SMPS vs LPS effect on Garage1217 Project Sunrise III

Discussion in 'Headphone Amplifier Measurements' started by atomicbob, Sep 15, 2017.

  1. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    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 this post.

    The data presented were acquired as follows:

    1. PrismSound dScope III, picoscope 5243B
    2. Garage 1217 Project Sunrise III headphone amp
    3. Electro-Harmonix 12BH7 tube
    4. 300 ohm load used for measurements
    5. Unbalance RCA cable - Sescom with Canare L-4e6s cable
    6. Balanced 4-pin XLR cable - Cable Matters 16 awg and Neutrik XLR connectors
    7. Shielded 14AWG and 16AWG power cables
    8. SMPS Meanwell GS40A24-P1J 24Vdc 1.67A 40W max
    9. LPS 1 Jay's Audio Talema 24Vdc 1A 25W max
    10. LPS 2 Keysight (HP/Agilent) U8001A 0 to 30Vdc variable 3A lab supply


    Presented is a comparison of 3 power supplies, their noise and respective effect on measurements using a Project Sunrise III headphone amplifier. The level chosen for this comparison, -24 dBu, highlights the effect on lower level signals as heard with an HD800 at approximately 76 dB SPL for a nominal of 85 dB SPL.

    Notable highlights:

    PS-III measurements improve with lower noise power supplies. Subjectively my ears tell me the same thing. Using an RME ADI-2 Pro and the U8001A enables an extraordinary blackground from which music appears.

    Setup picture:
    DSC_0010b.jpg



    Meanwell SMPS 1mVrms 4mVpp ripple 650mA 50 KHz BW:
    01 20150611 PS-III SMPS 24V 1mVrms 4mVpp ripple 650mA load 50KHz BW filter 2mV 5mS div.png

    Jay’s Audio Talema LPS 24V 815 uVrms 3mVpp ripple 650mA load 50KHz BW:
    02 20150611 Talema LPS 24V 815 uVrms 3mVpp ripple 650mA load 50KHz BW filter 2mV 5mS div.png


    U8001A 24V 110uVrms 670uVpp ripple 650mA load 50KHz BW:
    03 20150611 U8001A 24V 110uVrms 670uVpp ripple 650mA load 50KHz BW filter 2mV 5mS div.png


    PS-III THD, THD+N and 1 KHz spectrum - Meanwell SMPS:
    20170908-1 Project Sunrise III SE THD THD+N -24 dBu - w FFT - meanwell.PNG


    PS-III THD, THD+N and 1 KHz spectrum - Jay’s Audio Talema LPS 24V:
    20170908-2 Project Sunrise III SE THD THD+N -24 dBu - w FFT - talema.PNG


    PS-III THD, THD+N and 1 KHz spectrum - U8001A 24V:
    20170908-3 Project Sunrise III SE THD THD+N -24 dBu - w FFT - keysight.PNG

    A rather substantial improvement in THD, THD+N and spectrum occurs with the Talema 24V LPS over the Meanwell SMPS (provided with the PS-III.) Looking at THD and spectrum, another improvement occurs with the U8001A LPS.
     
    Last edited: Sep 15, 2017
    trl, Priidik, NekoAudio and 16 others like this.
  2. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    3,007
    Dislikes Received:
    7
    Trophy Points:
    113
    Location:
    Laguna Hills, CA
    Home Page:
    That's really impressive, honestly. I knew you'd be able to measure it, but the improvements are significant.

    I knew I was hearing improvements with my own setup using an HP/Agilent/Keysight E3617A with the Project Horizon III, but I love having objective data to back up my perception.

    Don't mind the precarious setup; this has been displaced by the Eddie Current Super 7 and is serving leaning tower duty.

    IMG_20170915_225932.jpg
     
    Last edited: Sep 16, 2017
    Dino, atomicbob and Cspirou like this.
  3. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    What we have is a case where measuring the correct attribute (spectrum and distortion vs frequency) correlates to the auditory experience.
     
