My feathers weren't necessarily ruffled! It is a little galling however when you talk about degrees and so on when it is quite plain that you know very little about cable technology. I will attemp to explain why, once and for all, cables make a HUGE difference to perceived sound, I have included some articles from external sources. This is a long post so bear with me - but you never know 1138 you may just learn something as I have included some information gelaned from the computer application of digital technology...

Due to the nature of digital signals and their processing, standing waves (signal reflections due to impedance mismatch) are very common in digital interconnections. So called interface induced jitter due to inter symbol interference is the result. Jitter is a commonly acknowledged problem within the industry and has been covered in various magazines including Stereophile, Widescreen Review and so on. It is a proven fact that jitter causes variations in sound quality. Jitter is not only caused by the machinatuons of the components digital clock but also by reflections within a cable.
The scientific application of 75 Ohm coaxial (SPDIF) or 110 Ohm balanced (AES/EBU) connections is reliant upon electrical behaviour and the chemicals used in the manufacture and damping of said cable may again cause variations in perceived sound.

The following is an article written by a highly respected authority on cables and will give you some idea as to the merits of different cables. We will concentrate on the most part on electrical cables but I assure you that this includes fibre optic cables as well.

Your house’s brick walls are made from a collection of bricks of reasonably equal dimensions, built together with cement. The walls are entirely rigid, this undoubtedly to your and your house companions’ great appreciation.
Cables in essence do not deviate that much from this picture.
Here we also have a collection of parts that among others are connected by means of so called “van der Waals forces”.
When observing metals in cables these “parts” are named crystals and, instead of cement (in your brick walls), we are dealing with their boundary surfaces. These edge surfaces originate from the cable’s production and especially from the time elapsed afterwards.

The production process of conductors involves massive forces on the metal wire and long waiting stops, the latter during which the bare metal remains exposed to the (polluted) open air.
Both mentioned forces and air exposure have quite an influence:
In addition to the mechanically induced formation of new edge surfaces and crystal defects during production, the wire’s crystal boundaries during its bare storage and processing are most prone to chemical reaction with oxygen and other airborne chemical compounds which forms chemical layers on the crystal edges.
This affects all metals; those that are closer to the noble metals in the periodic system of elements to a lesser amount than those that are more remote.

A metal suitable to be used as the conductor in e.g. an audio cable would be rather soft, pliable and ductile, have a low specific resistance, be quite insensitive to corrosion (noble) and be available in large enough amounts (price): regarding its good balance between all properties, of all metals, copper is the most commonly used electrical conductor.


Your environment’s and your own influence on your cables:
Copper is a reasonably suited material since it is soft, quite tough and has a low specific resistance. The formation of internal boundary surfaces due to mechanical and chemical effects however is unfavourable. As a result, the initial quality of signal transfer does not stay forever, but rather is influenced by all kinds of environmental and handling factors.
And this is just where the shoe pinches.


Cross section of a copper wire heavily broken down by bending and chemical corrosion: Severe damage and crystal boundary contamination are discernible. Cross section of an oxygen free pure copper wire as normally found in audio cables: Individual crystals and their boundaries are clearly visible.

Stating that a cable wouldn’t have any influence on sound at all would be the same as saying “I don’t care what colour my glasses are; all colours always look equally natural to me”.
Or: “I don’t care how thick (and long) my garden hose is; the garden sprinkler always works equally well”.
From your experience you probably don’t agree with these statements.
With the earlier given explanation on important parameters, by now you know that also with audio cables differences are to be expected.

The audio signal transfer through a cable for its most important part is determined by the conductor’s properties. With metal conductors besides that, the cable design and insulation materials used clearly take a second role.
With perfect conductors (hence in principle not with metals) latter two aspects suddenly do become important.


The world’s best cable:
The ideal audio signal transmission in a cable can only take place through a material totally clear of internal boundary surfaces where the material neither has mechanical or chemical decline present nor has these encroaching with time.

