ECP Audio Walnut X DAC technical measurements

Discussion in 'Source Measurements' started by atomicbob, Jan 25, 2022.

  1. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    16,409
    Trophy Points:
    113
    Location:
    On planet
    ECP Audio Walnut X.3 technical measurements

    Standard Prologue
    If you are unfamiliar with audio measurements please use a search engine with the query:
    "audio measurements" or "audio measurement handbook"
    Look for publications by Richard C. Cabot and also by Bob Metzler, both from Audio Precision. There are other useful publications as well. These will provide basic knowledge.
    Interpretation of the following measurements is beyond the scope of technical measurements posts.

    The data presented were collected as follows:

    1. PrismSound dScope III, picoscope 5243B
    2. Single Ended cables Mogami 2964 1 meter with Amphenol RCA connectors
    3. 100 Kohm load used for measurements
    4. dScope analyzer sample rate 48 KHz unless otherwise noted
    5. DAC 44.1 KHz sample rate, 24 bit depth unless otherwise noted
    6. Audioquest Forest and Schiit Pyst USB cables
    7. Vaunix Lab Brick USB hub
    8. Shielded 14AWG and 16AWG power cables

    Measurements are made in accordance with AES17:2015

    Measurements for record commenced after warmup of at least 48 hours.
    Measurements were performed over a period of several days.

    Index
    Post 1 - measurement setup description, highlights
    Post 2 - USB ASIO input - SE RCA outputs
    Post 3 - Filter response 20 Hz square wave
    Post 4 - reserved for additional data and corrections

    Notable highlights:

    The following exceptional measurements are usually associated with DAC balanced outputs. Walnut X proves a well designed SE output can produce similarly excellent results, some even superior.

    1. SE output Dynamic Range of 130 dB !!!
    2. SE output Cross-talk is essentially dual mono with 120 dB isolation !!!
    3. Power supply noise components below -155 dBFS !!!
    4. Low residual noise spectrum below -116 dBFS
    5. Low distortion spectrum with 4+HD+N below -101 dBFS
    6. Exceptionally low jitter with USB input

    An extraordinary WM8741 based design implementation by @dsavitsk and @TomB

    A big thank-you to @dasman66 for providing this Walnut X for measurement on the personal loaner tour.

    Listening setup:
    DSC_0787_small.jpg

    Measurement setup:
    DSC_0812_small.jpg

    Walnut X has received 40+ listening hours. The Walnut X is an excellent DAC and has great synergy with T4 headphone amplifier. Only a few Walnut X were produced, and even fewer T4; this combination is exceedingly rare. If fortunate enough to experience the pair the auditory reward will be worth the effort.

    Oddities:
    Amanero Technologies USB driver wouldn't work with ASIO4ALL on first day of measurements. Subsequent days it worked fine.
    When WASAPI used instead of ASIO, 0dBFS generator measurements were adjusted to -0.2 dBFS instead to avoid Windows audio system limiter.

    Click on any image in this report to enlarge the image and see fine details
     
    • Epic Epic x 9
    • Like Like x 6
    • List
    Last edited: Feb 3, 2022
  2. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    16,409
    Trophy Points:
    113
    Location:
    On planet
    USB ASIO input SE output measurements

    Walnut X Dynamic Range USB input SE output
    20220125 WalnutX dynamic range FFT ASIO SE t2.png
    Impressive performance for a Single Ended output!

    Dynamic range, in simplified terms
    1. find maximum output voltage at 0 dBFS
    2. find residual noise with a -60 dBFS 1 KHz stimulus, removed from analysis by window notch
    3. Dynamic Range is the difference between maximum output and residual noise
    Specifics are in AES17 section 9.3 (measurebators do your homework please)
    Example here:
    https://www.maximintegrated.com/en/design/blog/spec-dynamic-range.html

    Why we have a -60 dBFS stimulus
    Some clever codec designers include an output gate that shuts off when there is no signal present. This has the dual benefit of lowering output idle noise while also gaming the measurements. AES became wise requiring a -60 dBFS stimulus so any such gates are open during measurement and true Dynamic Range performance measured.

    Walnut X A04 THD+N THD nth-HD FFT USB input SE output
    20220125 WalnutX A04 THD+N THD nth-HD 4+HD+N 60Hz FFT SE 0dBFS - ASIO.png

    Walnut X 50 + 7000 Hz USB input SE output - Left Channel
    FFT_2_L1T13_1_A.png

    Walnut X Gain Linearity USB input SE output
    20220125 WalnutX amplitude linearity FFT ASIO SE.png
    Left channel Yellow
    Right channel Red
    Note the right channel gain linearity deteriorates more than the left channel, but both are better than many other WM8741 DAC implementations. At -100 dBFS the signal is below nano detail levels.

    Walnut X Crosstalk A to B USB input SE output - Left Channel
    G2_L1T10_1.png
    Very impressive for any type of output, SE or Balanced.

    Walnut X Residual Noise USB input SE output
    20220125 WalnutX residual noise FFT ASIO SE - 200 dB range t2.png
    It is apparent that careful attention was paid to power supply design and layout.


    Complete USB input SE output analysis report pdf attached
     

    Attached Files:

    • Epic Epic x 3
    • Like Like x 2
    • List
    Last edited: Jan 26, 2022
  3. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    16,409
    Trophy Points:
    113
    Location:
    On planet
    Filter Response 20 Hz square wave

    Measurement conditions
    1. -2 dBFS 20Hz square wave except when otherwise noted
    2. 44.1 KHz sample rate
    3. USB WASAPI input
    4. SE output to oscilloscope

    Walnut X has only one filter type enabled out of a number available in the WM8741: Linear phase
    20 Hz square wave response demonstrates DC output. Refer to the inset box in each of the oscilloscope displays.