    Cspirou and Dino like this.
  4. Cspirou

    Cspirou They call me Sparky

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    4,962
    Dislikes Received:
    5
    Trophy Points:
    113
    Location:
    Northwest France
    Thanks for the excellent measurements.

    The article I read by Benchmark about using a SMPS for audio said that they needed to be custom made and couldn't just use an off the shelf supply. I think the switching frequency is 500khz for the AHB2.

    I have a project I'm working on now with a Meanwell followed by an ultralow noise linear regulator. While you could do better, for portability I think it's ideal.
     
    Luckbad likes this.
  5. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    Benchmark probably has designed an SMPS that better serves the intended audio usage, but they are also probably paying attention to shielding and better filtering prior to the active amp circuits. I will be measuring both the Meanwell SMPS and Jay's audio LPS with a power LC filter when the parts arrive. They are currently lost on the road somewhere in Portland OR.
     
  6. peef

    peef Almost "Made"

    Joined:
    Oct 5, 2015
    Likes Received:
    108
    Dislikes Received:
    0
    Trophy Points:
    33
    The measurement of the Mean Well is really interesting. Eyeballing it, looks like there's some 60Hz, a lot of 500Hz, and some RF junk? Possibly more past 50kHz? Usually the Mean Well datasheets specify the switching frequency, but this one does not.
     
    Cspirou likes this.
  7. Grahad2

    Grahad2 Red eyes from too much anime

    Friend
    Joined:
    Jan 26, 2017
    Likes Received:
    1,218
    Dislikes Received:
    2
    Trophy Points:
    113
    Location:
    Singapore
    Congratulations must be given to THX's engineers, since they did most of it.
    I'm actually curious to see if the iFi iPower tech holds up well against the better LPSUs, but they don't have 24V versions.
     
    SHAWN P. WATSON likes this.
  8. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    3,007
    Dislikes Received:
    7
    Trophy Points:
    113
    Location:
    Laguna Hills, CA
    Home Page:
    @Grahad2 I can't imagine they hold up. I had one burn itself out at random, one cause intermittent buzzing noises, and all of them spark when plugged in if you connect it in the wrong order.

    While I think the sound is pretty good from the iCAN and SE, those power supplies are volatile as f.

    That said, I would be very interested to see measurements. Maybe they're technically great apart from the crappy quality control.
     
  9. Grahad2

    Grahad2 Red eyes from too much anime

    Friend
    Joined:
    Jan 26, 2017
    Likes Received:
    1,218
    Dislikes Received:
    2
    Trophy Points:
    113
    Location:
    Singapore
    I recall they measure better than many LPSUs, but it was done by someone who posted the noise numbers and not charts.
     
  10. jexby

    jexby Posole Prince

    Staff Member Friend JK47
    Joined:
    Sep 28, 2015
    Likes Received:
    4,889
    Dislikes Received:
    37
    Trophy Points:
    113
    Location:
    @zerodeefex's basement
    i have an iFi iPower 9V sitting around doing nothing if @atomicbob wants it for any measuring or hacking it apart for analysis.
     
    Luckbad likes this.
  11. Priidik

    Priidik Friend

    Friend BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    1,447
    Dislikes Received:
    2
    Trophy Points:
    93
    Location:
    Estonia
    @atomicbob it would be interesting to see 20 Hz results and something of more dynamic nature, like 20 Hz sqw in time domain, or pulsed loading.
    I have come to think besides noise performance the source resistance is just as important for psu-s.
    Looks like recent switchers, which allow for high output capacitance are quite good in that.
    Perhaps in low freq pulsing the switcher could win over small linear device?
     
  12. Armaegis

    Armaegis Friend

    Friend BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    3,701
    Dislikes Received:
    2
    Trophy Points:
    113
    Location:
    Winnipeg
    I think the bigger culprit was less radiated mag/elec fields from the SMPS affecting interconnects. They did a video on it somewhere, but I think it boiled down to yeah the SMPS has noise kerfuffles too but they were of a type that were lower in magnitude and/or easier to filter out.
     