Your undoubtedly very fine first bike has gone through rust and metal fatigue. And the several cars you may have had during your colourful existence due to same causes also already have ended at the iron foundry.

In their signal transmission behaviour, cables are much more critical than the conductive properties of your old bicycles.
Signals below -100 dB are important in determining the spatial definition in a recording.
With a bad cable these signals are at risk and/or decline into distortion. The harmonic structure is altered, producing more listening fatigue than listening pleasure.


The way many people listen:
With room reproduction in our country it is a custom not so much to refer to reality but rather to strive for one’s own sound image. This including further attempts to “dot the i’s and cross the t’s” of this idealised image by means of swapping components. With this, cables often are used as the “pepper and salt”; the flavourers and taste enhancers.
Nobody can hold this against you, but when using these flavourers to a considerable amount and, above all, in an unconscious manner, it wouldn’t be right to boldly deny having touched ye olde salt and pepper set.
Now you know what the consequences can be, it counts to retrieve the trail, realising yourself what truly counts.


Terminology:
Time (delay) compensation (an in connection with cables often used word) would only matter when you would be able to hear music up to at least 10 MHz. With common audio signals (and a common speed of propagation of the electrical signal in a cable of 200,000 Km/s), a single degree phase shift more or less does not matter. After all, your audio signal traverses 1 meter in 1/2,000,000,000 of a second, which is 5 nanoseconds. Within the same time span in air, sound covers a distance of only 1.72 micrometer. There are still listeners thinking they can detect this, whereas their ears may be as much as 1 mm. out of the straight with their eardrums 4.5 mm. in diameter and positioned at 45 degree angles. And what to think of your loudspeakers? Their distance to you often is unequal by more than 1 mm.
By the way: 1.72 micrometer is the 1/581-th part of a millimetre. Therefore you don’t have to worry too much about that anymore.
Likewise, the so called oxygen free kinds of copper are a nice and fine finding. However....., the copper concerned was only free of oxygen during its production. At your home this for a long time isn’t the case anymore. Only the printing on the cable stating “Oxygen Free Copper” still might be free of oxygen.

This shows the very basics of choosing and using cables and the differences that conductors can make to not only the quality of signal but also the delay of said signal. All of these factors together effect sound quality heard by the human ear whether used within a Hi Fi or Home Cinema.

The following paper deals with digital technology as a whole - right from its' inception to the very latest technology - we will 1138 be taking in computers along the way!

