highly dependent on the partnering loudspeakers. The forward midrange presented by these cables correlates closely with their effects on frequency response. High cable resistance also reduces dynamic impact with heavily-scored music.
- Inductance– Cable inductance is a prime cause of high-frequency attenuation and phase shift in loudspeaker cables. Inductance causes cable impedance to rise with frequency, reducing output in the very upper frequency range, sometimes preceded by response peaking. In addition, inductance increases distortion at the loudspeaker terminals and degrades the loudspeaker’s overall transient behaviour. Low inductance is required to achieve a flat frequency and phase characteristic, low distortion and good transient response.
- Skin effect– These are of minor significance in loudspeaker cables of moderate cross-sectional area but become important in cables with larger conductors where, together with greater inductance, they result in greater high-frequency signal loss.
- Insulation quality– Dielectric dissipation factor has proved to be a very strong indicator of sound quality in our listening tests. Most of the better-sounding cables we have tried use superior dielectric materials whereas PVC-insulated cables give the worst sound quality. Cables which measure badly for dielectric loss appear less able to reveal subtle detail, losing some of the atmosphere revealed by cables with superior dielectrics.
- Consistency of performance– Speaker cables interact both with the amplifier and the loudspeakers. Consequently, some cables give varied results in different systems. Those which have proved to perform most consistently are those with minimal inductance, capacitance and resistance. Unless an amplifier relies on inductance to maintain stability, keeping the speaker cables as short as practically possible optimises performance. High cable capacitance is best avoided because it can result in amplifier instability, which can degrade sound quality and negatively impact amplifier reliability.
- Directionality– Despite many manufacturers marking cables directionally, we have found no evidence under controlled conditions that speaker cables are directional. But we have found that merely constructing a cable differently can affect its inductance and capacitance, which may have an impact on sound quality.
- Solid-core vs stranded cables– Solid-core conductors have been introduced on the basis that, if made thin enough, a solid conductor will show less variation in loss at high and low frequencies than a thicker, stranded conductor. Our research suggests that it is more likely to be the insulation and geometry of many solid-core cables which are responsible for their generally higher performance than stranded conductors. In any case, paralleling up conductors, whether solid or stranded, reduces inductance, which has a far greater influence than skin effect. The stranded cables we’ve tested tend to have higher inductance and dielectric loss than many solid-core counterparts which generally use separately-insulated wires (resulting in lower inductance) and higher-quality dielectrics (resulting in lower leakage losses). We have found no evidence to support the contention that stranded cables suffer from distortion due to diode effects between strands.
- Metallurgy– Electrical conductivity is slightly superior for cables utilising high-purity copper. Greater improvements to conductivity can be achieved with silver-plated copper or pure silver conductors. Generally we have found that a cable’s geometry and dielectric material are more significant than conductor metallurgy in determining its sonic performance.
Conclusion: All our research confirms that the most accurate and consistent-sounding loudspeaker cable will have minimal DC resistance, inductance and capacitance combined with low loss dielectrics. Cables having a small cross-sectional area in an attempt to avoid the skin effect have higher DC resistance, with sonically obvious harmful consequences. QED’s engineers have shown that the ‘rule’ relating high inductance to lowcapacitance, and vice-versa, is oversimplified. Capacitance and dielectric losses can be reduced by using high-quality insulation material (low-density polyethylene) and by minimising insulation wall thickness and designing narrow webs (consistent with mechanical integrity), thereby increasing the ratio of air to solid dielectric. By optimally orientating multiple parallel stranded conductors around a central non-conductive core , e.g Aircore™ and X-Tube™ Technology, QED has been able to reduce both inductance and capacitance below that of a single conductor pair of the same DC resistance. Use of stranded conductors of sufficient cross-section keeps DC resistance low. The result is a range of low-loss, transparent-sounding loudspeaker cables of superior performance. The correlation between insulation and sound quality has also influenced the design of QED’s interconnect cables, where the use of foamed LDPE insulation to increase the air/solid ratio in the dielectric helps maximise sound quality.
