Advice for Schiit on how to improve Magni 3+ power supply

Discussion in 'How to Win Friends and Influence People' started by kris2014, Mar 31, 2020.

  1. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    Well I will add a little box to the list for recording comparison. I do agree we don't have many great out-of-the-box choices out there.

    Many years of training taught me one thing, manufacturer documentations are to be respected. Things are sometimes too complicated so there are just too many small caveats that were not well explained unless you made inquiries personally, but if one doesn't follow it to the letter... Well, they'd been warned.
     
  2. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Those wiggles, when extrapolated to a complete cycle, have a period of approximately 165 nS, which is consistent with a frequency of 6 MHz. Not even close to a 5 KHz oscillation.

    That waveform is a textbook classic underdamped response with only a few cycles of ringing. The ringing is at approximately somewhere between 5 ~ 7 MHz, which is 3 orders of magnitude higher than 5 KHz.

    I'm having trouble locating the 4.8 KHz spikes you describe on this graph cited, which happens to be 19+20 KHz IMD. From 1 KHz to 18 KHz there is nothing above -120 dBu.
     
  3. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    But these aren't directly measuring psu output either. Depending on current draw profile of load, it is possible to cause performance issues above 5kHz. At least we can to acknowledge that we do have these ringings.

    Let's hold off debating until I get to it.
     
  4. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    At the end of the day, users care about the system performance, not intermediate internal stages. By all assessments, both measurements and listening, Magni 3+ appears to satisfy user expectations well. I have long stated the Power Supply is at least 50% of any component's performance. But I am not observing the perceived deficiency being described.
     
  5. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    But hey, better is the enemy of good. I think this design is pushed closer to the limit where issue start to emerge with current iteration. The ringing wasn't observed in older Magni 3. Let's see if it is possible to squeeze a little more out of it without breaking the bank.
     
  6. Walderstorn

    Walderstorn Friend

    Pyrate
    Joined:
    Mar 20, 2016
    Likes Received:
    1,904
    Trophy Points:
    113
    Location:
    Europe
    Well man, if you can do it, it will be a feat and, i am sure, that we are all curious. Something positive can come out of this (in several different angles) so keep us updated.
     
  7. auvgeek

    auvgeek Acquaintance

    Joined:
    Jan 14, 2020
    Likes Received:
    31
    Trophy Points:
    18
    Location:
    around
    And yet, you're unwilling to give Schiit the same leeway?
     
  8. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Pontification is the enemy of execution.

    Data presented was collected using a picotech 5243b oscilloscope with tektronix x1 passive probe.

    Magni 3+ PCB - power supply +17V and -17V measured on the amplifier side of the 1R00 resistors
    IMG_2577_small.jpg IMG_2578_small.jpg

    Magni 3+ power supply +17V noise spectrum
    20200402 magni 3+ ps +17V spectrum.png
    We observe 60Hz ripple at approximately -75 dBu, a -80 dBu spike at 1.919 KHz and again at 27 KHz. Nothing here of any concern. The amplifier has a power supply rejection ratio further reducing power supply noise as will be seen several graphs below.

    Magni 3+ power supply -17V noise spectrum
    20200402 magni 3+ ps -17V spectrum.png
    Observations similar to +17V with an additional 120 Hz component at approximately -78 dBu. Again, nothing of any concern.

    Magni 3+ residual noise 300R load - left channel
    FFT_2_L1T16_1_A.png

    Magni 3+ residual noise 300R load - right channel
    FFT_2_L1T16_1_B.png
    From 1 KHz to 24 KHz there isn't any spectral component above -140 dBu. Put in perspective, -140 dBu is 77.5 nV.
    Mains hum is below -120 dBu or 775 nV.

    In an extremely quiet setting of 25 dBA, a 1KHz sine at -75 dBu would produce 25 dB SPL in HD6x0 headphones, on the threshold of audibility. Residual noise of Magni 3+ is at least 45 dB lower than HD6x0 threshold of audibility in a room ambient of 25dBA. Also remember dB is logarithmic, similar to earthquake scales.

    I am not observing any power supply design deficiencies, measured either directly on the power supply internal to Magni 3+ or at the headphone output. Quite the opposite. I suppose one could argue there is always room for improvement, of course. Always appreciate lower mains noise. But at what price. This is an exceptionally performing amplifier, especially considering the $100 US price.
     
  9. Elnrik

    Elnrik Super Friendly

    Pyrate
    Joined:
    Jan 10, 2017
    Likes Received:
    8,973
    Trophy Points:
    113
    Location:
    Denver CO
    Home Page:
    Case closed, as far as I'm concerned. Time to stop feeding the troll.
     
  10. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    I just realized this was made easier.

    What I did is removing C3-C6, then installing 2x 4.7uf tantalum caps in place of C5/6.

    Blue curve was before mod and white is after mod. Very obviously by removing those MLCCs and downgrade the big caps (from 2x 1000uf nichicon RSM to 2x 4.7uf tantalum), performance of psu increased... (7kHz sinewave playback and recorded with zoom H5).

    Quick comment on parts selection: I actually should use a 10uf Tantalum but due to the smd size I cannot. But diyers could consider putting them in place of C3/4. If bumped to 10uf, lower frequency noise would also be lower. Going above 10uf could cause resonance peak below 20kHz.

    [​IMG]1

    So we are back on the quetion, can we make magni 3+ better. I think the answer is positive. And I guess the bom doesn't need to go beyond $150 USD for indivudial.

