Loading...

Formats

Digital Audio Formats

Why bigger numbers and popular beliefs do not settle the question

Digital audio is complex, and digital audio design is as nuanced and full of trade-offs as designing any other part of the stereo system chain. That complexity often leads people to seek and adopt simplifying beliefs.

  • More bits must be better.
  • Higher sample rates must be better.
  • DSD must be better than PCM.
  • Native DSD must be better than DoP.
  • Upsampling must be an upgrade.

These ideas survive because each contains some truth. But none of them stands up well under scrutiny.

Modern digital audio is not defined by simple hierarchies. Once bit-depth and sample rate are adequate for the job, the result depends far more on how the whole system is designed and implemented.

Encoding, transport, and DSD carriage are different things

It helps to separate three different technology layers.

  • USB, S/PDIF, AES3 and I²S are transport methods.
  • PCM and DSD are ways of encoding audio as data.
  • DoP and Native DSD are ways of framing and signalling DSD over a transport method.

These distinctions matter.

PCM and DSD describe the form of the audio data itself. USB, S/PDIF, AES3 and I²S describe how that data travels from one device to another. If the data is DSD, it still has to be carried somehow. DoP and Native DSD describe two different ways of doing that, but when implemented correctly they deliver the same underlying DSD payload to the DAC.

Many arguments become confused because they treat these three distinct issues as though they were one. That is understandable because analogue transmission is conceptually simpler. Digital audio transmission is multi-layered, and each layer performs a different role.

PCM and DSD are different trade-offs

PCM and DSD are often discussed as though one must be inherently superior. In reality, they are different ways of encoding audio, and each has strengths and compromises.

PCM represents the waveform as multi-bit samples taken at regular intervals.

DSD represents it as a 1-bit stream running at a very high rate, with noise shaped away from the audible band.

DSD can reduce reliance on the steep anti-alias filtering associated with lower-rate PCM. But DSD pushes more quantisation noise into ultrasonic frequencies, and that too must be managed.

Neither is a free pass. Neither avoids filtering or quantisation trade-offs. Neither guarantees a better result. Neither abolishes the need for design skill and implementation quality.

For a time, DSD was marketed as though it had solved the problems of PCM. There were commercial drivers for that, not just a technical argument. That created a one-sided story which has lingered long after those commercial goals faded. Much of the early disappointment with PCM still lingers in audiophile thinking, yet those early problems were due to how PCM was being implemented, not to any fatal flaw in PCM. Digital engineering has moved on.

Preferences still vary, but with today’s better digital design the differences are often more about presentation and implementation than about one format being inherently better.

DoP and Native DSD carry the same DSD data

Once audio exists as DSD, it still has to be transported to the DAC.

DoP carries DSD inside PCM-style frames so it can travel over established PCM-based transport conventions, most commonly USB Audio, and in some cases S/PDIF and AES3. Interoperability becomes much more likely if both products adhere properly to the same standard.

Native DSD typically relies on transport-specific or manufacturer-specific implementations instead of the DoP convention, to achieve what is claimed to be a more direct transmission method. The usual arguments are:

  • the stream is declared as DSD from the outset
  • DSD data and signalling are handled more explicitly
  • it avoids the PCM-style wrapper used by DoP
  • it may allow support for higher DSD rates in systems designed for it

These are fair points, but Native DSD depends much more on how each manufacturer has chosen to implement support. And there lies the problem. Native DSD may work very well, or not at all with your equipment. Manufacturers do not all implement Native DSD support in the same way across products and platforms. That makes broad claims about Native DSD capability more impressive on paper than they often are in practice.

DoP’s strength is clear. It delivers the same DSD data to the DAC as any other competent method, but does so within long-accepted and widely used standards. If both products adhere properly to those standards, interoperability can reasonably be expected. Its DSD rate limits are simply the limits of the transport standards it follows.

Bigger numbers do not settle the question

It is true that:

  • Bit-depth matters.
  • Sample rate matters.

But once the rates are adequate, and the data reaches the DAC unchanged, the quality of the result depends much less on the numbers than on how the design behaves. It is also wrong to assume that higher rates cannot have downsides. Higher rates can stress a design and increase high-frequency noise.

Results will vary with the equipment used, and perceived differences are often influenced by the filtering choices involved. Higher sample-rate material is often handled with different filtering choices, and those can change the balance between smoothness, edge definition, and temporal character. Different listeners in different systems hear that trade-off differently, and opinions remain split on very high-rate files.

Upsampling is often misunderstood

Upsampling does not make a file more accurate. It changes where the reconstruction work is done, and it often changes the filters used.

DACs reconstruct a continuous waveform from discrete digital data. So some form of curve-fitting, filtering, or interpolation is always part of the process. When you upsample before the DAC, you are deciding that part of this work should be done before the DAC, using different maths.

It is a little like video scaling. Sending video already scaled to suit a display can help if the external scaling is better than the display’s own scaling, or if the display works best when fed a certain signal. But it only helps if that is actually how the particular display behaves.

Upsampling before the DAC is similar. But DAC manufacturers rarely tell you what their true preferred internal operating condition is, and that should tell you something. Whether upsampling helps depends profoundly on what the DAC itself does with the result, so there are no universal truths.

Formats and numbers can be a distraction

The things that often matter more to the final result are:

  • filtering
  • conversion method
  • clocking
  • noise behaviour
  • power integrity
  • analogue stage design

These things are not fully captured by headline specifications, and not all of them are easy to describe with the measurements most commonly quoted in marketing. So it is easy to be distracted by formats, bitrates, and claims that go well beyond what industry standards can honestly support. Digital audio design is full of nuance and trade-offs, not numbers to be maximised.

Standards Matter & Implementations Matter

It takes two to tango. A DAC and a DAS can only be expected to work well together when both are designed to the same standard.

Much attention is given to the headline figures associated with digital interfaces — such as 24/192 for S/PDIF and AES3, or 32/384 for USB — but standards are primarily about compatibility. They describe how formats are supported and transported. They do not, by themselves, promise one fixed upper limit or one best-sounding result.

The actual limits depend on the particular source and DAC. And even when a higher sample rate is supported and plays correctly, that does not mean it will sound best. In practice, operation near the edge of compatibility may be less convincing than a lower rate implemented more cleanly. Do not assume from the specification alone. Trust how the music feels to you.

 ▲