Biwiring: an exploration of the benefits
Biwireable loudspeakers have been available since at least the late 1980s. (You can easily tell if a speaker is biwireable because it will have four connection terminals on the back rather than two.) The idea is that instead of using just one run of cable to each speaker, a separate cable is used for each of the two pairs of terminals, once you have removed the shorting wires or plates that normally connect them. This means that you will use twice as much speaker cable than if you used the normal single-wire method, which has prompted cynics to use the phrase ‘buy wire’ to mock the idea. If you have not biwired speakers before, you might be tempted to try because of what you have read in the hi-fi press or online – but you may also wonder if the extra outlay is well spent. Proponents of biwiring point to the obvious sonic benefits they hear and cite the fact that speaker manufacturers fit the extra terminals as proof that there must be something in it. Detractors argue that manufacturers are merely maximising the marketability of their products and point out that there is little published evidence to prove that biwiring makes any audible difference. Biwire enthusiasts, meanwhile, theorise that separating the low- and high-frequency signal currents can eliminate distortions caused by interactions between them. Biwiring, then, is a controversial subject. Is it a worthwhile sonic upgrade or just a clever marketing ploy by the cable companies to encourage you to buy twice as much cable? We decided to do some research to see if an authoritative answer can be given to this question.
How are normal loudspeakers connected to an amplifier?– In Figure 23 a single speaker cable is shown connecting the output terminals of one channel of an amplifier to a loudspeaker. The speaker is a non-biwireable two-way type, having a woofer to reproduce low frequencies (LF) and a tweeter to reproduce high frequencies (HF). A passive electrical circuit (represented by the two boxes labelled ‘HF network’ and ‘LF network’) is used to filter the signal from the amplifier so that only the low frequencies are passed to the woofer and only the high frequencies to the tweeter.
The filter networks cannot create sharp cut-off points at the chosen ‘crossover’ frequency but instead apply more gradual attenuation with considerable overlap. In Figure 24 you can see the point where the two response curves cross (green for the woofer, red for the tweeter) is at about 2kHz, at which frequency the two drivers contribute equally to the acoustic output. If the LF and HF networks are correctly designed then the combined output of the two drivers (black curve) remains essentially flat through the crossover region.
How are biwired speakers connected to an amplifier?– Figure 25 shows how a typical biwired speaker is connected to an amplifier. It has four terminals instead of two, one pair for the HF network and one pair for the LF network. Two speaker cables are used to connect the HF terminals and the LF terminals separately to the amplifier output terminals.
How does signal current flow in single-wired and biwired speakers?– With a single-wire connection, signal current for both the drive units flows along the same cable. With a biwire connection, the low frequency and high frequency signal currents are separated. At high frequencies the impedance of the LF network is high and so the HF current is low; at low frequencies the impedance of the HF network is high and so the LF current is low. It’s not a matter of the low and high frequencies ‘knowing which way to go’ – it is just a natural consequence of the biwire connection.
What are the benefits of separating low- and highfrequency signal currents?– Wherever there is nonlinearity in a system – and loudspeaker drive units are nonlinear – there will be intermodulation distortion introduced in the form of sum and difference frequencies. Unlike some types of harmonic distortion, intermodulation distortion is by nature audibly objectionable. If separating low and high frequency signal current by biwiring reduces the loudspeaker’s intermodulation distortion then we can expect it to have an audibly beneficial effect.
If intermodulation distortion is reduced by biwiring, how can we measure it?– Now we are close to proving that biwiring has some genuine sonic benefit. We have accepted that the two cables carry different signal currents and that by keeping LF and HF currents separate we may have prevented the generation of intermodulation distortion. If it was possible to measure intermodulation distortion in a single-wired speaker connection and to demonstrate a reduction of that distortion in the same speaker when biwired, we could make a really good case for the audible benefit of biwiring being real. Some work on this subject has been published by Jon Risch1. He used a split-band test signal comprising 10 tones divided into two frequency bands, one low and one high, starting at 100Hz and 5kHz respectively. The frequencies of the tones were carefully chosen so that any intermodulation components which are added are minimally masked by harmonic distortion. Using the same methodology, we generated a similar test signal and conducted our own investigation to see if we could measure any differences between single-wired and biwired speaker connections. First we created the multitone test signal and burned it to a CD. Figure 26 shows a spectral analysis of the CD player’s output when playing the test CD. The high frequency tones can be seen as five distinct peaks in the graph from 5kHz to about 9kHz at around −10dB; the low frequency tones appear as five clear peaks from 100Hz to a little less than 200Hz, also at around −10dB. Note the low level of noise and distortion between the two bands.
Then we used a current probe to measure the spectra of the signal currents in a single-wired connection and in each cable of a biwired connection (woofer cable and tweeter cable). Results for the single-wire connection (blue trace) are shown in Figure 27, overlaid on the black trace of the test signal from Figure 26. Intermodulation products have significantly raised the distortion level, at some frequencies to only 40B below the level of the test tones.