Digital audio arrived on the scene 20-odd years ago with hype - and problems. Everything from mathematics to the New York Times was invoked to claim that digital was perfect. But the truth was that it sounded awful in most instances. LPs made from digital sounded bad, but CDs were supposed to reveal the real glory of digital. Then when CDs showed up, they sounded much worse.
The contrast between the promises and the reality polarized the audio world. In one camp, people who believed in the theoretical possibilities tried to look on the positive side and often minimized the seriousness of the problems of digital as it existed. In the other camp, people disappointed by the actual sound often became skeptical of the very possibility of digital's becoming good. (I ignore the cynics who just tried to make money by pretending that CD was perfect.)
The polarization made it hard to think about the issues in a clearheaded way. It was not easy to find impartial information on the vital question of whether the sonic inadequacies of CD arose from intrinsic inadequacies of the CD ("Red Book") standard itself or simply from inadequacies in the engineering execution of digital recording and playback within the standard.
Over time, some of these things have sorted themselves out. Much thought has been given to refining aspects of digital audio engineering, and CD sound quality has greatly improved as a result. But some residue has remained of the original perception of digital disaster.
Many who work professionally in digital audio have come to share the audiophile perception that improved sound could come about by an improved digital standard, in a way difficult or impossible within the CD standard. The "standard" means really two things: first, sampling rate, that is, how many digital "words" are used per second; second, word length, that is, how many binary digits or bits are in each word. For CD, the standard is 44.1 kHz sampling rate with 16 bit words. So CD provides 44,100 words per second with each word consisting of 16 digits, each either a zero or a one.
For reasons I shall try to explain, it seems that there are both not enough words per second and not enough digits per word. Of course, from the data transfer viewpoint, it is the total number of bits per second that really counts. But it is convenient and conventional to talk separately about the sampling rate and the word length. And in this context there are serious sonic reasons to increase both.
Furthermore, there is an excellent opportunity to make the improvement painlessly. The new video format, DVD, has enough data transfer capacity to accommodate "96/24" - 96,000 words of the length 24 bits per second. In fact, it already does this, in effect. This sampling rate and word length could well become the new music recording and playback standard. What I am going to discuss is why this ought to sound better and how it actually does.
This new standard would have exciting implications for the improved reissue of recordings that already exist if these have greater information content than CD has so far been able to deliver. But it has even more exciting implications for recordings in the future. Think of it this way- if a medium has what it takes to preserve the magic of old (analogue) recordings, it is bound to have what it takes to give new digital recordings equal or greater sonic magic. And without the hiss, wow and flutter that analogue tape is heir to.
A truly transparent recording medium, in which input and output are actually indistinguishable in listening, is all but here. Anyone who has ever listened to a "live feed" from a microphone of high quality and then heard what happened when that feed was recorded, knows what an important step forward in musical realism this will be.
The computer jargon description of a digital standard describes digital audio in terms of data transfer: so m my words, or numbers, per second with each word containing so many bits.
What Do Those Numbers Really Mean?
They are the values, written in binary, of the audio signal as a voltage at the particular moment that the number (or "word") is transmitted. So digital audio tries to describe the audio signal voltage by giving its value at many equally spaced points in time.
This idea is familiar to us in everyday Iffe. If you wanted to describe the variation of temperature during a day for a weather report, you might do it by writing down the temperature every hour with accuracy to the nearest degree, say. If you wanted to give more information still, you could improve in two ways: you could write down the temperature more often - every minute, for instance. And you could write it down more accurately, say to the nearest tenth of a degree.
The same things apply to digital audio. You can make it more accurate by writing down the signal voltage more often - higher sampling rate. And you can write it down each time more precisely - longer words, more bits in each word. Each of the two kinds of increased information gives a better description of the signal, but they are better in different ways.
