DSD For Dummies

What Makes DSD Audio (Potentially) Unique. A DSD Primer. Or DSD For Dummies.

  • Super(?) High Sample Rate.
  • Delta Sigma Modulation.
  • 1-Bit Samples.
  • Noise Shaping.
  • Low Quantization Noise.
  • Low Distortion.

DSD, Direct Stream Digital is a file format trademarked by Sony and Philips. The format was initially created for archiving audio signals and then commercially used for SACDs. At the time of this writing, DSD audio files are available as media downloads without needing to use the physical media.

Super High Sample Rates.

DSD 64 is called so because the sample rate or sample frequency is 64 times higher than the CD sample rate of 44.1KHz (44,100) samples per second.

64 x 44,100 = 2,822,400. Again, samples per second.

DSD 128

128 x 44,100 = 5,644,800 samples per second.

And so on.

Delta Sigma Modulation (Sigma Delta Modulation. Same. Same. Mostly).

Delta-sigma modulation, simply put, is a method for encoding signals into digital signals. It is just a different method of encoding than PCM (Pulse Code Modulation) is.

What delta-sigma modulation does differently is that it does so at a higher sample rate than PCM does, but with the higher sample rates, it only needs 1-Bit samples.

The delta-sigma modulation process uses a negative feedback loop to correct for the quantization errors that are inherent in converting an analog signal to a digital signal.

About the “errors”: Think pushing a wheelbarrow up or down a flight of stairs instead of up or down a ramp. The vertical distance between the steps (digital signal) is missing the slope (analog signal). No matter how low of a height difference there is between the steps, there will always be a ‘bump’. Bumps lead to errors which lead to noise. There is a mathematical remainder (error, noise) between the digital steps that the analog slope provided before the A to D conversion.

Delta means difference. Sigma means Sum.

1-Bit Samples.

Due to the high frequency sampling and the DSM process of quantization, with the negative feedback loop, only 1 (ish) bit is needed. Basically, the single bit indicates if the current sample voltage is higher or lower than the sample taken before it. When the bits indicating up or down are mapped out, a wave, like an analog wave, is the result. Hence the need for only 1 (ish) bit.

Noise Shaping.

The errors that are there from the conversion of an analog to a digital signal are moved up and out of the way. The noise that is the result of the errors created by quantization is still there but noise shaping places the noise in bandwidths outside of the desired frequency range, the frequency range of the original signal. When the file is accessed, a filter blocks out the noise that now resides in the higher bandwidths. The result is a higher Signal-to-noise ratio.

Essentially, Noise Shaping, sweeps the noise that comes from the quantization errors under the rug so that we do not see them. Or, as with audio files, we do not hear them.

Low Quantization Noise.

Noise Shaping provides this. Again, the noise is still there, it is just in places that are more out of the perceivable way. So it sounds like less noise is in the playback of the file.

Low Distortion.

For audio files (and audiophiles) distortion of an audio signal is usually very unwanted. Distortion occurs when the wave of an output file does not lined up like it is directly on top of the source file. When analog signals are converted to digital signals, there will be distortion – steps vs slope.

And again, Noise Shaping and, perhaps, more so, the high sample rate are our audiophile friends so that a file can be played back with low distortion.

Putting It All Together.

All the above make for a clean natural sounding audio file. If the original analog source file is good, the converted DSD file should also be good.

Yes, as in most things, if garbage in, then garbage out. So, DSD is not a magic bullet solution for all genres of music or for specific recordings.

So, What Makes DSD Audio Files Special?

In a nutshell:

The High Sample Rates.

Delta Sigma Modulation

Noise Shaping.

Without these you cannot achieve the resulting:

  • Low Quantization Noise.
  • Low Distortion.

Adding What @dvorak So Eloquently Wrote To Sum It Up (To Have It All In One Place):

Further Information.

As this is not all encompassing, here are some links to further information:

A Deeper Dive (Thanks to @tedsmith and Richard Murison)

A Few Videos:

Ted explains DSD - Paul McGowan, PS Audio

PCM vs DSD - Paul McGowan, PS Audio

DSD Explained part 1 - The Hans Beekhuyzen Channel

And For Advising Or Joining In The Conversation That Started It All:

With Thanks To:


If any corrections or ideas need to be changed or edited, please reply and I will be happy to review them.
mrcandude :sunglasses:

This is awesome @mrcandude!! Thanks!!!

1 Like

A very solid effort, and I love the spirit of sharing :+1:

Let me highlight that the specific magic of DSD is its 1-bit-ness, and the way that helps is that it makes it much easier to implement a DAC circuit with extremely high linearity (which means low noise and distortion).

It’s also the case that to make 1-bit-ness deliver useful signal-to-noise ratios in the audible frequencies, it is necessary to sample at extremely high rates and use sigma-delta modulation to shift the quantisation noise way up into the ultrasonic region.

And as a consequence of the high sampling rate and noise shaping, we can avoid the CD-era mistake of using “brick wall” low pass filters which contribute greatly to what people have come to think of as the “sound of digital”.

So with high linearity, low noise, low distortion and the lack of “digital-sounding” filters, DSD raises the bar on how good digital audio can be.


Another data point: Richard Murison wrote this for Copper a while back. He took flak from all quarters for it, but it is correct:

DSD – Is It PCM, Or Isn’t It? – PS Audio

I can also recommend all of his Quibbles and Bits column essays.


That article is what it sounds like when an actual expert says the thing I’ve been clumsily trying to get across! Love it.

Anybody trying to get their heads around DSD please read what Ted just linked to and then feel free to follow up with questions.

1 Like

An excellent read


Thank you! Kind words indeed!
I figured if I had these questions, there has to be others with the same or similar, as well…

And I love your “Magic” summary. I added it to the original post.
Thank you, for that too!

Thank you for this!
I added to the links in the original post.

This one came up in my feed.
From February 27, 2014. Over ten years ago!
The origins of Ted ? :smiley:

This is the board they were talking about:


Gus said something about briefcase size, yes?
What are the dimensions?

Board statistics

Number of copper layers: 6
Board outline(s) extent: X = 455 MM; Y = 330 MM (18" x 13")
Number of parts: 1471
Number of pins: 5284 (258 through-hole, 5026 SMT)
Number of vias: 3205
Number of holes: 3463

Three boards cost: approx $4900 parts, (which includes $1100 in audio transformers and $1200 for the custom VCXOs) and $4800 for boards and assembly.

[Edit: I had the wrong report the first time.]


Here are some pictures of our meeting with Gus:

Paul and his wife were going out for date night so he came in the next morning:


First pic you get a good sense of the size of the board. Whoa!