Prices. Choose optimal >
Order >
        AuI ConverteR 48x44 RSS Audiophile Inventory on Facebook Audiophile Inventory on Google+ Audiophile Inventory on LinkedIn Audiophile Inventory on Twitter     [Login/Register]

DSF File Format. What is DSF File Audio Extension [Article]


Audio Basis - educational articles

DSF file format (*.dsf file extension) is kind audio file with DSD (1-bit audio) content to store music and other sound stuff. Traditionally it is one of audiophile file formats.
Look below quality, conversion, playback issues, data structure, etc.

DSF file format

DSF file features >        DSF file quality issues (DSD vs. PCM) >       DSF vs. DFF >

How to edit DSF file >       DSF file converters >       DSF file players >      Metadata editor >




DSF file features

DSF file features

  • Sample rates: 2.8 (D64), 5.6 (D128), 11.2 (D256), 22.5 (D512), 45.2 (D1024), ... MHz (read more below)
  • Bit-depth: 1 bit / 8 bit (see below bit number discussion) [1]
  • Modulation type: sigma delta modulation (a.k.a. DSD)
  • Channel number: 1 ... 6 (see below)
  • Metadata (song, album names, year, performer, artwork, etc.): ID3v2 format


More details

Sample rate DSD calculated as 44100 Hz x K (coefficient: 64=2^6, 128=2^7, 256=2^8, 512=2^9, 1024=2^10, etc.)

Example: 44100 Hz * 64 = 2 822 400 Hz = 2.8 MHz

Also DSD sample rate may be based on 48000 Hz [2] (non-standard range). However the range can be implemented in DSF files too.


DSF audio file structure is:

  1. Header
  2. Audio data
  3. Metadata block

The audio data divided by blocks. Last block may be filled partitially. Wrong values into header can cause reading of unfilled place of last block. In some cases it can cause click at end of the DSF file.

Channel number is integer value. So channel number more 6 may be supported too. DSF have specified channel types and channel position in file structure from "mono", "stereo" to "5.1". However, the channel variable have reseved unspecified range for custom types. Of course, using custom types may cause issues with different software and hardware.



DSF file format quality issues (DSD vs. PCM)


If we want to compare quality DSD (sigma delta modulation) and PCM (pulse code modulation), we consider ear tests. As example, dsf vs flac or other. However, not all so simple.

First, mathematical models and hardware implementation should be separated.

Sound file is pure mathematical implementation of theoretical model of sigma delta modulation. As final criteria here quantization noise level (QNL) is considered. The level depend on frequency.

Hardware implementation distorted the mathematical model by non-linearity of real electronic components, their frequency characteristics, electrical hum, elecromagnetic interference.

Software implementations have different resampling, modulation and demodulation algorithms, that can cause different features and sound.


Read other articles about audio issues




BD - bit depth

NS - noise shaping

PCM - pulse code modulation

QE - quantization error

QNL - quantization noise level

SDM - sigma delta nodulation

SR - sample rate

UMB - useful music band



Bit depth (BD) and sample rate (SR)

DSD bit depth may have any positive integer value theoretically. Bith depth and sample rate may be considered in complex only. Both these values impact to quantizing noise level:

higher BD or SR => lower QNL.

Higher bit depth have lower quantization error (QE), and, therefore, QNL.

Higher sample rate provide wider signal band and the noise energy distributed in wider band. So QNL is reduced.

Quantization error and QNL are different things. QNL depend on QE. But for same QE, QNL may be different. At different frequencies, as example.


Noise shaping (NS)

To reduce the quantization error energy at useful music band (UMB), sigma delta modulator "pushes" the energy to high frequency range. It is . Band reserve shall be provided for NS. It reduce UMB into fixed SR/2.

There are ways of expanding signal band:

  • more steep the shaping;
  • more high sampling frequency value.

Steeper NS can cause modulator stability breakdown in overload.

Higher SR is larger file size, more computing resource and electrical energy consuming.