Of course even your improved shot at describing the temperature every minute with 0.1 degree accuracy is not perfect. For one thing, you are not giving the temperature each time exactly-you could have gone to the nearest 0.01 or 0.001 of a degree and even those would not have been truly exact. Second, if some temperature variation occurs really fast, you could miss it. It could happen hetween your sample times. If you are sampling every second, a little temperature blip that happens at the half-second marks in between will be missed.
These two types of errors have names. The first lack of precision in the values of the samples, is called quantization error or quantization noise (because it is a kind of noise, in the sense of that the signal is acquiring noise). The second lack of precision is called aliasing error.
Aliasing error is serious. If the temperature is changing really fast and you are writing it down only every hour, you might miss a major event if it happened to occur on the half-hour mark Your weather report could be truly misleading. Clearly, if you are dealing with a rapidly changing signal, you had better do frequent sampling.
Aliasing error is so bad potentially that it has to be gotten rid of from the start. And it is pretty clear how to get rid of it: you have to rule out signals that change too fast compared to how fast you are sampling, or, looking at it the other way around, you need to sample fast enough to be able to deal with how fast your signal is changing. In audio, this means that you need to filter out the really high frequencies - all frequencies at or above one-half the sampling rate. (I'll explain why in a minute.) This means that CD digital with 44.1 kHz sampling is good only up to 22.05 kHz in theory- actually, for a safety margin, the signals are allowed to run up only to 20 kHz.
So sampling rate determines the upper frequency limit. Ninety-six kHz will give an upper limit of half 96, or 48. This is good. It is about as high as the very best analogue tape and about the same as the practical upper limit of vinyl
Now, you can understand the sampling rate's having to be at least twice the highest frequency by a visual analogy. Suppose you are watching a pendulum swinging back and forth under a strobe light. Suppose the pendulum completes a cycle back and forth in one second -1 Hz, in other words. If the strobe flashes fast, say ten times a second, then the pendulum's motion will be visible. It will look jerky, but you will be able to see what is happening, what its real motion is. But if the strobe flashes too slowly, the whole situation will become ambiguous. If the strobe flashes twice a second, then it could happen that every time it flashes, the pendulum is straight down. The pendulum is pointing straight down twice a second, once going left, once going right) if those moments occur when the strobe flashes, then, of course, the pendulum will appear stationary. AH information about its motion will be lost!
If you slow the strobe down even more, to a little less than 2 Hz, the pendulum will appear to oscillate back and forth, but very slowly, at a frequency much less than its actual 1 Hz. We have all seen this on the stagecoach wheels in Western movies: as the wheels speed up in reality, that is, as the wagon accelerates, the wheels first rotate realistically, then later, as speed increases, they slow down visually in the film. Then they stop and then start to rotate slowly the wrong way.
This would happen in audio, too. If the signal contains frequencies higher than one-half the sampling rate, then - as with the 1 Hz pendulum watched by the less-than-2 Hz strobe - the too-high frequencies will fold back down to a much lower frequency. So frequencies that are too high compared to the sampling rate generate lower frequency error signals that are, in effect, unrelated to the music. That is why aliasing error is intolerable and signals have to be heavily filtered to eliminate frequencies higher than half the sampling rate.
This works the other way around, as we H. If the signals are so filtered, then there is no aliasing error and the sampling will give you complete information about the signal except for the quantization error. (Both amplitude and phase are correct in the reconstructed version of the filtered signal.) People often want to know happens to the signal value between the samples. The answer is that m order for a signal to be between samples entirely, it would have to be so fast that it would necessarily have frequency components at or above half the sampling rate. If there are no such components, then what happens between the samples is completely determined by what happens at the sample times. This sounds a bit strange in description, but intuitively it is believable - if the pendulum is going to go left from its center position and get back to center within one sample period, it has to make a whole cycle within two sample periods. And that means its frequency has to be at least half the sampling frequency!
Sampling rate determines the upper frequency limit. And word length also has a direct meaning in audio terms. It determines the dynamic range of the system. The difference in dB between the loudest and softest that the system can register is six times the number of bits in the words. (Exactly, it is 2log3 times, about 6.0206.) For CD's 16 bits, the dB range is 6 x 16, or 96 dB. So the nominal dynamic range of CD is 96 dB. However, the dynamic range in listening terms can be considerably greater if the system is used properly, as we shall see in a moment.
