How is Chord Dave?

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by mtoc, Jun 9, 2016.

  1. Torq

    Torq MOT: Headphone.com

    Pyrate
    Joined:
    Jan 15, 2016
    Likes Received:
    8,193
    Trophy Points:
    113
    Location:
    Seattle, WA
    Stage widens a little with DAVE and the Utopia vs. some other DACs, and DAVE's imaging with speakers is very impressive, but it doesn't suddenly turn the Utopia in HD800 or Abyss in terms of stage (using the direct headphone output).

    What does help with the Utopias stage are things like the matrix feature on the SPL Phonitor X, the 3D+ function on the iFi Pro iCAN, and various software "spatializer" tools like Waves NX.
     
  2. ufospls2

    ufospls2 Friend

    Pyrate
    Joined:
    Jan 10, 2017
    Likes Received:
    2,010
    Trophy Points:
    93
    Location:
    Canada
    Using the internal amp currently.
     
  3. EagleWings

    EagleWings Friend

    Pyrate Contributor
    Joined:
    Jul 20, 2017
    Likes Received:
    1,709
    Trophy Points:
    93
    Location:
    India
    In my experience, there are pros and cons to Dave's Headphone Out, as Dave's analog stage does not have a traditional amplification circuit. Volume is handled in the digital domain.

    Pros:
    - Works very well for linear-impedance, non-demanding HPs
    - Outstanding resolution, transparency, precision and plankton
    - Super black background and precise imaging

    Cons:
    - Can't handle hard-to-drive Planar HPs and doesn't have the best synergy with High-Z Dynamic HPs
    - Although the stage is very deep and in a way holographic, it still can't match the sheer size of the stage that some good desktop amps reproduce
    - Can't do bass impact like the desktop amps

    As a result, Dave's HO is not the most ideal for the popular full-size Dynamic HPs like the 800s or 650s. But I can only imagine the sheer resolution the Dave + Utopia combo can reproduce. Some of the best candidates for Dave's HO are, the easy-to-drive planar headphones like the LCD-i4, LCD-4Z, Aeons, Ethers etc.

    The tone of the HO is a bit neutral. So, people who are used to a fuller sounding mid-range, may find Dave's HO a bit on the dry side. But it's tonality and sound characteristics just work so well for classical, especially orchestral music.
     
    Last edited: Aug 8, 2018
  4. m17xr2b

    m17xr2b Friend

    Pyrate Banned
    Joined:
    Mar 26, 2016
    Likes Received:
    3,988
    Trophy Points:
    93
    Location:
    United kingdomland of fish and chips
    Finally cured the Dave itch, I've been thinking for a while of using the Dave as a DAC and having the single ended output as a casual listen.
    With Blue + Dave it was quite a letdown for my sennheisers. The DAC seemed capable enough but the amp doesn't have the capability to power the 105dB HD650 to even mid to high levels of quality. The uber microdetail was missing, bass couldn't hit the lowest the can is capable and soundstage didn't extend as per the usual. Very neutral but also somewhat polite.

    Pavane/Yggdrasil etc. + BH Crack is a far better solution for any high impedance sennheiser than just the 14K Blue/Dave combo.
     
  5. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    • Like Like x 10
    • Epic Epic x 4
    • List
  6. soumya

    soumya Acquaintance

    Joined:
    Jun 24, 2018
    Likes Received:
    42
    Trophy Points:
    18
    Location:
    Mordor, Middle Earth
    I can't express how badly I needed someone with an access to AP to post zoomed in FFT spectrum of Chord Dave. My colossal vote of thanks to you!
    I have been trying to reverse engineer WTA filter for past several weeks and while I think I have cracked it - I needed to validate my filter's response with Dave's measurements. The timing could not have been better!

    Sometime in 2020 , Watts posted a pretty length post on Head-Fi comparing the WTA to Kaiser and Rectangular. While he was smart enough not to provide any time domain measurements, he did post an AP FFT comparing the 3 window functions.
    And he went on explaining how WTA is different than other conventional windows. That got me thinking and triggered this whole exercise.