Compare this with the measurement taken from the tweeter cable in the biwired connection, tested under the same conditions, shown as the red trace in Figure 28. First, note that the LF tones have been attenuated by over 30dB which proves that low- and high-frequency currents have indeed been separated in the biwired connection. Second, see how intermodulation distortion has been significantly reduced below the upper set of tones but still within the passband of the tweeter.
Now let’s look at the spectrum for the woofer cable (green trace in Figure 29). The high frequency components have been attenuated this time, again proving that the LF and HF signal currents are separated in a biwire connection. There has been a less dramatic reduction in the level of the intermodulation products but there is still some improvement.
If your speakers have four binding posts then you can biwire them but you will need twice as much cable. Low-frequency and high-frequency signal currents will then travel in separate cables. Our measurements show that this reduces intermodulation distortion caused by nonlinearity in the speaker system. In light of this evidence it’s sensible to conclude that where the opportunity exists and funds allow, biwiring should be explored as a means of improving the performance of any suitable high fidelity loudspeaker system.
Reference: Risch, Jon M, “A New Class of In-band Multitone Test Signals”, Paper 4803, Audio Engineering Society 105th Convention, September 1998
Notes: The test signal CD was played using an all-in-one CD player/amplifier set at half volume, into a pair of floorstanding speakers. Current probe measurements were recorded using a Tascam US144 USB audio interface and analysed using TrueRTA. The single-wire measurements were taken with two runs of speaker cable connected in parallel (speaker terminal shorting connectors in place); the biwired measurements used the same arrangement but with the shorting connectors removed.
Some writers in the general press and on the internet suggest that it’s not worth upgrading to high-quality digital audio and HDMI cables. As a leading manufacturer of cables and accessories for more than 40 years, we have a wealth of experience in this area and, at our cable assessment and design facility, we own some of the world’s most advanced HDMI digital measurement equipment. As well as using this to develop QED cables, we also use it to evaluate other HDMI cables on the market. So here’s our (well-informed!) contribution to this debate.
We agree: a lot of HDMI cables are the same!– We are shocked by some of the cables we measure, particularly when premium cables prove to offer little or no measurable improvement over lesser models from the same manufacturer. On these occasions we find ourselves in agreement with our more sceptical journalist friends. Just because a cable looks flashier or is priced higher than another cable, it is not necessarily better. It may just be standard cordage disguised with some ‘bling’.
Misconceptions: There is a common misconception that digital data transfer in HDMI is perfect, that noise immunity and error correction work together to ensure that data received by the sink equipment (the receiver) and subsequently shown as a picture on the TV screen is always the best it can be. In fact there is no such error correction in an HDMI link. Rather, errors that creep in due to logic level transitions in the physical layer (the cable) are minimised by encoding the data stream so as to minimise those transitions. At the sink (TV) a simple test is used to verify the stream and it is then converted into a viewable picture, complete with any cable-induced timing errors. These timing errors, known as jitter, can affect the perceived quality of the picture so it is important to keep them to a minimum. This requires that the cable introduces as little ‘skew and slew’ (clock skew and slew rate limitation) into the signal as possible, which can only be achieved by the use of carefully controlled cable geometry and top-quality materials.
Just ones and zeroes? Yes, but there is more to it than that: “Digital signals are either on or off and immune to noise – so as long as there is a signal the other end of an HDMI cable that can still be read by the TV then the received signal is perfect, isn’t it? How can the type of cable make any difference to the received signal– expensive HDMI cables are a waste of money, right?” You’ll often see this argument deployed but it relies on the erroneous assumption that any errors or noise introduced by the transmission of digital signals along a cable are masked by the natural noise immunity of the signal’s discrete logic levels or repaired by error correction algorithms in the receiving equipment. The true situation turns out to be much more of a grey area.
What is inside an HDMI cable and what signals does it have to carry?– HDMI stands for High Definition Multimedia Interface. It is a way of conveying encrypted digital video and audio signals from a source (eg a DVD or BD player) to a sink (a TV or AV amplifier) and is capable of bandwidths that far exceed that of standard definition TV signals. The High Speed standard, which is the maximum bandwidth described in the HDMI 1.4 specification, uses 10.2 Gigabits per second transmission for a deep colour picture resolution of 48-bit 1080p60 (1080-line progressive scan at 60 frames per second). This requires the HDMI cable to convey 10.2 billion discrete logic levels every second in order to meet the High Speed HDMI standard.