To put this figure of 96 dB dynamic range in perspective, it is natural to look at what the corresponding figures are for the vinyl record and for analogue tape. This is more complicated than it might appear, however, because the noise of vinyl and of tape is not uniformly distributed over the frequency range. At the higher frequencies ( I kHz on up), vinyl typically has about 80 to 85 dB between its maximum possible undistorted signal level and the minimum possible groove noise. (This is not to say that commercial records ever attain Anything like this figure m practice.) The figure is somewhat variable, since the maximum allowable signal level is lower at the inner grooves than at the outer. At low frequencies, the noise goes up - way up - and the maximum signal level goes down. In the bass, a record is lucky to have 30 dB of signal-to-noise ratio. You can hear this bass noise for yourself by turning up the volume on a "silent" groove. However, and a big however it is, the ear's relative insensitivity to bass and the bunching together of the Fletcher-Munson curves of equal loudness makes that 30 to 35 dB in the bass equivalent to a much larger range in terms of audible dynamic range expressed in midrange terms. Human hearing itself has much less dynamic range in the bass! So one really has to think of the dynamic range of records in "weighted" terms, where low frequency noise is partially discounted. In this sense, 80 dB is a reasonable approximation to the theoretical dynamic range of a record.
Analogue tape requires less in the way of weighting to get a meaningful figure. With a plausible and not too extreme weighting, the signal-to-noise ratio of workaday recorders and tape formulations is around 75-80dB. The quietest tape formulations combined with an exotic recorder like Tim de Paravicini's one-inch machine can reach 90 dB or very dose to that. In this regard, it is worth noting that the recorders and tapes of the early days of stereo were much noisier than this. The old big-particle tapes were robust, but they were noisy. And so were the machines. (A quick look at old Audio reviews will reveal that, for consumer recorders, 50 dB was a more than respectable figure. The pro machines were better, but nothing like modern standards was envisioned for the combined noise of tape and machine.)
The bottom level of the dynamic range of CD is a definite level above zero, and this irreducible interval is in effect the source of quantization noise. One cannot distinguish levels any better than the difference between all zeros and a single one in the last place (the "least significant bit"). But at first sight this quantization error does not seem so bade So what if we make a little bit of error in exactly how loud the signal is? And in fact, quantization error does not make much difference as long as the signal is loud. For a large amplitude signal, the quantization error is small as a percentage of the signal, and the ear tends to judge the importance of errors in terms of their percentage of the signal, rather than on their at solute size.
The big problem comes when the signal, or some audibly separated and important part of the signal like a soft accompanying instrumental line, is low in level compared to that lowest level that the system can record. Then the quantization error is a larger percentage of the signal. And - bad news - it is harmonically related to the signal itself. For, say, a pure tone, the jumps from one digital level to another occur m a pattern that has the same periodicity as the tone itself. So the quantization produces harmonic distortion, and harmonic distortion with lots of high-order harmonics at that (when the signal itself is in the middle range). That high-order harmonic distortion was what made early transistor amplifiers sound so bad.
And that is what it did to early CD, too: screaming attached to the signal, fimgernails on blackboards, the horrffyingly edgy, nasty sound of high harmonics in large quantifies. And large they were - a midrange signal that is 90 dB down from the top level has more than 25 percent harmonic distortion in this setup. Help is at hand, however, from (unlikely as it may seem) mathematicians. In working on digital signal processing in general, they figured out that if you add a little noise to the original smooth signal, the quantization error decouples itself from the signal and just becomes part of the noise. Moreover, signals below the least-significant-bit level become observable. The whole system becomes like analogue - infinite resolution except for the signal's gradually fading into the noise floor as its level goes down. (The resolution in time also becomes infinite in the same sense, but that is a little harder to see.)
Now something amazing comes into play - the ear/brain's ability to hear into the noise floor. With proper "dither," as the added noise is called, one can hear far below (about 15 dB below) the least significant bit. So material at 105 dB, even 110 dB below the top level, is still audible in the CD standard, measured relative to the top level. Properly dithered CD has a lot of dynamic range and low levels of harmonic distortion at all signal levels. But if you do not do the dithering, the dynamic range and resolution are much worse and the system has a lot of harmonic distortion at low levels. Dither is indispensable!