    Please note that recorders like H5 is a very high impendance load and won't be performing as well as audio anaylzers in terms of noise floor, so I cannot get my system to go down to -150dB as those properlly geared for this. Also, this amp potentially doesn't like high impedance load like H5, as I canont get it to work properly under low gain mod.

    There are a few other points can be improved. The first thing is that those 1R00 resistors in R2/3 would introduce a 0.1V drop with every 100mA the circuit draws, reducing its value to 0.33R and install a 1000uf behind it would stablize voltage rails for the opa1678s. The second thing is that those output capacitors next to headphone jack could use 2x 470 ohm resistors connecting their negative end to the GND. This would reduce on/off pop. The thrid thing is that the 2x 2200uf input caps are a bit far from the regulators, so bypassing them with 0.1uf MLCC would help.

    One extra bit of opinion. Manufacturer datasheets are written by really smart people, who sometimes forgets muggles don't know things as well as they do. So the writers failed to explain something wrongly assuming that users know them, and users found the datasheets "not working", which is... I leave that to you.
     
    Last edited: Apr 3, 2020
  11. penguins

    penguins Friend, formerly known as fp627

    Pyrate
    Joined:
    Sep 26, 2018
    Likes Received:
    3,451
    Trophy Points:
    113
    Location:
    SouthernCA
    ^ could you provide some thoughts on the sound when you get a chance?
     
  12. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    Will do. I previously recorded 2 sample tracks. When I get back to it I would record again and compare.
     
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
  14. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    Quick note on the price changes due to mod.

    Parts removed (buk pricing from digikey)
    2x Nichicon URS 1000uf/25V (~0.26USD each)
    2x 0805 MLCC probably 0.1uf x7r (TDK ones are 0.06 USD each)

    Parts added
    2x 4.7uf 25V tantalum (ones I bought are 0.74 USD each, but AVX has some for 0.2 USD each)

    So it is possible to save some money with better results, especially with heresy. As Heresy uses 2x Panasonics HD 2200uf/25V in place of the nichicons, which is 0.81 USD each .

    Total change is ~1 USD more than original. But considering that I save some of space on a 4-layer pc board, it probably breaked even if layout was changed accordingly to bring components closer, which would further enhance performance (LM317 should be on the bottom of the board swith pins stitting next to input caps) and cut down manufacturing cost.

    Void
    [​IMG]Untitled by
     
    Last edited: Apr 3, 2020
  15. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    Purpose aligned on this topic. However, I hope you are convinced that I wasn't talking without constraint on this.
     
  16. schiit

    schiit SchiitHead

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    9,937
    Trophy Points:
    93
    Location:
    Texas and California
    Home Page:
    So I guess I'm a "muggle?" Cool. I will refrain from any veiled or direct insults myself, just as we refrain from ever commenting on another company's products.

    What I will say is that some of your assumptions about what caps are on the board, their values, and their functions, are technically incorrect. (As in, there are no output coupling caps and there are plenty of 0.1s on the board for local bypassing.) And I will stand by our design, based on our measurements and experience.

    And, as far as recommendations go, maybe I've been doing this too long, but I'm not big on tantalums due to their potentially firey failure mode, which appears to still be a thing: https://electronics.stackexchange.c...ntalum-capacitors-safe-for-use-in-new-designs

    Yes, I know, we can debate endlessly about "what if," and "well, maybe," but here's the thing: anything can be second-guessed. One of the favorite things our teachers used to do on the English (writing) side of things was to lob us a piece of a published work by a famous author for critique. It was always savaged. And it was always from a published work considered not only fit for print, but exemplary of its type.

    Yes, I know, this example is not electronics, but it's a good analogy. I look at some products and think, "Why didn't they do it like that! That's crazy!" But it's also not my place to comment on other products. And, you know what, I may be wrong--I may have misinterpreted what they were trying to do, or missed something else in the design.

    That said, if you'd like, send the modded board back to us, and we'll run it on the APx to see what kind of performance differences we see. I'll just overlay the results on a standard board. Or, maybe AtomicBob would be game. That's entirely up to him, though.
     
    Last edited: Apr 3, 2020
  17. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    Why not, I am interested in knowing the performance too.

    Sorry about misinfo on the output caps, I just checked again. This requires spme rethinking.

    As for muggles... It depends on the prespective. For those who are smart enough to work out these tiny devices in the 1970s, many of us are subpar intellectually. In fact, there are only about 2.5% of total US population have PhD.
     
  18. schiit

    schiit SchiitHead

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    9,937
    Trophy Points:
    93
    Location:
    Texas and California
    Home Page:
    Good to know that I'm subpar intellectually. It confirms what I've always suspected!
     
  19. kris2014

    kris2014 Acquaintance

    Contributor Banned
    Joined:
    Feb 1, 2020
    Likes Received:
    38
    Trophy Points:
    18
    Location:
    us
    Since my statement includes myself as well, I stand by it. At least I won't challenge the datasheets as if I know things better than they do...
     
  20. schiit

    schiit SchiitHead

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    9,937
    Trophy Points:
    93
    Location:
    Texas and California
    Home Page:
    And I will always report our experiences, even if they go counter to the datasheets. So I'm now a terrible person for not immediately bowing to authority? Another insult? That would make three, if anyone's counting. Or maybe I'm mistaken. It's been a stressful few weeks here.
     

Share This Page