5. HDMI cable's design
This section of the article gets a little technical, but bear with it, as it does explain thoroughly how HDMI cables work, and the factors which can affect the cable's performance.
The primary work of an HDMI cable is done by four shielded twisted pairs which carry the colour, sync, and clock signals. Control of the cable impedance is critical to keeping the rounding of the bit edges under control; the more the impedance wanders off of spec, the more the signal will round, and the closer the cable comes to failure. Where a coaxial cable's impedance can be controlled within two percent of spec, it's a challenge to keep a twisted pair any tighter than about plus or minus 15%.
The HDMI signal will fail if attenuation is too high, or if the bit transitions become excessively rounded so that the receiving unit can't reconstitute them accurately. There's no really reliable benchmark for just how much attenuation is acceptable, or how round the shoulders can be, before the "sparklies" will start. (Yes, there are specs for these things in the official HDMI spec document, but real-world devices vary so much that meeting the spec is no guarantee of success, while failing it is no guarantee of actual failure.) But while wire gauge has much to do with the former, it's really the latter that's important; and wire gauge has nothing to do, at least directly, with impedance control.
Transmission line impedance, in any cable, is dependent on the cable's materials and physical dimensions. For purposes of an HDMI cable, these are:
1. the shape and size of the paired wires;
2. the thickness, and dielectric properties, of the insulation on the paired wires;
3. the dimensions of the shield over the pair.
These seem, in principle, like simple things to control – that is, until one spends a bit of time in a wire and cable factory and finds out just how many little problems there are. Wire is never perfect; its dimensions and shape vary from point to point, and small dimensional variations can make for significant impedance changes. Wire can suffer from periodicity (in fact, strictly speaking, it not only can, but always, at some level, does) because it's been drawn over a wheel that was microscopically out-of-round, and that periodicity will cause the wire to resonate at particular wavelengths, which can really wreak havoc. The plastic dielectric has to be consistently extruded to the correct diameter (and thousandths of an inch matter here!); if it's foamed, it needs to have highly consistent bubble size so that one side of the dielectric isn't airier than another, or one foot airier than the next. The two wires in the pair need not to wander in relation to one another; as they "open up" or are pressed tightly together because of tensioning on the wire-twisting machine (or tension applied to the cable by other handling, or by shield application, or...), or because the finished cable is being flexed, the impedance changes. The shield is a factor in the impedance as well, because both signal wires have capacitance to the shield, and if the foil is wrapped more tightly in one place and more loosely in another, that, too, will cause impedance to vary. (And these are just a few of the obvious problems; manufacturing processes involve other problems that nobody not involved in manufacturing would ever think of. For example, the lubrication that is used to assist in wire drawing needs to be washed off the wire before dielectric is extruded over it; what if the side from which a jet of cleaner is fired at the wire gets cleaner than the opposite side, and the dielectric winds up conforming differently to one side of the cable than the other? What about the other thousand things we, not working in a cable factory, have never even begun to think about?) As a result, although every manufacturer's HDMI cable is built to meet a nominal 100 ohm characteristic impedance, every metre of every cable is different from every other. The best one can do is to hold impedance within a range, centered on 100 ohms; the official HDMI spec calls for 100 ohms plus or minus 15%. The tighter that tolerance can be kept, the better the performance will be.
Worse still, impedance is not a one-dimensional characteristic. HDMI cable operates over an enormous frequency bandwidth, and impedance in a twisted pair is frequency-dependent. A twisted pair's impedance will rise relative to frequency; how much it will do so, and how evenly and regularly, will depend upon subtle physical characteristics. So, strictly speaking, no cable can actually be within tolerance for impedance over the whole operating range of the cable; it can only be within tolerance by the method the spec designates for measurement.