Most of the discussion in the main article has been based on theory - on the theoretical information capacity of the various digital standards. But we don't listen to theory, and it is natural to ask how well the theory can be put into practice. In particular, is true 96 kHz/24bit recording really possible? Is it available now? And what about 96/24 playback?
The 96 kHz sampling rate is not a problem. It seems fast in human terms, but it is actually a slow "cycle time" by computer standards. Sending 24-bit words around at that speed is no problem, either. Computers do much faster data transmission routinely. And the required data transfer off DVD is already up and running. In short, in the purely digital realm, we are just cruising. The problem is how to get to the digital realm in the first place with this kind of accuracy.
Specifically, the problem is the incredible accuracy involved in the 24-bit words. Think about what is involved: one has to measure a varying voltage to an accuracy of one part in 2 x 2 x 2 ... x 2 (twenty four 2 factors), or one part in about 17 million. The accuracy involved here is extreme if one is to have the resolution to justify calling the process true 24-bit.
For making accurate measurements, the measuring device itself must be very stable. You can't weigh yourself accurately on the bathroom scale in the middle of an earthquake. In electronic terms, this translates into, "An accurate measuring device must have low internal noise." To get the precision involved in the 144 dB of maximum to minimum in 24-bit digital, the analogue-to-digital converter must have 144 dB of signal to-noise ratio. And there's the knot.
To put this in perspective, note that the electronics of the High End audio world (e.g., Bryston, Boulder) have S/N ratios between 120 and 130 dB. This is really impressive in itself. To accommodate the full 120dB of dynamic range of 20-bit digital in a power amplifier, say, is already a substantial accomplish meet. But it is a long way to 144. To get from 124 to 144, for example, the equipment has to be 100 times quieter!
We have almost come to believe that technology can do anything. But there are limits. In this case, we are coming up against one of Nature's own barriers. Basic electronic parts, the devices of which components are made, have internal noise. And 144 dB down is getting into the realm of the intrinsic, thermal noise of that most basic device, the resistor. The Brownian (thermal) motion of air molecules sets a lower limit for human hearing. And similar things limit electronics. There are ways at audio components could be quieter than they are. Low-noise scientific experiments use battery power, for instance, and audio could do this more widely, too. But short of coming the equipment to cryogenic temperatures, one is still going to bang up against the thermal noise barrier. |
As to playback, similar problems arts' in making amplifiers, as noted. And consider that no speaker with 144 dB of signal-to-noise ratio has ever existed or is likely to exist in the foreseeable future. Speakers are mechanical devices and they make noise when they move. Ask a speaker designer if the extraneous noise from playing a midrange tone can be gotten 144 dB down from the tone, and you will get a rueful laugh. We are, after all, living in a world where 0.1 percent distortion is admirable, even superb, performance for a speaker. That is error 60 dB down from signal. Of course that is different from the noise figure, but it gives the ballpark range. 144 dB is miles away, and a promise no one is going to be able to keep.
The whole 24-bit thing could begin to seem like a scam - A to D impossible, D to A and amplification impossible, speaker behavior not even remotely close. But the idea that this is all some sort of fraud would be mistaken. The digital world has curious and subtle, even paradoxical, features. One is that it is a good idea to have some extra bits even if they would seem to consist only of noise.
There are several reasons for this: first, the ability of people to hear into the noise. Even if the last few bits are mostly noise, there is still some signal there and we can hear it. Natural sound involves signals' fading into noise (think of the ring of the final chord in the concert hall as the reverberation fades away). This is natural and it is what makes reproduced sound seem natural. Second, there is the need for extra bib: for processing, as I have already mentioned. Third, there is the fact that human hearing is logarithmic. The audible significance of a signal is a matter of the percentage that the signal is of its surroundings. So as overall volume drops, tiny things become more significant. (One of the interesting developments in speaker design currently is a project of DALI to develop drivers that respond well to very tiny signals - down at or below the absolute threshold of hearing. This elimination of low-level "sticking" behavior has surprising implications in listening.) Finally, the extra bits tend to make the slightly higher level bits: function more linearly. In practice, a system that goes to 24 bits, even if the last four are noise, is likely to have really good linearity through 20 -bits.
In summary, don't get too skeptical '"True" 24 bit, meeting the S/N spec, isn't here and may never get here. But those last bits are still worth having. It is import tent to get big things right, but in the digital world, the little things count, too.

I hope that this post has given you all a clearer picture of digital technology. In theory we should have perfect sound but as can be seen - the reality is very far from that. There are so many tiny variations in so many components within our home cinema and hi fi systems that have both positive and negative effects. IIt has been well proven by scientists greater than I that differing cables imbue a soundwave with a particular character - this cannot be refuted by me or 1138. I urge all who read this post to take cabling very seriously - it can make or break a hi fi and home cinema system.

If any of you would like further information or would like to read any further papers, please get in touch with me personally.

Kind regards

Justin Pledger
Dolby Laboratories