This Sticky Thread starts off with a really great summary of the concepts behind transmission of digital signals via cable by member, House de Kris.

WARNING:

This thread is intended as a technical reference (for learning) , and we reserve the right to edit the hell out of any post. No non-technical opinions, please? Meaning that this is not a place to compare cables, or to express your long-held belief that analog is better, or silver is better, or whatever. All such comments will be edited out.

All comments amplifying, explaining, clarifying, questioning, or expanding on Kris' comments are very welcome, as are technical questions of all kinds. They're not only welcome, they're encouraged!

DIGITAL INTERFACES USED IN AUDIO – SPDIF COAX

There are a number of interfaces used in the audio world for moving digital audio data from one component to another. Like chess, understanding these interfaces is simple on the surface, but complexity escalates exponentially as the layers of the onion are pealed away. To keep things simple, this thread will fixate on the SPDIF on coaxial implementation of digital transfers only. Others methods include SPDIF on Toslink, AT&T glass optical, AES/EBU professional. This is not intended to be an engineering tutorial on the subject, but rather a layman’s aid in understanding the terms and issues involved. As such, engineers may not be too satisfied with this discussion, but remember, the intended audience is the layman.

THE DATA
The intent of the digital interface is to get data from one location and ship it to another. In this case, SPDIF is a protocol that defines what the format of the data is, and how it should be interpreted. In addition to the bits that define the instantaneous sampled levels used to reproduce the audio, the SPDIF interface also includes the clock that tells when each of the samples should be used to recreate the audio signal. Therefore, the destination is always a slave of the source. SPDIF can support multiple clock rates and multiple word widths. SPDIF uses a bit modulation technique called bi-phase. Bi-phase is just that, a modulation scheme for the data, it provides no error detection and correction.

THE PHYSICAL INTERFACE
The coaxial SPDIF interface specifies a 75ohm controlled impedance environment for the signal. This implies a 75ohm source, a 75ohm interconnect, and a 75ohm destination. This “75ohm” designation is of the characteristic impedance of the transmission line that carries the signal of interest. Being an impedance, it is an AC phenomena, not DC. It cannot be measured with DC instruments. But, it is a very real 75ohm load being presented to the incident waves traveling down a PC board trace or wire. In any controlled impedance digital interface where edge placement accuracy is desired, mismatches between any of these three primary components (source, interface, destination) will cause edge distortions which could cause timing related errors later on downstream. If all three primary components are perfectly matched, the waveshape developed at the source will appear exactly at the destination assuming a lossless interface, no matter how long the interface. Lossless cables don’t exist in the real world, though, and the edge degradation due to bandwidth limitations will degrade the edge fidelity.

REFLECTIONS
In a controlled impedance environment, when the edge of an incident wave propagates down the line, it is both an incident voltage wave and an incident current wave. In a 75ohm line with a 1V edge, it will have a 13.3mA edge associated with it. If the characteristic impedance of the line changes (called a discontinuity) as the signal travels the line, a different current is required for the same voltage in order to satisfy Ohm’s Law. The excess voltage or current that results will be reflected from the discontinuity point and send smaller versions of the incident wave back down the line in the opposite direction from where the discontinuity occurred. This reflected energy will keep going back and forth down the line until it is entirely absorbed in the source, destination, or interface. If there are a number of discontinuities in a line, there can be many reflections heading in either direction at any point in the line.

TRANSMISSION LINES
A transmission line occurs anytime a signal carrying conductors has a specific relationship to ground. The signal will create magnet flux lines between the conductor and ground. If the space between the conductor and ground is consistent, and the dielectric properties of the insulating material are consistent, then a characteristic impedance is developed. Thus, on a PC board it is relatively easy to create traces with a specific characteristic impedance. By specifying the thickness of the board between signal and ground, the width of the signal trace, and the board material, just about any impedance is possible (in reality, it is difficult to get greater than 80 ohm traces with today’s processes). Keeping everything else constant, making a trace wider will lower its impedance (capacitive discontinuity), while making it thinner raises it (inductive discontinuity). From this it should make sense that a typical via in a PC board appears to be a capacitive discontinuity in the transmission line. This could happen when a trace moves from one layer to another.

