Wrong. Your DAC sucks. Read on.

Now, everyone knows different digital sources sound different when used with the same DAC. Even using the same source, some claim sonic differences depending on the digital cable used, paint their CDs green or trim the edges.

I don't think these people are hallucinating, however there is no need for mystical explanations.

You change something, and the sound changes.
The DAC is the same.
How was the DAC influenced ?
Let's make a list of all the things that could do that :

• Data bits, of course.
• Noise injection
• Clock jitter

Bit errors

Testing for this is simple : burn a CD with a test waveform or music, then play it and record the SPDIF sent by the CD player using a good soundcard with digital input and compare the files on the computer. Of course you have to compensate for the time delay between the two.

Surprise : most CD players can read CDs. Unless there is a nasty scratch on the CD, there will be no bit errors.

Some CD players will corrupt the stream however, due to brain-dead processing that it is not possible to deactivate, like digital volume controls. I have an example right here. The easiest way to solve the problem is to test your CD player or transport for bit-accuracy, and once assured it can indeed read CDs, this is one less worry.

So, to sum up : if you have a transport that delivers the wrong bits, it is defective ; sell it on e-bay. If your transport delivers the right bits, the sonics being not what you expected don't obviously come from the bits being wrong, so stop polluting the forums about the effect of green paint on bit error rate when there were no bit errors to start with.

Noise injection

A CD player contains a lot of nasty things : digital hi-frequency circuits, PWM-controlled tracking servos, generally a small computer, multiplexed fluorescent or LED display, and most of the time a really crappy power supply and little if no shielding at all.

It is conceivable (although EMC regulations fight against this) that the CD player could emit nasty interference through the power cable and through the air and thus pollute the rest of the system, including the DAC. I don't think it is very important though.

It is pretty inevitable, though, unless serious measures are taken, that the master clock and other very sensitive analog parts inside the player will happily "process" all that junk noise, which will compromise the sonics.

Remember that the power draw in a CD player is also modulated by the tracking servos, which closely follows the movement of the CD, which resonates with the music you're playing.

Jitter

Most CD players and transport's SPDIF digital output will contain a rather large amount of jitter (up to a few nanoseconds in the most desperate cases).

Jitter is phase noise in the clock of a signal, ie. the signal edges don't appear when they should, but rather a bit earlier or a bit later.
Really massive amounts of jitter (>100 nanoseconds) can make a bit arrive too late, so out of sync that the receiver chip thinks it's the next bit and gets confused, resulting in a bit error. However this never happens in real life because even the cheapest CD player is designed so that its digital output "works" ; and it can be easily diagnosed by recording the output of the CD player with a digital soundcard and comparing with the actual data on the CD.

However tests seem to indicate that jitter above 20 ps is audible. 100 nanoseconds is 5000 times larger than that, so there is room between actually working and sounding good.

Phase noise is not a mystical sort of noise, it is simply regular noise added to the clock signal. As this signal has got a finite slew rate, adding noise to it will shift the edges. Hence, jitter.

Effects of jitter

A DAC chip has two inputs :

• Binary sample data input (digital)
  These are the bits stored on the CD. If the system functions correctly, the DAC chip will see the correct data bits. No arguing, please.
• Clock input (analog)
  This is when problems start, because the clock is an analog signal. The information carried by the clock is carried by the instant at which its voltage crosses a certain threshold. Therefore, if noise is added to the clock signal, these instants shift and you have jitter.

Knowing the bits are right, the only thing that can make the DAC chip sound different is the jitter in the clock. I won't speak about the output stage or DAC power supply noise here, let's just concentrate on the incoming clock. Where does it come from ?

Sources of jitter

Noise added to the clock signal

The master clock is generally located in the transport in the form of a crystal oscillator.
Picture a tiny quartz crystal, cut to the right size and oscillating (moving) like a bell at a certain frequency. Motion is converted to and from electricity via piezo-electric effect. The crystal is controlled by an electrical circuit in a tight feedback loop, making it an oscillator. Like a bell, it has a specific frequency it likes to vibrate at (a high Q resonance peak), which makes it quite precise.