An HDMI cable incorporates four twisted pair cables (two insulated wires twisted into a helix) to carry four data channels used for Transmission Minimised Differential Signalling . Figure 30 shows the internal construction of a typical HDMI cable with the TMDS lines (label 4 and its equivalents) and several other conductors that are used for communication between the source and sink, transmission of remote control signals and video encryption in the form of High Definition Content Protection, all of which work at much lower data rates then the TMDS lines. There are also conductors for power and ground.
TMDS: There is a limit to how quickly the signal can be switched from logic level 1 at 0.6V to logic level 0 at −0.6V due to cable capacitance. The twisted pair conductors act like the two plates of a capacitor and when the signal changes from a low to a high voltage or vice-versa this capacitance must be charged or discharged before the required voltage level can be attained. This takes a finite time and has the effect of rounding off the square edges of a perfect digital signal. If the data rate is too fast, the cable can still be charging up from a zero to a one when it’s time to discharge form a one to a zero again, and the receiving equipment may not see any logic
transition at all. In order to prevent this, the serial data stream that makes up the HDMI signal is split into three parallel signals that are sent down three separate twisted pairs within the HDMI cable. This reduces the maximum data rate each conductor must be able to carry to a third, so the bandwidth requirement is reduced to 3.4 Gb/s for a High Speed cable. These twisted pairs are known as TMDS lines. TMDS stands for Transition Minimised Differential Signalling, the mechanism of which is explained in more detail in the next section. Because it is the logic level transitions that cause errors in digital signalling, these are minimised using a simple algorithm. The ‘differential’ part of the interface description comes from the fact that the transmission line is made up of a twisted pair and separate ground wire. This hot, cold and screen geometry is used in analogue signalling too, where it is known as a balanced line and used to cancel the effect of external electrical interference Figure 31 shows the signal that appears at the output of the HDMI source equipment on each of the three TMDS lines. This signal is assumed to be perfect and is, as yet, unaffected by the electrical characteristics of the HDMI cable. Known as an eye diagram, this measurement is used to test the HDMI cable for compatibility with the HDMI standard. It is a composite density map of a million samples taken over a few seconds of a random bit pattern running at the 3.4 Gb/s data rate defined by HDMI.org. There are three such data lines which add up to 10.2 Gb/s altogether – the maximum data rate for HDMI and that defined as “High Speed” by the HDMI standard. The width of the graticule (or window) represents a time period of just 571ps (571 millionths of a millionth of a second), enough to encompass two logic level transitions or three bits of data. All the possible transition combinations are present: signal sections which remain at logic one across the whole graticule; signal sections that stay at logic zero across the whole graticule; and signal sections where one or more high/low or low/high transitions occur.
The colours in the diagram represent the density of the sample points, with blue representing only a few samples and red the most. It can be seen that the signal is very rarely caught in the transition between zero and one, with most of the samples being found in the discrete levels of either logic one or logic zero. The purple area top and bottom and in the middle are the “don’t know” areas where the receiving equipment would not be able to interpret the incoming signal as a one or a zero. If any of the samples fall into this area then there is an error in the data. In addition to the three data lines within an HDMI cable there is a fourth similar shielded twisted pair which carries the master clock signal. In HDMI the data is split up into chunks of 10 bits and because of this the master clock only has to run at a tenth of the data rate, i.e. a maximum of 340MHz in a High Speed HDMI cable. This helps to keep the clock accurate because the lower its frequency, the less its waveform is affected by cable parameters. In order for the sink device to read the digital data from all three of the data channels, it must be told when to look at the incoming signal. This is the purpose of the master clock but because it runs at one-tenth of the data rate, the TV must multiply the clock frequency by 10 and then realign it with the data window in each of the three TMDS data lines. This reproduced bit rate clock does not have a guaranteed phase relationship with any of the three data lines. In addition the HDMI specification allows a certain amount of skew between any two data lines. Because of this the clock phase must be adjusted individually for each data line to sample the incoming serial bits correctly. This involves aligning the clock rising edge to the middle sampling window of each data line. This is done by delaying the data by varying amounts based on the incoming data stream and how much it is skewed or jittered from the ideal ×10 data clock that has been recovered. If the jitter is too large then errors can creep through.
Display Data Channel: The Display Data Channel or DDC (Figure 30) uses two single conductors within the cable. It forms an I2 C communications bus for the Enhanced Extended Display Identification Data (EDID) protocol which is a way that the source and the sink can communicate with each other and negotiate the data rate for the video signal. For instance, a DVD player set to output 1080p will announce this to the TV which will tell the DVD player to go ahead and set itself to receive 1080p.