    Walnut X 20 Hz square wave -2 dBFS 100uS/div
    20220122 SigGen WalnutX 20 Hz square -2dBFS 3550mVpp 100uS div 5MHz filter USB SE fltr ripple.png

    Detailed examination

    Why square wave observations are typically made at -2 or -3 dBFS
    Walnut X 20 Hz square wave 0 dBFS 100uS/div
    20220122 SigGen WalnutX 20 Hz square 0dBFS 4900mVpp 100uS div 5MHz filter USB SE fltr ripple.png
    Note distortion of the filter ring
    Keep in mind the number of DACs that leave sufficient headroom to avoid filter ringing distortion at 0 dBFS are in the minority.

    Walnut X 20 Hz square wave -2 dBFS 100uS/div
    20220122 SigGen WalnutX 20 Hz square -2dBFS 3550mVpp 100uS div 5MHz filter USB SE fltr ripple.png
    Filter ring no longer distorted

    Walnut X 5uS/div Lin BW calculation
    20220122 SigGen WalnutX 20 Hz square -2dBFS 3550mVpp 10uS div 5MHz filter USB SE BW calc.png

    Bandwidth estimation: BW (KHz) = 0.35 / RT (uS)
    Where RT = 10 to 90% Rise Time
    0.35 / 23.25 uS = 15.1 KHz

    Why I favor upsampling to 88KHz:
    Walnut X 20 Hz square wave -2 dBFS 100uS/div 88 KHz sample rate
    20220122 SigGen WalnutX 20 Hz sqr -2dBFS 3550mVpp 100uS div 5MHz filter USB SE fltr ripple 88K.png
    Note the filter rings fewer cycles than at 44KHz above.
     
    • Like Like x 4
    • Epic Epic x 2
    • List
    Last edited: Jan 26, 2022
  4. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    16,409
    Trophy Points:
    113
    Location:
    On planet
    reserved for additional data and / or corrections.
     
  5. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,307
    Trophy Points:
    93
    Location:
    Canada
    I find this rather surprising. The Walnut appears to run everything through an 24/192 ASRC, based on TomB pointing out the SRC4190 hardware chip, and seemingly no way to switch it on/off. So presumably, this is an effect of the ASRC, and not the DAC chip.

    I would have thought the ASRC would do a really good job of keeping things fairly constant across incoming sample rates that are integer multiples of each other. Maybe see some differences when comparing between 44.1 / 48 base frequencies. Interesting that the difference is so obvious.

    Then again, this presumes the Amanero is passing the native signal without modifying it in any way. Is there any way to force the Amanero to upsample to 24/192 before it hits the ASRC?

    [EDIT] Answering my own question, the Amanero seems to pass the incoming sample rate un-changed, and no way to set it differently. Likely puts this phenomenon back on the ASRC.
     
    Last edited: Jan 26, 2022
  6. dsavitsk

    dsavitsk Friend

    Friend
    Joined:
    Oct 2, 2015
    Likes Received:
    1,519
    Trophy Points:
    93
    Home Page:
    The output sample rate of the ASRC (SRC4192) is determined by the master clock you feed it. Here, that master clock is simply the recovered bit clock from the Amanero, so the ASRC is acting as a jitter reducer but not up or down sampling.
     
    • Like Like x 4
    • Epic Epic x 2
    • List
  7. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,307
    Trophy Points:
    93
    Location:
    Canada
    Huh, interesting! Is that a typical usage of that ASRC? I just recall way back when I built the Gamma2, the output was set as a fixed upsampling.

    (FWIW, I said SRC4190, because that's what it says on TomB's images in the first post of the build thread)
     
  8. dsavitsk

    dsavitsk Friend

    Friend
    Joined:
    Oct 2, 2015
    Likes Received:
    1,519
    Trophy Points:
    93
    Home Page:
    Beats me. Years ago I build a differential NOS TDA1543 DAC (best DAC ever, b/t/w/ :headbang: and the circuit that eventually led to both the Walnut and the DSHA). I had put an ASRC on the board for upsampling, but the chips couldn't handle the higher speed. So I used it that way instead. It worked, sounded better than not using it, and I have used them like that ever since.
     
    • Like Like x 6
    • Epic Epic x 2
    • List
  9. TomB

    TomB MOT: Beezar

    Joined:
    Dec 14, 2015
    Likes Received:
    1,718
    Trophy Points:
    93
    Location:
    Atlanta
    Home Page:
    The SRC4190 reference is a mistake. They're all SRC4192.

    This originated from one of those Mouser search hits ... like the several hundred 220K resistors I now have when I needed 220R instead and grabbed the first item from the Mouser search hits ...
     
  10. gixxerwimp

    gixxerwimp Professional tricycle rider

    Friend
    Joined:
    Sep 30, 2015
    Likes Received:
    5,192
    Trophy Points:
    113
    Location:
    small island claimed by China
    Just for clarity, this is the
    Walnut X.3, correct?
     
  11. TomB

    TomB MOT: Beezar

    Joined:
    Dec 14, 2015
    Likes Received:
    1,718
    Trophy Points:
    93
    Location:
    Atlanta
    Home Page:
    Since this was Dasman's DAC, then yes - X.3 with the Lundahl amorphous core transformers.
     

Share This Page