    So I figured out this - WTA at a very high level is a convolution of 2 windows - rectangular for most parts in the window (to optimize sync behaviour in time domain) but then it also needs to taper down to 0 as steep as possible so as to avoid leakage due to Gibb's phenomenon.

    Here's my take at reverse engineering 164000 taps WTA filter
    FFT of the filter
    [​IMG]
    Looks familiar ?

    Here's the white noise spectrum
    [​IMG]

    White noise zoomed in
    [​IMG]
    Deja Vu, right ?

    But what really matters - the Time Domain performance of the filter which no filter designer posts.
    [​IMG]
     
    • Epic Epic x 3
    • Like Like x 2
    • List
  7. soumya

    soumya Acquaintance

    Joined:
    Jun 24, 2018
    Likes Received:
    42
    Trophy Points:
    18
    Location:
    Mordor, Middle Earth
    That said , I am a bit surprise that Dave's stop band attenuation is only ~ 160dB. That's roughly 27 bit DR. Was expecting a full 32 bit DR as he did for M Scaler.
     
  8. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    Nice stuff!
    Any chance you could post the time domain result for something like HQP, PGGB or other popular upsampling options?


    Keep in mind that it's an analog measurement so the physical device and/or the measurement system itself will be a bottleneck in some instances
     
  9. soumya

    soumya Acquaintance

    Joined:
    Jun 24, 2018
    Likes Received:
    42
    Trophy Points:
    18
    Location:
    Mordor, Middle Earth
    That's daunting - only the FIR filter designer would have the actual window coefficients (not to be confused with final FIR filter coefficients). That said if they are using some form of composite window which is partly sinc in nature, we could make some smart assumptions by looking at the frequency response FFT provided the number of taps at a given Fs is known.


    So I wasn't verbose here - I was referring to this FFT that Rob Watts posted sometime in 2020 comparing the WTA to 2 other conventional windows when loaded in to Mscaler.
    [​IMG]
    If you notice, the analyzer was able to capture till - 192 dB. He doesn't mention how he captured this FFT , at what sampling rate etc.
    I don't know much about AP 555x and the accompanying software it comes with. But can AP 555x capture digital output over a dual BNC interface like Chord DACs? I very much doubt that.
    The other possibility is may be the AP software allows loading local wav files for analysis ?
     
  10. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    The APx555 can't accept chord's dual-BNC protocol as far as I can find.
    But there is an I2S module so it's entirely possible he was testing at the bus level prior to any SPDIF/Dual-BNC conversion?
     
  11. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,775
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Purely academic, but what's with the 100Hz and harmonics on the balanced output. Looks related to UK power (50Hz).

    [​IMG]
     
  12. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    That's digital clipping I think. (Hence why it stops at 22khz) Only occurs at max output
    Reducing vol by 1dB it goes away:

    [​IMG]
     
    Last edited: Mar 16, 2022
  13. SoupRKnowva

    SoupRKnowva Official SBAF South Korean Ambassador

    Pyrate Contributor
    Joined:
    Sep 26, 2015
    Likes Received:
    4,249
    Trophy Points:
    93
    Location:
    Austin, TX
    Im more curious about what is causing that little spike at like 270hz, its there in most of the measurements
     
  14. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    I'm not sure to be honest. Power freq here is 50hz so I doubt it's that.

    Looking back at the measurements it seemed to not be present when the DAC was in 'pre mode'?
    I'll set it back up and double check that at somepoint.

    In any case it's incredibly low so likely not much of a concern
     
  15. SoupRKnowva

    SoupRKnowva Official SBAF South Korean Ambassador

    Pyrate Contributor
    Joined:
    Sep 26, 2015
    Likes Received:
    4,249
    Trophy Points:
    93
    Location:
    Austin, TX
    I was also hoping you might do a few measurements investigating the no noise floor modulation claims? running with longer FFT lengths to get the noise floor lower like Rob says he does when he measures it. and maybe compare it to another dac with identical measurements?
     