Can we reduce bit depth and keep quality?

Yes. To keep UMB level noise, noise shaping is applied. Quantization error energy from low frequency range "pushed" into high frequency area. However we lose part of high frequency band in full range [sample rate / 2].

NS can cause audible artefacts at some apparatus due non-linear distortions (intermodulations).



DSD QNL depend on sigma delta modulator and demodulator implementation.

Generally, DSD 64 have the higher QNL than DSD 128, DSD 256, DSD 512. Because DSD64 have lesser noise band reserve than higher sample rates. The band reserve placed at high frequency range. The range intended to pushing noise energy. Lesser sample rate - lesser band reserve - lesser noise energy may be pushed without delta sigma modulator's stability breakdown.

In this part the article we will consider noise level in useful music band. Above UMB noise-shaped level grow significantly.

In the table below trends bounding with DSF diskretization frequency increasing. It was important to remember, that the trends depend on modulator and demodulator implementation.

DSF sample rate trends
  Noise floor in useful musical band Useful musical band width
CD about -100...120 dB 20 kHz
DSD64 about PCM 24 bit about 20 … 22 kHz
DSD128 lower DSD64 wider DSD64
DSD256 about PCM 32 bit wider DSD256
DSD1024   about 100 kHz




From point of view mathematical model DSD and PCM have differences:

  • Noise level,
  • Transmitted band.

I.e. no difference between PCM and DSD, between kinds of pulce code and sigma delta modulations when they are considered as formats.
Because format is abstract mathematical model, that give potential abilities for implementation in the real life.

These features may be implemented differently as hardware and software.

It is developer's choise what increase to reduce noise: sample rate or bit depth.

I.e. instead BD increasing, designer can increase SR, and contrary. Or noise shaping can be apply instead.

Thus no univocal advantages of PCM or DSD as format (mathematical abstraction).
But implementations may have advantages.



How to edit DSF file

How to edit DSF

Audio data may be cutted by block without changing of binary content.

Editing of audio content is provided via conversion to pulce code modulation, content altering, converting to delta sigma modulation back.

It is lossy transformation. But, professional editing software have loses comparable with resampling.

There is ringing audio matter. However, the author still don't know about exact researches of the ringing impact to ear perception.

Advice: To minimize the loses, use conversion DSF to PCM and PCM to DSF once.

DSF file editing



To edit .DSF file format need convert DSD to PCM and back. Quality of conversion resembles resampling quality.





DSF file format's binary audio content may be exactly copied into DFF.

However DFF have no standard storing metadata way.

DFF consists of blocks. It allow to insert block with metadata. However, DFF software should recognize metadata block.

Warning: some software can convert DSF to DFF or back with modifications (non-bit-perfect mode). So checking of every converting tool is recommended.



DSF file converters

DSF converters

Distingushed by noise level, aliases, processings, service functionality and price.




DSF file players

Music players DSF


  • Foobar 2000 (Win)
  • Audirvana (Mac)
  • JRiver (Win, Mac)
  • Amarra (Win, Mac)


Metadata editor DSF

Metadata editor


AuI Converter 48x44 >

What Inside DSD Converter of Audio Files >



Read more
  1. How open DSF files audio >
  2. DSF Format Audio Files >
  3. DSF audio file format >
  4. DSF Audio File >
  5. File Extension DSF >
  6. DSF to PCM converter >
  7. DSF File Converter >
  8. DSF file converter audio >
  9. DSF oversampling. D64 vs. D128 >
  10. Audio Converter DSF to FLAC >
  11. How Impact to Audio Quality of PCM to DSF Conversion. 1-bit DSF vs. PCM >
  12. How oversample DSF audio file >





[1] DSF file format specification

[2] Prism Sound: DSD technology review


Read other articles about audio issues

Copyright © Yuri Korzunov [Audiophile Inventory (], 2010-2017. All Rights Reserved.

Site map      Terms and Conditions