SBAF DAC Talk II

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by Maxx134, Jul 22, 2018.

  1. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    Adcom GDA-600 update haha!

    ref: SBAF DAC Talk II
    ref: https://tangentsoft.net/audio/opamp-bias.html
    ref: https://www.superbestaudiofriends.org/index.php?threads/class-a-bias-opamp.5813

    I researched a bit more on that Adcom GDA-600 alternate Class-A bias... the parts are J555 Current regulator diodes (CRDs) and not available anywhere (need to find a better modern equiv.)... I believe this is ~roughly the 2nd method biasing scheme in the tangentsoft link.

    Screen Shot 2022-07-26 at 12.23.45 AM.jpg Screen Shot 2022-07-24 at 9.08.14 PM.jpg

    Low and behold an old GFP-715 preamp uses the J555 on its op-amps... so I harvested them (put in sockets and replaced with biasing resistors for now) The thing needs work in other areas anyway, or may even get parted out/scrapped. But gonna leave it in a working state, and sockets to change later!

    dhGix.jpeg IMG_0223.JPG IMG_0224 2.JPG

    Also I'm temp using OPA1655s (single circuit version OPA1656s) with dual/single adapter to separate out the two and make adding the J555's easier (see pic) plus they're on hand... anyway want to explore everything, could be better or worse, gotta try it, who knows maybe modern OPA1656s don't need biasing!! Keeping it modular to swap back out etc.


    So far it sounds really good and there's a change, letting it play now- just finished!


    IMG_0312.jpg
     

    Attached Files:

    Last edited: Aug 9, 2022
  2. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    IMG_0234.jpg

    [See prev post] SBAF DAC Talk II

    Ok, being very thorough on this, it's interesting (hope it's not gobbledegook)... deleted the main-board bias resistors (note the hidden J555 pads)... lets me test straight no ClassA OPA1656, moved those resistors to the opamp mini board to test method1 ClassA OPA1656, also had to delete them to not conflict with the J555 OPA1655x2 (method2 ClassA), it's a bit rich/too liquidy the J555 + resistors, actually weird... incorrect circuit ie has to be one or the other.

    • OPA1656 Straight [deleted main-board ClassA resistors]
    • OPA1656 Adcom orig ClassA 8.25kohm resistors
    • OPA1655x2 Adcom orig ClassA optional J555 CRDs
    • OPA2156 SOON as I can get them -> OOS everywhere!

    So 3 easy swappable versions to find the best!

    1 ~maybe 2 of them sound quite stunning, going to be listening extensively and experimenting with it at friends' houses!
     

    Attached Files:

    • Like Like x 4
    • Epic Epic x 2
    • List
    Last edited: Jul 28, 2022
  3. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,320
    Trophy Points:
    93
    Location:
    Canada
    Very nice!

    I'll throw a curveball...... does the OPA165x sound better when sourcing or sinking that bias current? I remember the discussion from way back when the AMB MMM came out with 'class A' OPAMPs, and different OPAMPs apparently behave quite differently in this respect.
     
  4. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    Thanks! Seems like that would be the same thing? The opamp wants to return to zero... why is pulling to +15 vs -15 different? Is that what you mean? I'm kinda sticking to the original circuit, purposely.
     

    Attached Files:

    Last edited: Jul 28, 2022
  5. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,320
    Trophy Points:
    93
    Location:
    Canada
    I'm not an expert by any means, but my understanding is that an OPAMP stays in class A bias when its output sits slightly away from ground, and its output transistors never switch off.

    So yes, this can be achieved by the OPAMP sitting closer to the positive rail (sourcing current, most common setup) or sitting closer to the negative rails (sinking current). Different OPAMP architectures reportedly sounded better one way or the other.

    I honestly can't remember where I saw the discussion - Head-Fi, Headwize or DIYAudio - it was like 15 years ago......
     
  6. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,547
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    I think you can certainly give it a try. CMOS is typically not good for added heat and have a lot of issues with offset. Be careful @Scott Kramer.
     
  7. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    -V isn't ground... it's -15volts

    But will do it, I'll order 2 more OPA1656s and bias them to V+ as a 4th option.

    @Azimuth I check the heat, they never get too hot in all 3 circuits, it's toasty even without bias!
     
    Last edited: Jul 27, 2022
  8. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,320
    Trophy Points:
    93
    Location:
    Canada
    I'd bet a modern CMOS OPAMP has less offset than the original, unless it was a *very* good FET model for its time. I mean, the GDA-600 is over 20 years old now, isn't it?

    I take no responsibility for anything murdered during this experiment!
     
  9. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,547
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    I will echo what @Beefy said @Scott Kramer. If you bias to class A and it goes "poof", it goes poof.

    Check the schematic an make sure that bias is not already applied or check with a multi-meter. Adcom was notorious for doing this already.

    Edit: Nevermind, I looked at your previous posts above and you found what I saw on the schematic. Good stuff.

    What is also interesting is all that stuff on the output...the inductor, the cap across the RCA output. Strange. Unless THAT is the HP digital filter.
     
  10. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    Cmon' guys... this is done & working with and without Adcom's ClassA, they haven't fried --> please dig back into my thread on it. Been running fine for a few months... this new stuff is to explore no ClassA and the more advanced current regulator diode ClassA.
     
