Hacking Schiit

Discussion in 'Modifications and Tweaks' started by Marvey, Jul 22, 2021.

  1. Pancakes

    Pancakes Friend

    Friend
    Joined:
    Aug 13, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    Atl
    So what's involved in a DAC? PSU, DAC chips/analog boards, some digital input(s)...where does the filter live? The EEPROM?

    I know next to nothing about DAC design but probably someone who does can put together an Yggdrasil A1 for well under a grand.
     
  2. Beefy

    Beefy Almost "Made"

    Joined:
    Apr 10, 2021
    Likes Received:
    1,306
    Trophy Points:
    93
    Location:
    Canada
    In most cases, the filters live within the DAC chip, because they are purpose built audio DACs. They run directly from an audio standard input, SPDIF, AES or I2S, do the filtering on-chip, and spit out music.

    To the best of my knowledge, the Schiit multibit DACs are quite different. They are not audio DAC chips, and they run natively off a non-audio SPI interface. So the SPDIF/AES/I2S audio stream comes in to a dedicated signal processor that applies all the digital filtering, converts to SPI, and the DAC just processes that into music.

    Without extremely solid skills to implement a digital filter and SPI converter, this is a very difficult project.
     
    Last edited: Oct 7, 2021
  3. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,193
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Right, looks like you'd need to essentially make your own DSP board or find a way to obtain one meant for the Yggdrasil.

    Some of their multibit DACs, like the 1st gen Bifrost and Gungnir at least, have the DSP components on the same board. Doesn't look to be the case with the Yggdrasil.

    Nonetheless, could be a fun project for someone equipped to handle the DSP aspect.
     
  4. Pancakes

    Pancakes Friend

    Friend
    Joined:
    Aug 13, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    Atl
    Interesting, Bifrost 2 has a separate analog board from what I can tell. Maybe someone can build a Byggy.
     
  5. JellyRhino

    JellyRhino Facebook Friend

    Joined:
    Jan 27, 2020
    Likes Received:
    165
    Trophy Points:
    33
    Location:
    Montreal, Canada
    When I look at the Bifrost 2 cards on Schiit's website, I can see the ARUZ DAC chips on the same board as the outputs. They're at the bottom if I'm not mistaken.
     
  6. Pancakes

    Pancakes Friend

    Friend
    Joined:
    Aug 13, 2020
    Likes Received:
    1,049
    Trophy Points:
    93
    Location:
    Atl
    Yeah that exactly what I was saying. That's how the Yggdrasil is as well. Maybe some smartass out there can make it work. I dunno...I'll stick to analog stuff.
     
  7. dBel84

    dBel84 Friend

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    2,773
    Trophy Points:
    113
    this thread has me looking at the old Theta I was trying to fix and had put on the back burner - time to keep an eye out for a modi 3+
     
  8. Cspirou

    Cspirou They call me Sparky

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    7,308
    Trophy Points:
    113
    Location:
    Northwest France
    Last edited: Oct 17, 2021
  9. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,341
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    So the TL082 is not the servo? Do you know what is it doing then?

    Have you also tried the OPA2604 as a servo? That one was my favorite when swapping around servo op amps in my Dynalo amp. Actually it runs on two of the single channel versions the OPA604. But it actually was a big difference over the TL071 that was originally in there.
     
  10. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,341
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    And just to verify...it's U17 & U18 (bottom) for the I/V and U10 (top) for the buffer?

    Edit: Apologies, I now see Marv's first post is a Modi 3+ and not a Modi MB. So looks like for the Modi MB, only one is needed for U10 and two are needed for Modi MB for U17 & U18.
     
    Last edited: Nov 16, 2021
  11. Grattle

    Grattle Friend

    Friend
    Joined:
    Feb 17, 2017
    Likes Received:
    448
    Trophy Points:
    63
    Location:
    USA
    For Modi MB, I only replaced 2 opamps.
     
  12. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,193
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    I believe the TL082 is the servo, yes. I was only mentioning that some of their products might use a different/newer part instead. For all I know, maybe newer Modi MBs no longer use the TL082 for the servo.
     
  13. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,341
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    Turns out the OPA2604 is a discontinued item from TI. Apparently they still make the single channel OPA604.

    Then I found that the OPA1656 is the replacement for both the OPA2604 and OPA2134:

    https://e2e.ti.com/support/audio-gr...xt-generation-opa2134-and-opa2604-replacement

    So try a OPA1656 as a servo as well? Worth a shot maybe. I found a place that has OPA2604's SOIC's in stock, but they are $5 a piece and have a $25 minimum order. Looks like they buy discontinued parts and then slap a minimum order on to help stay afloat. Anybody else curious about OPA2604? - will cost me the same for 1 or 5.
     
  14. ergopower

    ergopower Friend

    Friend
    Joined:
    Mar 21, 2018
    Likes Received:
    607
    Trophy Points:
    93
    Location:
    South Central PA
    IF an OPA1656 works, then so ought an OPA2156. I can mail you one if you want to try it.
     
  15. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,341
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    I went ahead and bought three 1656's. I will have to monitor the heat to see of it oscillates and gets super hot. For the offset, I would imagine the more precise 2156 might be a better choice. I send you a PM.

    On mine the DC offset op amp looks different.

    20211117_180533.jpg

    Had to look it up. Appears to be an ST Microelectronics version of the TL082. Likely just a subbed part from the board manufacturer. No biggie. Just something that someone may also trun across.
     