Impedance control is important for another reason: timing. As impedance varies, so will the time it takes a signal to travel down the cable. Electricity travels at nearly the speed of light; how close to the speed of light it travels depends on the dielectric, and is referred to as the "velocity of propagation." The objective, in putting together the four pairs in an HDMI cable, is to have them be identical; but in actual practice, each pair in a four-pair set will have its own delay. If the delay of one pair is sufficiently greater than the delay of another pair, the receiving device will not know which "red" pixel belongs to which "blue" and "green" pixel, or if the clock circuit is off, it may be impossible to time any of the colour signals reliably. Since this delay depends on the consistency and dimensions of the dielectric, and the consistency and dimensions of the dielectric are important factors in impedance, the same requirement for consistent impedance applies here; if impedance is too inconsistent, timing will be too inconsistent, and the whole system will fail.
One way of looking at cable performance is to chart the attenuation for a given length of cable against frequency. For any cable, attenuation (measured in dB) will increase with frequency; this attenuation comes from a few factors. Loss to resistance goes up with frequency, because higher frequency signals are able to use less and less of the cross-section of the wire (this is known as "skin effect") and so have less copper to travel through. Losses to reactance – capacitance and inductance – also increase with frequency. Then, what we call "return loss" adds the most irregular, and difficult-to-control, component to the loss. "Return loss" is the loss to impedance mismatch, and is so called because it represents the portion of the signal which is lost when, upon encountering a change in the impedance of the circuit (this may be a change in impedance along the cable, or a change of impedance on entering or leaving a connector, or a circuit board trace, or encountering a different impedance than expected at the load end of the circuit), it reflects back along the cable towards the source rather than being delivered to the load. While basic resistive and reactive losses are pretty reliable and have a definite relationship to frequency, return loss can be quite irregular. A graph of return loss against frequency, rather than showing a nice, consistent curve, is characterized by sharp, spiky lines. Why is this? Well, return loss has to do, more than anything else, with those manufacturing tolerances and their impact upon impedance. Every wire, at some level, has some periodicity, and so resonates somewhat at some unintended frequency. Every dielectric extruder fails, at some level, to extrude the dielectric consistently; every spooler that winds wire or dielectric-covered wire, every wire twister, every unreeler that handles that wire as it goes back into another stage of processing, every foil-wrap and drain-wire machine, every planetary cabler (which bundles and twists the pairs together with one another), every jacket handler and extruder--all of these machines, in all of these processes, apply microscopic irregularities to the cable which show up as return loss. Return loss can't be eliminated, at least not in a real-world cable; but it can be, within limits, made as small and as consistent across a range of frequencies, as possible.
Generally speaking, devices handle very linear or predictable losses very well. If one knows that one part of a signal will come in a thousand times weaker than another part, it's easy to "EQ" the incoming signal to boost the weak part to match the level of the strong part. But return loss can't be EQ'd out because it's too uneven and unpredictable.
Return loss, not resistance, is the critical consideration in determining the quality of an HDMI cable; if one were comparing cables with similar resistance, capacitance, and inductance values against one another, and consulting a chart of attenuation relative to frequency, what one would generally see would be that cables with superior return loss characteristics would show a flatter attenuation curve than the others. This is very important in HDMI because the required bandwidth for an HDMI signal is enormous, and the higher the frequency, the harder it is to control return loss.
Generally, in looking at HDMI cable products currently available on the market, we've found that these issues get overlooked. Instead of trying to control impedance well, which will result in flattening the curve on the attenuation chart, manufacturers generally try to control resistance. Why? Well, resistance is a lot easier to control. Bigger wire (smaller AWG number) has less resistance, and choice of materials can play a role, too (silver-plated copper is lower in resistance than bare copper, and bare copper is lower in resistance than tin-plated copper, for example). But as the frequency demands placed on the cable increase, bigger wire doesn't really help all that much (and, for a whole slew of reasons having to do with manufacturing process control, it can actually hurt), because it's not the total loss that's limiting performance; it's the non-linear component of the loss that's the real problem.