However, its mechanical nature makes it sensitive to vibrations.
The external circuit can be either cheap or low noise, not both.

As a result, a crystal oscillator will be disturbed by vibrations, power supply noise, electromagnetic interference, and the inherent noise of the oscillator circuit. All these will create jitter in the master clock right from the start. Check inside your CD player to see what countermeasures were taken (hint : none).

A good starting point is a Tent Clock which incorporates a low-noise circuit, good crystal, and is shielded against interference. It only needs a proper power supply and vibration control. These are still far from obvious...

In its path from its source (crystal oscillator) and the DAC chip, the clock signal will pick up noise (jitter) at every corner. The more chips it goes through, the more jitter it will accumulate. So, it is simple to understand that a short clock path (a few cm. with 1-2 chips in the way) will be a lot less problematic to clean that a path including a CD player ASIC, spdif cable, CS8412 and various logic ICs.

In this sense, integrated CDPs have a lot more potential that the transport + DAC combo.

Noisy thresholds

The clock is an analog signal, but the important time information it carries can only be extracted by comparing it with a threshold voltage. This is done at the input pins of any digital chip it goes through. However, if the threshold is noisy and moves all the time, it is the same as if the noise was added directly to the clock. In digital ICs, the comparison threshold is directly derived from the power supply voltage. Therefore, the power supply voltage noise and ground noise of any IC on the critical clock path is as jitter-sensitive as the clock itself. This includes the DAC input pins. Another reason to minimize the number of chips on the clock path, and not to use ICs processing the clock for other tasks (for instance, the Marantz CD63SE uses the same buffer IC to process its clock AND data, adding jitter).

Why the DAC chips don't use differential clock signalling, I don't know. It would certainly suppress a problem.

SPDIF

The coding used in SPDIF transmission has an advantage : it transmits the clock and data on only one shielded cable. However, the two are so well mixed that, combined with the low pass filtering courtesy of the digital cable, the clock cannot be fully recovered without the random audio data signal contaminating it. This contamination depends on the filtering characteristics of the cable and associated connectors and circuits. This neatly explains the sound of digital cables without resorting to snake oil.

Read this AES article which analyzes SPDIF in-depth.

False problems

From this, we can explain a lot of things with real reasons and without resorting to snake oil.

• The sound of different transports depends on their jitter profile (how much, at which frequencies, etc).
• The jitter created by the transport comes from a lot of sources.
  Some of them are pretty straightforward, like incompetent layout or cheap components, others are pretty obscure, like power supply modulation caused by vibration compensation done by the tracking servo. I see no reason why the average (sloppily designed) transport would not be sensitive to the brand of CD-R used, for instance. People hear a difference in the sonics of computers used as SPDIF sources according to CPU load and the layout of files on disk ; this shows that noise and supply problems are everywhere.
• The sound of digital cables has a straightforward explanation.

This sucks !

You bet !

However, in my opinion the job of a DAC is to reproduce the music encoded in the digital audio data and to reject jitter. The job of a DAC is not to allow you to hear the jitter, but to allow you NOT to hear it. Therefore, I view any DAC which does not include successfull jitter reduction measures as worse than useless.

Therefore, a DAC which shows more sonic differences between transports or cables delivering the same correct bits from the CD but different jitter, is not "more transparent", but rather sucks more !

So far the only person I know who understood the problem and found a solution is Guido Tent with his famous DAC, kudos to him and his team. There are probably others, but they are not many.

So, the best solution is to eliminate the problem entirely and not use SPDIF :

• Use I2S (but you still have all the noise problems)
• Put the clock in the DAC and slave the transport (been there, done that, it's easy, it works, it sounds good)
• Ditch SPDIF entirely and use a transmission protocol designed from the start thinking about jitter problems.