Sinc Filters

Discussion in 'Computer Audiophile: Software, Configs, Tools' started by soumya, Mar 10, 2022.

?

Which filter did you like

  1. A

    0 vote(s)
    0.0%
  2. B

    100.0%
  1. soumya

    soumya Acquaintance

    Joined:
    Jun 24, 2018
    Likes Received:
    42
    Trophy Points:
    18
    Location:
    Mordor, Middle Earth
    While I am done hardening the coefficients for my sinc filters, I need some feedback if my work is in intended direction or not.
    To get a measure of it, I am creating a poll.

    There are 3 tracks which have been up-sampled 8x from 44.1 kHz source to 352.8 kHz.
    One of them is the Time Domain Optimized Sinc filter that I have been working on. And the other is HqPlayer Sinc-L

    I request community members to participate in this poll - download the wav files, play them on any 384 kHz capable DAC and try A/B-ing them. Please ensure that there is no other DSP in the path. Just feed the PCM files straight in to your DAC.

    Here 's the link
    https://drive.google.com/drive/folders/1ozDVkPN4vSt3nNj5j3Z0wZLIXkfN8J8g?usp=sharing
     
    • Like Like x 3
    • Epic Epic x 1
    • List
  2. Woland

    Woland Friend

    Pyrate
    Joined:
    Jan 13, 2021
    Likes Received:
    1,311
    Trophy Points:
    93
    Location:
    a friendly land
    This is so awesome! There's a lot of focus on this site on spending to upgrade gear, while much more attention is deserved on the enhancements that can be achieved at little or no cost and hassle with DSP. Thanks for the demo, @soumya ! I'm in camp B.

    Would you be able to provide 192kHz versions so friends with Schiit DACs and other bitrate limited DACs can participate?

    Is it possible to get the source files to compare both to no-upsampling?

    Can you plot the raw data (eg in Audacity) to show the differences? I did this but the file times don't quite line up, and I'm not sure what sample would show the differences most clearly.
     
    Last edited: Mar 10, 2022
  3. soumya

    soumya Acquaintance

    Joined:
    Jun 24, 2018
    Likes Received:
    42
    Trophy Points:
    18
    Location:
    Mordor, Middle Earth
    Hey there,
    Thanks for participating !

    I uploaded the source files. Usual disclaimer - no copyright infringement intended. Purely for academic purposes.

    While I will eventually create filters for 4x upsampling ratios, but I am curious how much will it be beneficial to users of Schiit DACs since even at those rates the built in Closed Form filter is still in effect and can't be defeated. For NOS R2R DAC users, yeah, I totally get it.

    Uhm that's tricky since the HQP filter output is captured out of a loopback adapter where as my filter has been run through a combination of SoX + camilladsp. So , as you observed the samples won't be aligned.
     
  4. Woland

    Woland Friend

    Pyrate
    Joined:
    Jan 13, 2021
    Likes Received:
    1,311
    Trophy Points:
    93
    Location:
    a friendly land
    a-b offset is 190108

    Here's a graph of the region (1ms) with the biggest differences.. they're not big.


    ab comparison.jpg .
     
  5. Ksaurav402

    Ksaurav402 Friend

    Pyrate Contributor
    Joined:
    May 7, 2019
    Likes Received:
    2,062
    Trophy Points:
    93
    Location:
    NY
    At 352.8 wouldn't some DS DAC still upsample it? I know Schiit BF2 does 8x but it only takes 192khz so difficult to bypass Schiit filter unless someone has Yggdrasil MIL which does 4x but with Metrum and Sonnet if you can provide 192Khz then I can listen to vanilla flavour with no sprinkles.
     
  6. soumya

    soumya Acquaintance

    Joined:
    Jun 24, 2018
    Likes Received:
    42
    Trophy Points:
    18
    Location:
    Mordor, Middle Earth
    All DS DACs upsample to MHz region before the data is fed in to the modulator.
    However, this is done in 2 stages -
    1. Digital FIR based 8x interpolation
    2. Digital IIR or simple Zero Order Hold

    The 1st stage is what's responsible for actual digital low pass filtering and hence has the biggest effect on how the DS DAC will sound (after the DS modulator of course).
    The 2nd stage does no interpolation (low pass filtering) ; merely increases bandwidth so as to make room for noise shaping by modulator.
    Now AKM , Cirrus Logic et al seem to use ZoH. ESS , whereas, applies an analog style IIR filter which is still very gentle (as typical analog Butterworth filters are). So no time domain distortion.

    I will upload the 176.4 kHz files for you to test later today.
     

Share This Page