Does the quality of a digital signal matter?

The points Kijanki made about timing, jitter, and reflections on impedance boundaries merit added emphasis and explanation, imo.

The S/PDIF and AES/EBU interfaces which are most commonly used to transmit data from transport to dac are inherently prone to jitter, meaning short-term random fluctuations in the amount of time between each of the 44,100 samples which are converted by the dac for each channel in each second (for redbook cd data).

As Kijanki stated, "Jitter creates sidebands at very low level (in order of <-60dB) but audible since not harmonically related to root frequency. With music (many frequencies) it means noise. This noise is difficult to detect because it is present only when signal is present thus manifest itself as a lack of clarity."

One major contributor to jitter is electrical noise that will be riding on the digital signal. Another is what are called vswr (voltage standing wave ratio) effects, that come into play at high frequencies (such as the frequency components of digital audio signals), which result in reflection back toward the source of some of the signal energy whenever an impedance match (between connectors, cables, output circuits, and input circuits) is less than perfect.

Some fraction of the signal energy that is reflected back from the dac input toward the transport output will be re-reflected from the transport output or other impedance discontinuity, and arrive at the dac input at a later time than the originally incident waveform, causing distortion of the waveform. Whether or not that distortion will result in audibly significant jitter, besides being dependent on the amplitude of the re-reflections, is very much dependent on what point on the original waveform their arrival coincides with.

Therefore the LENGTH of the connecting cable can assume major importance, conceivably much more so than the quality of the cable. And in this case, shorter is not necessarily better. See this paper, which as an EE strikes me as technically plausible, and which is also supported by experimental evidence from at least one member here whose opinions I respect:

http://www.positive-feedback.com/Issue14/spdif.htm

Factors which determine the significance of these effects, besides cable length and quality, include the risetime and falltime of the output signal of the particular transport, the jitter rejection capabilities of the dac, the amount of electrical noise that may be generated by and picked up from other components in the system, ground offsets between the two components; the value of the logic threshold for the digital receiver chip at the input of the dac; the clock rate of the data (redbook or high rez), the degree of the impedance mismatches that are present, and many other factors.

Also, keep in mind that what we are dealing with is an audio SYSTEM, the implication being that components can interact in ways that are non-obvious and that do not directly relate to the signal path that is being considered.

For instance, physical placement of a digital component relative to analog components and cables, as well as the ac power distribution arrangement, can affect coupling of digital noise into analog circuit points, with unpredictable effects. Digital signals have substantial radio frequency content, which can couple to other parts of the system through cables, power wiring, and the air.

All of which adds up to the fact that differences can be expected, but does NOT necessarily mean that more expensive = better.

Regards,
-- Al

P.S: I am also an EE, in my case having considerable experience designing high speed a/d and d/a converter circuits for non-audio applications.

Mceljo, with all due respect your friend seems to have missed my point.

My point was NOT that bit errors would occur in the link between transport and dac, due to logic threshold problems or due to any other reason. I would expect that any such interface that is not defective, and that is Walmart quality or better, will provide 100% accuracy in conveying the 1's and 0's from one component to the other.

My point in mentioning the logic threshold of the receiver chip was that variations in its exact value, within normally expectable tolerances, may affect whether or not the receiver chip responds to reflection-induced distortion that may be present on the edges of the incoming signal waveform. (By "edges" I mean the transitions from the 0 state to the 1 state, and from the 1 state to the 0 state). And thereby affect the TIMING of the conversion of each sample to analog.

Signal reflections caused by impedance mismatches, as I explained and as the article describes, will propagate from the dac input circuit back to the transport output, and then partially re-reflect back to the dac input, where whatever fraction of the re-reflection that is not reflected once again will sum together with the original waveform.

If the cable length is such that the time required for that round trip results in the re-reflection returning to the dac input when the original waveform is at or near the mid-point of a transition between 0 and 1 or 1 and 0, since the receiver's logic threshold is likely to be somewhere around that mid-point the result will be increased jitter.

Again, no one is claiming that bits are not received by the dac with 100% accuracy. The claim is that the TIMING of the conversion of each sample to analog will randomly fluctuate. The degree of that fluctuation will be small, and will be a function of the many factors I mentioned (and no doubt others as well), but there seems to be wide acceptance across both the objectivist and the subjectivist constituents of the audiophile spectrum that jitter effects can be audibly significant.

If your friend disagrees with that, he should keep in mind two key facts, which he may not realize:

1)The S/PDIF and AES/EBU interfaces we are discussing convey both clock and data together, multiplexed (i.e., combined) into a single signal.

2)The timing of each of the 44,100 conversions that are performed each second by the dac is determined by the clock that is extracted from that interface signal.

Best regards,
-- Al

Paul, I believe the thread you are referring to is dealing with a one-box cd player (a Music Hall CD25.2).

Although it has a digital output and can be used as a transport in conjunction with a separate dac, presumably the discussion pertains to its analog outputs, which are generated by its own internal dac, and processed through its analog circuitry. Given that, as Mapman indicated, tonality and color can certainly be affected by the design and quality of the dac and the analog circuitry in the player.

Jitter is the predominant consideration just in the digital parts of the signal path, up to and including the dac chip. And it becomes a MUCH more critical consideration when the transport and dac are in separate components, because of the impedance matching, reflection, noise, clock recovery, and other interface-related issues that have been discussed above.

Best regards,
-- Al

In Jea48's linked article the implication was that the level of jitter was related to or at least different for different frequency levels of sound (presumably after the DAC). Someone straighten me out on this. It seems to me that the bit stream speed is independent of the bit content. If this is correct than should not the jitter be either constant of possible a function of the disc itself (like radial position or burn/pressing quality)?

Paul, you raise a good question, and I believe that the key to the answer is that jitter should be thought of as noise in the time domain.

As you will realize, an analog signal will always have some amount of noise riding on it, which causes its amplitude to fluctuate to some degree, in a manner which is to some extent random. That noise will typically consist of a great many frequency components, all mixed together. Essentially a mix of ALL frequencies within some finite bandwidth, with different frequencies having different magnitudes.

Similarly, the random or pseudo-random timing fluctuations that characterize jitter in a digital signal will have a spectrum of a great many jitter frequencies all mixed together. In other words, there may be slow fluctuations in the timing, that are of some magnitude, accompanied by rapid fluctuations in the timing, that are of other magnitudes.

Some frequency components of the jitter spectra can be data dependent, because a major contributor to the electrical noise that is a fundamental cause of jitter is the rapid transitions of transistors and integrated circuits between the 0 and 1 states, and vice versa.

BTW, re the references in your two posts to disk speed, radial position, etc., keep in mind that fluctuations and inaccuracies in the rotational speed of the disk (which figure to be far larger in magnitude than the electronic jitter we have been discussing) are, or at least should be, taken out by subsequent buffering in the transport's electronics.

Best regards,
-- Al