    Last edited: Jul 28, 2022
  11. Cspirou

    Cspirou They call me Sparky

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    7,447
    Trophy Points:
    113
    Location:
    Northwest France
    Pete Millett biased Opamps with a CCS in the Butte

    http://pmillett.com/butte.htm
     
  12. Armaegis

    Armaegis Friend

    Friend BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    5,946
    Trophy Points:
    113
    Location:
    Winnipeg
    I think the first I read of it was way back on Tangent's site...
    https://tangentsoft.net/audio/opamp-bias.html
     
  13. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,320
    Trophy Points:
    93
    Location:
    Canada
    Nice. He really did have the best, most easy to understand articles on so many topics. The issue of sinking or sourcing current was much more esoteric.....
     
  14. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    Yes, that's def. on my radar to verify; I did delete those LC parts... it's likely simple fcc (Adcom's original docs even talk about this) protection / cert, and those same circuits are on every input / output even the spdif. The first leg of the op-amp is where I believe the digital filtering is happening. Doubt they'd rely on that little dumb ass inductor to do audio filtering lol.

    I took it off coax2 input also... just to see if there's an effect vs coax1/aes/optical etc.
    Nice, referenced that a few times. Thanks! --> SBAF DAC Talk II
     
    Last edited: Jul 28, 2022
  15. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,547
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    Decent coax and AWS inputs are transformer isolated, but an inductor might be a band pass filter , which may glorify keep the digital noise down. I’d leave it.
     
  16. gixxerwimp

    gixxerwimp Professional tricycle rider

    Friend
    Joined:
    Sep 30, 2015
    Likes Received:
    5,264
    Trophy Points:
    113
    Location:
    small island claimed by China
    If only people would read your posts carefully, eh ;)
     
  17. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,320
    Trophy Points:
    93
    Location:
    Canada
    I just look at the pretty pictures.
     
    • Like Like x 3
    • Agreed, ditto, +1 Agreed, ditto, +1 x 2
    • List
  18. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    IMG_0324.jpg

    Nudging a PSAudio DLIII a bit forward (hopefully)…

    old shenanigans with early pi ref: Raspberry Pi2 and Hifiberry Digi+: A cheap transport that made me happier

    ...when I ordered DigiKey parts for the Adcom, also snagged a crazy XO for my less old school DLIII... a $27 Crystek. Installed that replacing a generic XO from 2006.

    Anyway it's the perfect (in theory and what I’m going for) XO for HiRes 192, integer 96 (192 being the native non-src/f’d with input) content on Qobuz/Local, that the Adcom can't play without downsampling... so far digging it, hyper-detailed yet subtle and nice; beauty with cymbals/piano with lots of room ambiance. It's better vs the old XO, it matters here.

    Also deleted the SPDIF input transformer (that back in the day was an upgrade?), digital sources are so good and isolated, why add this extra distortion?

    Notes:
    • The BurrBrown dac chip is a segmented R2R/DS hybrid type.
    • Has a BurrBrown SRC4192 (sample rate converter) chip, converts 44.1/88.2/176.4 properly with this XO, maybe not ideal... considering a 22.~MHz XO for that, but the enable/disable pin works differently on these, need to research it. This could indeed be an improvement, stock it also has the same problem.
    • Discrete output stage, no opamps!
    • EXT clock in is disconnected for now... until I figure out the Crystek enable/disable pin. This works with the Pi2AES, so want to figure it out!
    • It operates as if stock, still bulletproof.
    TLDR; Pictures!
     

    Attached Files:

    Last edited: Aug 9, 2022
  19. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,547
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    This one of the downfalls of this DAC. I was researching old BurrBrown DAC's and just saddened that Paul decided to just run everything native instead of dealing with different clock rates. If you can figure out a way to bypass that and just directly feed the CS8416 receiver, that would be more ideal. Because basically that SRC4192 upsamples everything to 96 or 192kHz.

    Otherwise, it looks like a great DAC as you say...PCM1798, only 8x oversampling, large torodial transformer, discrete output. It is a small SMD circuit board as well for such a big box.
     
    • Like Like x 1
    • Agreed, ditto, +1 Agreed, ditto, +1 x 1
    • List
  20. Scott Kramer

    Scott Kramer Friend

    Friend
    Joined:
    May 3, 2016
    Likes Received:
    1,414
    Trophy Points:
    93
    Agree, still a kinda sorta interesting, sexy DAC... Rick Cullen (original designer) modified. OG stock stock this actually had a PLL chip (no XOs)! I'm cheating in that Rick simplified and made it basically the standard XO architecture used today.

    SRC... YES, awhile ago explored setting the BYPASS pin on the SRC, no effect (verified on an ocilloscope) some other complexity in the way (startup init bits); even left a header for i2s. Not interested in that today. 192 HiRes is native on it; even improved via the crystek, and lucky to have other cool dacs!

    Finishing it off with selectable 176.4 native HiRes, which is at least integer math 88.2/44.1 would be a great final touch (it's too.friggin'.easy)! [REALLY want to go with the crystek XOs, they matter plus actually available.]

    I attached an NDK datasheet (these correctly disable on the PSAudio and on pi hats/Pi2AES/DigiOneSig) also added the Crystek datasheet to compare...

    It sure looks like it should work. Measured the voltage for it in the PSAudio (3.3V enabled / 75mV disabled). It just stays enabled no matter what, so can't add the 2nd 44.1x XO or EXT.

    Does anyone know or have an idea? Thanks!
     

    Attached Files:

    Last edited: Aug 9, 2022

Share This Page