Likewise, in coax cables the impedance is set via physical parameters. In this case, it is the distance from center conductor to shield (which is set by the thickness of the insulator), and the properties of the insulator (dielectric constant). Any variation in the distance from center conductor to shield will show up as a change of impedance, or discontinuity.

COPPER OR SILVER?
The type of metal used for the center conductor has no effect on the speed of propagation down the line. Silver does have greater conductivity than copper, and this is why it is sometimes specified. Silver is more expensive than copper, though, so the typical use of silver is to plate a copper center conductor to minimize losses due to skin effect at high frequencies. Losses due to skin effect look like bandwidth limitations in the frequency domain.

TEFLON OR POLYPROPYLENE?
The dielectric choice directly impacts the speed of propagation of the incident wave. It will also have an impact on the distance from center to shield. All other considerations being the same, a cable with a polypropylene dielectric will be thinner than one with Teflon. The polypropylene dielectric cable will also have longer propagation times. Air-impregnated Teflon, or Teflon foam, is superior to solid Teflon for speed purposes.

REALITY SETS IN
With the spec for SPDIF on coax so well defined, and no new science being developed for this interface, it would seem logical that all manufacturers of audio equipment and cables would be able to abide by the simple rules of maintaining a 75ohm path throughout the equipment. Unfortunately, this is not always the case. When source, interface, and destination don’t adhere to the standards that would ensure the best possible situation, we are left with always dealing with trying to optimize the sub-par. Thus, changing things that should, on the surface, have no logical reason to affect the sound, does.

MEASURING CABLES
There are ways to measure the characteristic impedance of cables. The most common is TDR (Time Domain Reflectology). A TDR instrument will send a fast edged step function down a line, and then interpolate the reflected energy into equivalent characteristic impedances at any physical point in the line. That way, the displayed trace then directly represents the cable being measured. A TDR can also make measurements as to the minimum and maximum reflected energy. Another similar test is TDT (Time Domain Transmission). This instrument will observe the signal on both the souce end (like TDR) but also at the destination end. This allows easy measurements of propagation delay and loss.

It should be noted that lumped capacitance values for coax used as a transmission line has no value. The capacitance is balanced by the inductance to yield the characteristic impedance.

In a related series of posts, House De Kris is kindly measuring many digital cables via TDR. Look in the "Thinking Out Loud" forum and read threads beginning with "TDRing..."

Impedance mismatch and reflection has more to do with much higher frequency range such as video in the mega and giga hertz. Yes there is plenty of reflection going back and forth from source to amp to speaker but it's happening at such a high frequency that our audio range is not essentially affected.

Transmission of long ranges (hundreds and thousands) of feet is where the problem starts for video people. In our audio world, our cables are in the 10's of feet or even less. There is not much to worry. For the purists, yes, there might somethings you could do but not real worth the money. And even if you were affected by these waves, you will never hear them. The average male even in the 30's is below 20khz.

wsjoe

True, the notion of reflections is very appropriate for high frequency systems, not audio. But, a digital audio interface has bandwidth requirements way beyond audio, into the video range. In addition, it is necessary to have sufficient bandwidth to ensure accurate edge placement at the receiver. Having enough bandwidth to make the data look square-ish is not enough. When sub-nanosecond timing accuracy is desired, extra care must be used when shipping digital data from one point to another.

I completely agree with you about any reflections between amp and speakers as having no effect at all. But, that's not what this thread is about.

I don't support the notion that having a digital signal where each frame is as square as possible in the corners would contribute to a better signal decoding at the D to A end. The digital signal is fed to a level shifter, or Schmitt Trigger, that decides if it is a valid 0 or valid 1. The actual incoming signal is NEVER used, but serves as a guide towards recreation of a new data stream that reflects the incoming 0s and 1s.
Where the problem arises is when the clock at the source end is unstable, so the waveform jitters about in time. The Schmitt Trigger cannot correctly identify the start or end of the jittery square wave. Sometimes this problem is tackled through the use of a Phase Locked Loop clock circuit, but selection of the maximum and minimum range over which a PLL can accurately lock in and onto a jittery waveform is not for the faint hearted designer involved in precision clock designs. Harmonics can easily defeat a PLL. Cable reflections and standing waves are the kind of waveforms I have in mind.

Thanks, its a good article to read and understand. Much appreciated.