The PCM To DSD Conversion Project

DarqueKnight
DarqueKnight Posts: 6,760
edited November 2017 in Going Digital
Introduction

m0eyaguodq3m.jpg
Figure 1. The digital trinity: dCS Puccini U-Clock master cloce, Bryaton BDP-2 digital player, dCS Debussy DAC. The diskless life keeps getting better.

I recently completed a conversion of all PMC FLAC format music files to DSD (direct stream digital) format. DSD is the format used to encode SACD discs. My motivation for this project was not better sound quality, although that was an unexpected benefit. My motivation was convenience.

My DAC, a dCS Debussy, mutes the first few seconds of a song if it has a different sample rate than the previous song. The muting is caused by the DAC needing time to lock on to the new sample rate. I also have a dCS Puccini master clock, and it requires time to lock on to the new sample rate also. Therefore, I have two delays to contend with. When I questioned dCS about the issue, this was their response:

"Our Phase Locked Loop circuitry has a 3-stage lock which takes a few
seconds longer than simpler designs. This is a factor in the sound
quality, as you need a very narrow bandwidth in order to filter out
jitter. We are considering a final update to the Debussy to add some
higher sample rates, but I don't know if it has sufficient capacity
to reduce the locking time.

Note also that the U-Clock has to lock before the Debussy can start
to lock, so you have 2 delays in your system.

The USB interface is primarily intended for connection to a PC.
Streaming programs such as JRiver have the option to play a few
seconds of carrier after a rate change, to give the DAC time to lock
and unmute. It might be worth asking Bryston if they have the
capacity to add a delay."


My Bryston BDP-2 digital player does not have the capability of adding gaps between songs. The workaround I used was to make silent files of 2 seconds for every sample rate in my music collection. When making playlists, the appropriate "gap file" was inserted where needed. This was cumbersome, but effective. My playlists typically have a mix of FLAC files of different sample rates and DSD. The 2 second length was arrived at by trial and error.

In order to get away from that cumbersome workaround, I began investigating PCM to DSD conversion software. Most such software is aimed at the professional recording studio market. I evaluated three that are aimed at the home user:

1. JRiver Media Center ($50) - sample rate conversion and format conversion tools are part of this digital player's features.

2. Aul ConverteR ($250).

3. TASCAM HiRes Editor (Free).

The TASCAM converter required a conversion from FLAC to wav prior to converting to DSD. The resulting FLAC files had diminished detail, clarity, and spatial properties. This was also true of the Aul converter, but to a lesser extent. Aul could convert directly from FLAC to DSD. JRiver produced the best sounding conversions, with DSD files that had improved clarity, detail, and spatial rendering than the original FLAC file. I already owned JRiver, therefore no additional investment was required in software. I did invest low 4 figures in new hard drives and hard drive enclosures. JRiver's PCM-to-DSD performance was commendable when compared to a $2,000 professional recording studio package:

http://archimago.blogspot.com/2014/04/analysis-comparison-of-dsd-encoders.html

I was surprised, and disappointed, to find that the software I used to rip my vinyl records to DSD, VinylStudio, did not have the capability of converting PCM to DSD. PCM files could be imported into the software and changed to a different sample rate and/or a different format (e.g. FLAC to wav), and DSD files could be imported and converted to PCM, but PCM files could not be converted to DSD. When I asked the software maker's reason for not including a PCM to DSD feature, they replied:

"...it would not gain you anything anyway. You would get:
an increase in file size
no increase in quality
a file you cannot edit

VinylStudio's support for DSD is to allow you to record from devices that support it, and, if desired, to convert the resulting file to PCM."


PS Audio, the maker of the NuWave Phono Converter that I used to convert my vinyl records to DSD format, recommended a maximum sample rate of 96kHz, but they included higher PCM and DSD sample rates because they knew there was a demand for it. This quote is from the NuWave manual:

"Our recommended maximum sample rate for all vinyl phono or Analog inputs that do not exceed a usable input bandwidth of 48kHz, is 96kHz. As there is no usable information on a vinyl LP above30kHz, it will always sound preferable to restrict the maximum output sample rate to 96kHz."

At the beginning of my vinyl ripping project, I experimented with 96k, 176.4k, 192k, and DSD64, and I obtained the most natural and detailed sound with the best spatial properties as I went up in sampling rate.

Converted File Sizes

My mixed-sample-rate digital music collection currently took up 62% of a 2 TB hard drive. The all-DSD collection takes u[ of half of a 6 TB hard drive. Average differences in converted file sizes with respect to sample rate were as follows:

1. 44.1k/16 bit FLAC files produced DSD64 files 4 times larger.
2. 88.2k/24 bit FLAC files produced DSD64 files 2.11 times larger.
3. 96k/24 bit FLAC files produced DSD64 files 1.89 times.
4. 176.4k/24 bit FLAC files produced DSD64 files 0.7 times smaller.
5. 192k/24 bit FLAC files produced DSD64 files 1.18 times larger.

