The CDT20R-T²L

Jitter - the major enemy of CD playback

Besides the limitation of CD's sampling frequency (which is becoming less and less an issue with improved filtering and replay methods which move the resulting artefacts outside the audible range) major sound degradation is caused by what the audio engineer calls 'timing jitter' or simply 'jitter'.

Quite a while after CD was launched, people started to realize that even extremely small timing errors in the 'bits' coming from a CD and fed to a digital to analog converter were audible, making CD sound harsh and non-musical. Here one must recognize the difference between the digital information encountered in a Personal Computer and that of a CD; the two are often incorrectly compared. A computer is only interested in the sequence of zeros and ones. It does not matter at all whether these bits come very slowly, say via a 300baud modem or via an extremely rapid 100Mbaud Ethernet link. Irrespective of speed, the information will be the same; it simply may take longer to get it.

The digital information on a CD, however, is extremely time sensitive. If the bits don't appear at the da converter at exactly the correct time it will generate sound elements which were not on the original recording. This in itself is easy to appreciate, but the accuracy of the timing is crucial and rather amazing. It is another demonstration of the human ears' ability.

So where could jitter come from (and my list will be far from comprehensive!): first of all lets assume that we have a CD player whose disc transport is so good that all information comes from the CD without any reading errors. Reading errors would mean some zeros or ones have been lost; some of them would be recoverable due to the error correction scheme (Reed-Solomon) stored with the original data.

Reading error-free data from the disc is not easy; quite often the laser tracking system has to work very hard to keep the laser in focus. This in itself puts more of a burden on the power supply system due to continuous peak current demands.

Data coming off the disk is timed by a in-built clock which in itself is often the cause for added timing errors. Whilst modern oscillators (which produce the timing baseline) are extremely precise, they are not perfect. And most importantly, almost all oscillators are extremely sensitive to vibration. So isolating the oscillator from the potential movement, both chassis mounted and air-borne is essential. However, the oscillator must be as close to the process as possible, but it is essential that its timing information is not affected by electrical noise, again either within the pcb or air-borne.

Sonic Isolation reduces jitter
The CDT20R already included much more sonic isolation than say the 8000CDM, but our intense listening tests, carried out when optimizing our flagship theatre AV32R av processor, led to improved knowledge of suitable damping material and its specific location. So applying this experience to the CDT20R, confirmed and verified by many hours of listening tests, that sonic isolation was indeed very beneficial. In actual fact a significant amount of the rebuild cost is caused by the special damping material used for the sonic isolation.

A revised S/PDIF interface reduces jitter introduced through the digital interconnect between transport and da converter.
Jitter is also introduced by the digital link between the transport and the outboard da converter. There are several reasons that jitter is added along this link: lack of bandwidth, electromagnetic noise, connector choice, imperfect screening, impedance mismatch and what we call differential impedance.

When designing the digital cable F3-10-DIG TAG McLaren Audio carried out innumerable listening tests to select the cable material, the connector and its screening. Very soon it became apparent that it was more or less impossible to design a digital cable which had the required bandwidth and the impedance stability along the cable without introducing jitter.

You may ask yourself why would a digital cable for transmitting audio signals be so difficult to design. We transmit very high frequencies down a coaxial cable in radio-frequency applications every day. The reason is the jitter. Remember that jitter is the timing imperfection of zeros and ones sent to the da converter. Zeros and ones are detected as a change between two voltage levels. A precise timing, independent of the detection threshold is only then possible when the voltage changes virtually immediately; in other words the voltage raises with an infinite slope. The disadvantage? To transmit a signal which changes so rapidly needs a signal cable of unlimited bandwidth and such a signal radiates quite heavily.

So what happens if our digital interconnect is not perfect, (no digital interconnect can be made with unlimited bandwidth)? Quite simply, the signal is modified, adding timing inaccuracies. But it is worse than that. If we want to transmit a digital signal we basically transfer data from one box (the transport) to another (the da converter). The connection is called a transmission line. This is a connection where the so called impedance becomes more important than its ohmic resistance. Changes in the impedance will lead to modification of the signal in form and shape and even to reflections, i.e. the signal travels backwards to interfere with the original signal. Smallest changes of the impedance will be significant if we need a transmission line for a signal where the transmission between the voltage threshold has to be performed rapidly. We found that it is almost impossible to generate such an interconnect. We can in the meantime measure the impedance along a wire, whilst usually the impedance is only measured as an average between both ends. We call the impedance as a function of its location along the cable differential impedance.