  16. purr1n

    purr1n Burned out

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    81,295
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Parts subs are normal. Some stuff is more available than others. Don't be surprised if the AD8512s suddenly become OPA2156s or something else that is compatible.

    You can try playing around with the servo opamp. I'm kind of curious myself. Servos feedback a low pass at ultra low frequencies near DC, so theoretically it should not affect the sound very much.

    OPA1656 and OPA2156 have been confirmed to work with no oscillation on the AP. What you really want to look for is a fast settling time. OPA1656 does not have this spec listed in its datasheet, but it works. This what stuff like the APx555 is for - not for people like Amir and his cultists to be SINAD retards.

    --

    Other mods to try - on Magni 3+:
    • Bypass the relay, remove the relay, jumper last two pads closest to front of the board.
    • Replace small ceramic cap right after the input (RF filter?) with a film cap. 100p. I think it's C100 and C200, but boards do get revised and stuff changed around.
    • Replace small ceramic cap at C103 and C203 (feedback) with film cap. 33p or maybe higher. Just make sure no oscillation.
     
  17. Cspirou

    Cspirou They call me Sparky

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    7,308
    Trophy Points:
    113
    Location:
    Northwest France
    Older audio DACs like the PCM1704 had external filters too. Like the DF1704 chip.

    I believe the Soerkris DAC gives you easy access to your own custom filters as well.
     
  18. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,341
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    This is what I figured. The TL071, 072, 081, and 082 are super old and now made by a few different companies since the patent is expired. Same for stuff like 4556, 4562, 5532, etc.

    My question is that if applied a negative gain, the 1656 and 2156 flip the polarity. So hopefully there is not negative gain applied after the AD5547.

    Thanks for checking the oscillation. I remember I tried swapping a few in my Dynalo and servos and one got superhot, so I knew it was likely oscillating. This is about the only way for me to tell, and if of course sounded terrible. I know the servo should not make much difference , but it can. I will say it mostly affects the dynamics and less of the FR or clarity. It just seems to give more and have better space and sound less strained. It may make less of a difference when just a line level device.

    It is also you mention settling times, because I was looking at the difference between one of my favorites in the I/V position, the AD827. TheOPA1656 and the 1656 did have a better slew spec, but the 827 did have a better settling time. The 827 is only made in PDIP and would need an adapter for Modi, which will not fit the Modi MB because those type PDIP to SOIC adapters require about 7/8" in height. And since the MB op amps are on the bottom of the board, only same size SOIC's are going to fit. Although it might work in the Modi 3+ since the final op amp is on top.

    Anyway, back to settling times for the servo position, here is a breakdown of a few:
    • TL072 - 0.91 us (0.00091 ns)
    • TL082 - 0.91us (0.00091 ns)
    • OPA2604 - 1.5 us (0.0015 ns)
    • OPA1656 - 800 ns
    • OPA2156 - 600ns
    So even if the settling time is more with TL072 vs 2604 (0.91 vs 1.5 us), somewhow the 2604 (or really single channel 604) sounds better at least in my Dynalo. I know the MKI Gilmore used TL071 and the current MKII uses the OPA604.

    Looks like I probably need to pick up a 2604 and experiment. I have a 2156 and 1656 coming now, but not as excited about it now looking at these specs because objectively does not look as good. I will have to check as I know specs don't mean everything.

    Tech Note: All the TI specs for settling are at 0.01% with Gain at -1, and at 10-V steps because those were the only ones the same across all the spec sheets. Of course the settling time goes way down with less voltage change and looking higher % of settling. I did not include AD827 because specs were not comparable. AD827 spec, if you are curious, is listed at 65ns at 2.5V steps and 120ns at 5V steps at 0.1%.
     
  19. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Friend MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,193
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Wouldn't 0.91us be 910ns? And 1.5us be 1500ns?

    Regardless, I know some individuals are all about slew rate and settling time for I/V opamps. I just went with what I thought sounded best after trying many options that seemed to meet use case requirements on a fairly holistic level.

    The original Bifrost MB has more clearance room between the DAC board and chassis. But newer iterations might be rocking 8X oversampling, so sound won't always be quite like the Modi MB.

    You could always go crazy and get a larger chassis for the Modi MB. ;)
     
  20. Azimuth

    Azimuth FKA rtaylor76, Friend

    Friend
    Joined:
    Nov 9, 2015
    Likes Received:
    5,341
    Trophy Points:
    113
    Location:
    Nashville, TN
    Home Page:
    I think I was thinking pico rather than micro. µs. So you are correct. 0.91µs = 910ns
    • TL072 - 0.91 µs (910 ns)
    • TL082 - 0.91µs (910 ns)
    • OPA2604 - 1.5 µs (1500 ns)
    • OPA1656 - 800 ns
    • OPA2156 - 600ns
    This means that the OPA2156 would have the faster settling time out of all of them technically speaking.

    True. Subjective impressions to me mean more that the spec sheet says.

    The Bifrost board is on a riser already, so you would have to double check clearance for sure. For the Modi MB, a larger box to accommodate the extra depth would be another level. The PDIP to SOIC adapters look crazy enough.

    PS - Alt+230 to make the micro µ symbol for future reference for anyone. This is very helpful in capacitor values.
     

Share This Page