Consumer Electronics Control: The Consumer Electronics Control or CEC (Figure 30) uses a single conductor and common ground within the cable. CEC allows the user to control up to 10 different devices over the HDMI link using only one remote control. For instance, it can be used to control the volume of an AV amplifier using the TV remote handset codes.
How is the video signal transmitted by HDMI and how are errors dealt with?– The TMDS signal transmission method used by HDMI is the same as that found in the DVI standard, with which it is backwards compatible. The idea is that errors are minimised, rather than corrected, by encoding the signal in such a way that transitions between logic high and low are kept to a minimum. This is achieved by performing a simple XNOR or XOR operation on each successive bit of the pixel data and its predecessor. Two control bits are then added to each pixel data byte to form a 10-bit video character. One of these extra bits is added to indicate which of the two logic functions was used. Additionally, if the previous character contained a lot of ones and the next character will also, the system can invert that character to maintain a mean zero DC offset in the signal, so a second control bit is added to the signal to indicate whether the character has been inverted or not. In this way the 10-bit pixel data character will contain a maximum of five transitions. Each pixel data character is bookended by a control data or blanking character of 10 bits which is used to indicate the boundaries of the pixel data character. There are three such channels in the HDMI signal. All this is clocked out on a fourth differential transmission line at between 74.25 – 340MHz depending on the video standard, with one clock pulse per data character so that the clock rate is one tenth of the bit rate. When the signal is decoded the other end there will be no picture until the receiver has synchronised with all three video data streams and this is achieved by detecting the blanking periods and synchronising them with the one-tenth-bit-rate clock signal. The blanking signals are only distinguishable from the video data because the number of transitions in the 10 bit character is higher (therefore once the video signal has been corrupted enough that the pixel data is indistinguishable from the blanking data synchronisation is lost and the video will cease or not even start). If the incoming signals can be synchronised, the data is then decoded according to the settings of the two control bits to restore the correct number of data transitions.
discrete logic levels that it needs to detect. In HDMI, a voltage is decoded as binary 1 as long as it lies between 70mV and 780mV. Unfortunately, as noted earlier, the master clock must be recovered from the data and is only corrected every tenth bit. A data signal displaying 120ps of jitter will cause a corresponding variance in the recovered clock. This means that it is possible that the TV will get its clock signal up to 60ps early or late which is enough to make it see data that is at a “don’t know” level and thus cause a bit error. The HDMI spec says this is OK as long as the bit error rate is below 10−9, or up to four bit errors per second in a Full HD (1080p60) picture. Noticeable pixel errors will start to creep in before sync and eventually the entire picture, is lost. Timing errors are caused by this mechanism too and the signal can also be affected by crosstalk from adjacent signal carriers. Within the HDMI cable there are three differential signalling lines plus a master clock signal, so there is plenty of scope for this kind of crosstalk to take place. Finally, because the transmission lines utilise differential signalling there can be slight differences between the two conductors in each twisted pair that introduce timing errors known as intra-pair skew. This can be caused because the two conductors are slightly different lengths or because the dielectric (insulation) is inconsistent or the twist rate varies. All these geometrical and electrical parameters must be carefully controlled within the cable if jitter is to be minimised. Some manufacturers have recognized this problem and incorporate upsampling and local master clocks in their equipment to try to reduce or eliminate jitter, but there is inconsistency between manufacturers so it is important to try to preserve the signal integrity along the cable as a first line of defence, before any post-transmission recovery is added or required. Jitter can also affect picture quality perception way before any bad pixels start to appear, a phenomenon investigated in ‘The Effects of Jitter on the Perceptual Quality of Video’, Claypool & Tanner, Worcester Polytechnic Institute USA 1999.
In this paper the authors report “we found that jitter degrades perceptual (video) quality… and that perceptual quality degrades sharply even with low levels of jitter as compared to perceptual quality for perfect video.” What does this degraded perceptual quality look like? Here is an extract from some research carried out at QED HDMI Labs in 2012: “In an effort to try to characterise this problem we artificially jittered a standard colour bar video signal until we could see a change in the video output from a TV. It was significant that errors in the picture were visible even though the HDMI link was still up and the cable was
able to maintain a steady picture.” The two eye diagrams in Figure 32 show the amount of timing error we were able to add to the cable before problems became apparent. The trace on the left
Digital audio optical cables
Ciąg dalszy: QED- o wpływie kabli (2)