The bit rates of PCM and DSD are not directly comparable. While it is true that a higher PCM sampling rate can capture more information than a lower PCM rate, you can't say that a two chanel PCM sampling rate of 192k/24 bits (9.216 megabits-per-second) can capture more information than a two channel DSD64 sampling rate of 5.6448 megabits-per-second. DSD samples are not the same as PCM samples. PCM encodes information by varying the amplitude of the sample in proportion to the amplitude of the analog signal. DSD encodes information by varying the energy density of the sample in proportion to the amplitude of the analog signal, while maintaining a constant amplitude for each sample. In this way, comparing a PCM sample is somewhat analogous to comparing a two dimensional picture of an object to the actual three dimensional object.

When I began research for this project, I was not aware that there was a rancorous "PCM vs. DSD" debate similar to other audiophile/anti-audiophile debates such as "tubes vs. transistors", "sighted vs. blind testing", and "stereo vs. mono".

Proud and loyal citizen of the Digital Domain and Solid State Country!
Post edited by [Deleted User] on

Comments

  • The Sound

    Converting my FLAC files did not turn frogs into princes. It turned princes into kings. Frogs still croaked like frogs, while well-recorded songs assumed improved clarity, detail, weight, and spatiality. My Debussy DAC upsamples all PCM sources which are 44.1k or multiples of 44.1k to 2.8224 MSps/5 bit PCM. PCM sources which are 48k or multiples of 48k are upsampled to 3.072 MSps/5 bit PCM. I asked dCS if the improved sound quality I was hearing with converted FLAC files was due to the DAC being relieved of the upsampling process. Here is their reply:

    "Using a separate Upsampler relieves the DAC of some of the oversampling
    burden and results in better sound quality as you have heard for
    yourself. This is why we have sold large numbers of Upsamplers over the
    years. The reason this makes a difference is not clear, we have no
    other convincing explanation for the effect.

    Upsampling to DSD generally has the maximum benefit. We think that the
    very large amount of out-of-band noise inherent in DSD data streams
    heavily dithers the DAC, resulting a smoother and more rounded
    performance. These days, the upsampling can be done by a PC instead of
    separate hardware."


    Dithering is the process of adding high frequency white noise to a digital signal in order to "fill in" the spaces between samples and "fill in" the spaces between the actual analog value and the rounded off digital value. While it's true that the human ear can't "hear" ultra-sonic noise, the effects of ultra-sonic noise can be sensed by the ear, just as infra-sonic ( below 20 Hz) sound can't be "heard", but can be sensed (felt) consciously and subconsciously by the ear and body.

    Associated Equipment

    Bryston BDP-2 digital player
    dCS Debussy DAC
    dCS Puccini U-Clock word clock
    PS Audio PowerBase isolation platforms for BDP-2, Debussy DAC and Puccini U-Clock
    Black Diamond Racing isolation Pits and Mk IV Cones
    Pass Labs XP-30 line level preamplifier
    Pass Labs X600.5 monoblock power amplifiers
    Revelation Audio Labs Prophecy Cryo-Silver Split Configuration USB cable
    Revelation Audio Labs Prophecy Cryo-Silver AES coaxial cable
    Revelation Audio Labs Passage Cryo-Silver DB-25 power umbilicals for XP30
    AudioQuest Sky XLR interconnects
    AudioQuest Everest speaker cables
    PS Audio PerfectWave AC-12 power cords
    PS Audio PerfectWave P-10 AC Regenerator
    Polk Audio SDA SRS 1.2TL loudspeakers (heavily modified)
    Salamander Synergy Triple 30 audio credenza

    Proud and loyal citizen of the Digital Domain and Solid State Country!
  • headrott
    headrott Posts: 5,484
    Very nice and very interesting write up Raife. Not sure how I missed this thread when you posted it.

    Of course, some "audiophile-bashers" may use this statement:

    "Using a separate Upsampler relieves the DAC of some of the oversampling
    burden and results in better sound quality as you have heard for
    yourself. This is why we have sold large numbers of Upsamplers over the
    years. The reason this makes a difference is not clear, we have no
    other convincing explanation for the effect........"


    ;to say that it shows how much "hooey" we audiophiles are full of because even the Upsampler manufacturer (dCS) could not scientifically explain why the separate upsampling equipment used could cause the improved sound quality.

    I do agree with and like your statement regarding dithering of high and low frequencies:

    "Dithering is the process of adding high frequency white noise to a digital signal in order to "fill in" the spaces between samples and "fill in" the spaces between the actual analog value and the rounded off digital value. While it's true that the human ear can't "hear" ultra-sonic noise, the effects of ultra-sonic noise can be sensed by the ear, just as infra-sonic ( below 20 Hz) sound can't be "heard", but can be sensed (felt) consciously and subconsciously by the ear and body."
    Relayer-Big-O-Poster.jpg
    Taken from a recent Audioholics reply regarding "Club Polk" and Polk speakers:
    "I'm yet to hear a Polk speaker that merits more than a sentence and 60 seconds discussion." :\
    My response is: If you need 60 seconds to respond in one sentence, you probably should't be evaluating Polk speakers.....


    "Green leaves reveal the heart spoken Khatru"- Jon Anderson

    "Have A Little Faith! And Everything You'll Face, Will Jump From Out Right On Into Place! Yeah! Take A Little Time! And Everything You'll Find, Will Move From Gloom Right On Into Shine!"- Arthur Lee