So where does this leave us? First of all in making a digital interconnect we need to make a cable which has a constant (or as much as possible) differential impedance, we have to use good connectors and we have to screen the cable very well to prevent signal degradation by adding very high frequency elements from the outside. Our F3-10-DIG is such a cable. Whilst being excellent, it is not perfect. So we have to conquer the problem of the cable's limitations differently. We have to generate a signal which can work with real digital cables, with cables of a limited bandwidth and a manufacturable differential impedance but does not add 'in-built jitter'.

The CDT20R already had a new S/PDIF waveform output circuit to generate a signal which was less critical to real-life digital interconnects (even very good ones). The circuitry added, in simple words, a very precise slope to the transmission between the signals. Our engineers added the slowest rise time they could make to stay within the S/PDIF specification and to work with all da converters connected (even if not a TAG McLaren Audio one). This improved the sound quality by resulting in less jitter, and made the combination of transport and da converter less sensitive to the digital interconnect, but it was just the first step.

Our experts started to model the output circuitry to generate a waveform, falling within the S/PDIF spec but with the least possible bandwidth requirement. Simply think of a smooth signal with a steep transmission only where it counts.

When first presented to our audio experts they were very sceptical about the idea as it was, in their opinion, quite an unusual approach for passing a signal with low timing errors. So it was agreed that some of these interfaces would be built and auditioned in blind tests. The result surprised our experts at TAG McLaren Audio - the computer models had made a big step forward. The result is, what we call, the latest generation of our improved S/PDIF output circuitry, resulting in improved sound due to less jitter introduced along the digital interconnect.

NB: even our latest generation of S/PDIF interface requires a good digital interconnect, such as our F3-10-DIG. Buying a stand-alone transport without investing in a suitable high-spec digital cable is a false economy. But a word of caution: audition the cable of your choice - there are more low-quality digital interconnects (some costing a fortune) than you might think as it is extremely difficult to make the correct one. And bear in mind, you may even have sample variations, so better buy what you auditioned and liked.

Having reduced the jitter at source through better sonic isolation, a sophisticated design and good screening, transporting the signal over a digital link with our improved S/PDIF interface still leaves us susceptible to jitter. And here's where the final modification comes into play.

The TAGtronic Sync Link reduces jitter
The theatre AV32R av processor is widely acknowledged for its excellence in reproducing movie sound, being a marvellous preamplifier and outstanding 24bit/96kHz da converter. The AV32R features also what we call the TAGtronic Syncronisation Link (often referred to as Tagtronic Sync Link), a link which allows the AV32R to use its own clock as the master, even with external digital source equipment.

The ideal (theoretical) transport (whether CD or DVD) is a transport which sends the zeros and ones to a da converter without the timing information. This is easier said than done, as it would either require storage in the da converter (some very high-cost CD players offer this option) or the transport and da converter need to synchronize their clocks. One way, often employed in a 'two-box' solution is to send the timing information down a separate link from the transport to the da converter. This - at first glance - seems a good idea, but then, as we have seen above, it is basically impossible to send any critical timing information over a distance without introducing timing errors, or jitter. So it is much better to synchronize the timing clock within the transport to the master clock within the da converter. In other words the da converter carries out the timing of the data, and tells the transport how fast or slow the bits need to come. In order to achieve this the clocks in both products need to be continuously synchronized, which can happen in a control loop between both products. This has to be done pretty fast as otherwise bits would be lost. So the TAGtronic Sync Link continuously 'adjusts' the clock within the CDT20R-T²L transport to provide the AV32R with the data at the correct timing. The information, as to whether the clock within the transport is too slow or too fast, is sent as an analog signal (which can be sent along long distances). But the link still requires a well-screened analog interconnect, such as our F3-10-ANA, to prevent 'noise' which might affect the link.

Another beauty of the Sync Link is the fact that it is not limited to one single unit; in actual fact the TAGtronic Sync Link can be chained, so you will be able to connect both a CD player and a DVD player (and other digital sources equipped with our Sync Link) to our AV32R and benefit form the Sync. The result is a virtual elimination of jitter, resulting in a better stereo image, better transparency and greater musical purity. Another layer of sound perfection.

So after the better sonic isolation and the improved S/PDIF waveform circuitry, the third addition to the CDT20R-T²L is the TAGtronic Sync Link interface, allowing it to receive the control signals from the AV32R.

Bear in mind, even if you do not have an AV32R (yet!), both the improved sonic isolation and the latest generation of our improved S/PDIF waveform circuitry work with all da converters. So upgrading to the CDT20R-T²L its already a worthwhile upgrade - until you have added the AV32R to your equipment.

The TagMclaren sign is a registered trademark of Mclaren Group Limited. Copyright 2004 Pointfield Limited - a member of the International Audio Group.