  16. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    I'll do that shortly and will post once done :)
     
  17. GoldenOne

    GoldenOne Friend

    Pyrate
    Joined:
    Jul 17, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    England
    Ok here is the DAVE.
    To replicate the situation that Chord described in the DAVE tech presentation (available attached to this message) I set a -2dB sine into the DAVE in DAC mode which produced a 2.475v output on the RCAs.

    This was Chord's graph:

    upload_2022-3-16_19-59-42.png

    Firstly it seems that he'd set dBrA to something above the 2.5v he was measuring, I'm not sure exactly what but it looks like the fundamental is about 6dB below so I did the same and set dBrA as 5v.

    It is also not stated what windowing or FFT settings he was using and whether the high performance sine analyzer was enabled. It looks like he'd used a lot of averaging.
    In any case we're looking at relative differences so it doesn't matter too much. I used the AP-Equiripple window, 1.2 million point FFT, 44.1khz ADC sample rate and 10 averages.

    Firstly I did one with the high performance sine analyzer enabled to check how the harmonics line up:

    upload_2022-3-16_20-27-31.png

    The harmonic distortion on Chord's graph is MUCH lower, with all harmonics at MOST about -145dB relative to the fundamental.
    I'm not exactly sure how this was achieved given as I don't think I've ever seen a DAC with all harmonics below -145dB and so I'm doubtful as to whether that graph is entirely honest or perhaps just tested in an unorthodox manner. It does seem somewhat unbelievable if I'm honest. But all I can say is the unit I have here does not match that performance. It's still very good, and considering when the DAVE was first released this likely would have been one of if not the lowest THD+N DACs available.

    I couldn't do the silent output test with the high performance sine analyzer enabled as it creates a dip in the noise floor. (The high performance sine analyzer as the name implies is meant for testing sines not noise, or more complex signals). I also assume that therefore Chord didn't have the high performance sine analyzer enabled for his test as that dip was not present? So for the actual noise floor comparison I switched that off.
    We can indeed see that the noise floor modulation is nonexistant and regardless of what the DAC is outputting the result is the same.

    [​IMG]
    Now lets check a different DAC. I picked a fairly high performing option, the X26 pro.

    [​IMG]

    There is indeed some additional noise present when the DAC is operating at a high output vs a low one. Incredibly low at under -165dB, but still there. So it does indeed seem that the Chord approach is particularly good in this area.

    I also decided to recreate some of the other tests he'd done in that presentation.

    Jitter:

    I'd already done this in my measurement post, however I do not use the high performance sine analyzer when doing J-Tests as it can sometimes cause problems such as oddly shaped noise floors or altering the level of jitter products close to the fundamental.
    In any case, I redid it with the HPSA on to compare to Chord's graph:

    Rob's Graph:
    upload_2022-3-16_21-2-57.png

    Mine via USB (HPSA on):
    upload_2022-3-16_21-7-28.png

    Mine via BNC SPDIF (HPSA on):
    upload_2022-3-16_21-10-38.png

    There's some low level jitter products which aren't present on Chord's graph.
    Still truly excellent jitter performance, but the claim that Chord's DACs are 'immune' to jitter is not true as there is some present inherently and it can be increased via a jittery source too.

    Two-Tone 19/20khz Test:

    Chord's graph:
    upload_2022-3-16_21-15-44.png

    Firstly, interesting to note that this graph from chord shows the roughly 270hz spike that was present in my measurements but was absent from Chord's earlier graphs. The noise floor modulation test for example it's not there in Chord's graph.

    Mine (zoomed on the X axis to show things a bit clearer):
    KEI7yKEypi.png

    About 10-15dB higher on the 18khz/21khz products than chord's own graph.


    Overall it seems that some of the graphs Chord has put out are.....well 'optimistic' to say the least....
    But the claim that they handle noise floor modulation well is true.
     

    Attached Files:

    • Like Like x 4
    • Epic Epic x 2
    • List

Share This Page