The "best" way to convert DSF to FLAC?

Discussion in 'Computer Audio' started by mmcc, Mar 14, 2018.
  1. mmcc
    Hi all - is there an accepted "best" way to convert DSF files to FLAC?

    I understand that it is a lossy process, but I'm curious what the best options to choose are.

    I'm on Ubuntu Linux, and I compiled dsf2flac, which offers a bunch of helpful options:

    -r, --samplerate=Hz Output sample rate (possible values="88200",
    "176400", "352800" default=`88200')
    -b, --bits=bits Output bitdepth (possible values="16", "20",
    "24" default=`24')
    -n, --nodither Don't add dither before quantization (default=off)
    -s, --scale=dB Scale adjustment. Raw DSD has a modulation depth of
    approximately 0.5 so with no scaling the PCM peak
    level is approximately -6dB below 0dBFs
    (default=`4')
    -d, --dop Encode DSD data directly into FLAC file without
    conversion to PCM using DoP format (DSD over PCM)
    (default=off)

    What's the best sample rate? Bit depth? Dithering is ok?
     
    Last edited: Mar 15, 2018
  2. gimmeheadroom
    Hi,

    Usually ffmpeg is the swiss army knife, I rarely have to use anything else.

    ffmpeg -i yourfile.dsf newfile.flac

    It uses the file extensions (at least for output) to determine what format. And the defaults are usually good so you don't have to specify tons of options.

    BTW why are you doing this? There are several players for Linux that play DSD...
     
    Last edited: Mar 17, 2018
  3. varkokonyi
    Thanks, ffmpeg works really well. I wrote this to convert a whole folder of files

    for x in ./*; do ffmpeg -i "$x" "${x::-4}".flac; done

    The reason I am doing this (don't know about mmcc) is I like VLC, and I couldn't get the dsf files playing there.
     
  4. gimmeheadroom
    Yes, it's quite useful. I think find with -exec is more flexible that for, especially if you have embedded spaces or escape characters. I went through the process of converting a few hundred gig of APEs. I was too lazy to script the .cue file changes though.

    I'm to the point now where I'm so dissatisfied with Linux for media I bought my first winbloze box in more than 15 years just to use as a foobar appliance. I use linux utilities to process files, since it's much easier and more flexible, then I ftp everything to the windoze box to actually play it.
     
  5. varkokonyi
    Which Windows did you get? On the latest win10, you can install linux as a VM without any additional requirements. You get a Linux command prompt, the only catch is that there is no GUI, CLI only.
     
  6. gimmeheadroom
    It came with Win10 pro and it's actually not horrible. I like it much better than xp, win7 etc. I guess it's like Android (which I hate) on a "desktop" but I'm used to it already, so...

    I have a bunch of boxes running Linux and UNIX. I don't need another Linux box. I really don't like Linux either but I hate it less than Windows.

    Hi from CR btw, I see we are somewhat neighbors :)
     
  7. Roseval
    Inherent to DSD is a high amount of quantization noise.
    In case of DSD64 this starts at approx. 21 kHz
    In case of DSD128 at 42, etc.
    If you use DSD64 and a PCM sample rate of 88 there might be a high amount of quantization noise in the PCM
    Use a spectrum analyzer to check.
    If this is the case, look for a low pass filter to get rid of this noise or lower the PCM sample rate.
    http://www.thewelltemperedcomputer.com/KB/DSD.htm
     
  8. varkokonyi
    Headroom: Hi neighbour! What I mean that for example, the two systems share filesystems, so you don't have to FTP files back and forth

    Roseval: So converting it at a higher sample rate makes the noise worse? I don't have a scope, unfortunately.
     
  9. Roseval
  10. gimmeheadroom
    Ah yes, you are right. But I am tired of playing home sysadmin :frowning2: so for now I don't mind. I use the ftp server as a backup for the music files anyway.

    @Roseval I didn't do special testing but to me on my equipment DSD just sounds better than FLAC.
     

Share This Page