  13. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    It would be interesting. However in practice this is difficult to measure with the resolution desired at the levels that would be meaningful. Consider the voltage and current pulse for a 20 Hz square wave to drive an HD800 to 100 dB SPL and the relative stiffness of the supply at 40 dB below (60 dB SPL). This gives us the following (HD800 assumed 300R and 97 dB/mW):
    100 dB 2.191 Vpp 7.3 mA peak
    60 dB 21 mVpp 73 uA peak
    Measuring dynamic effects at 73 uA is rather challenging.
    Then there is the issue of regulation and noise on the programmable electronic loads used for power supply dynamic performance analysis. Have a look at the specifications for one of the better units (now discontinued):
    http://www.keysight.com/en/pd-1000001519:epsg:pro-pn-6060B/300-watt-dc-electronic-load?cc=US&lc=eng
     
  14. Priidik

    Priidik Friend

    Friend BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    1,447
    Dislikes Received:
    2
    Trophy Points:
    93
    Location:
    Estonia
    Hey I used to calibrate those. These do deserve their pricetag and cult status. For hp amp use though, way too beefy and perhaps noisy.
    I have messed around with bunch of resistors and checked the results on scope, ofc scope doesn't have much dr.
     
    atomicbob likes this.
  15. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    The picotech 5243b is one of the few scopes with 16 bit hardware A/D resolution (1 ch only, 2ch is limited to15 bits). I looked long and hard to find this particular feature. But even with this resolution and 10R current shunt, 100 uA measurements are at the threshold of my reach.
     
  16. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    A simple LC filter to achieve lower ripple noise with the MeanWell SMPS and Talema 24V LPS consisting of
    a. Hammond 155B 6mH choke
    b. Nichicon 3300uF 63V capacitor

    LC circuit and simulation

    09 20170919 LC circuit.png


    Performance improvements measured:

    Talema 24V LPS:

    07 20170919 Talema LPS + LC 24V 52 uVrms 248 uVpp ripple 650mA load 50KHz BW filter 2mV 5mS div.png

    MeanWell GS40A24 SMPS:
    05 20170919 PS-III SMPS + LC 24V 309 uVrms 1mVpp ripple 650mA load 50KHz BW filter 2mV 5mS div.png

    MeanWell GS40A24 SMPS with 6mH choke and 6600 uF capacitance:
    PS-III SMPS + LC 24V 123 uVrms 488 uVpp ripple 650mA load 50KHz BW filter 2mV 5mS div.png


     
    Last edited: Sep 19, 2017
    mrflibble, Luckbad and Dino like this.
  17. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Friend
    Joined:
    Sep 27, 2015
    Likes Received:
    3,007
    Dislikes Received:
    7
    Trophy Points:
    113
    Location:
    Laguna Hills, CA
    Home Page:
    Dude, that is no joke. Show us how to build it, then make a nice looking version and sell it for $100.
     
    SHAWN P. WATSON and Dino like this.
  18. atomicbob

    atomicbob dScope Yoda

    Friend BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    10,364
    Dislikes Received:
    4
    Trophy Points:
    113
    Location:
    On planet
    Here is the not-so-nice looking lab test version:
    IMG_2145a.jpg
     
    bilboda, Dino and Luckbad like this.
  19. winders

    winders Mentally ill. Do not engage.

    Friend CBC
    Joined:
    Feb 13, 2017
    Likes Received:
    2,109
    Dislikes Received:
    366
    Trophy Points:
    113
    Location:
    San Martin, CA
    That's impressive!! So much to do so little......
     
  20. Joe Bloggs

    Joe Bloggs Rando

    Joined:
    Jan 14, 2017
    Likes Received:
    19
    Dislikes Received:
    0
    Trophy Points:
    3
    Location:
    Hong Kong
    I don't see noise from the power supplies being injected into the audio output, I see ground hum at 60Hz and mostly odd harmonics.
    1. The SMPS ripple like crazy (relatively speaking) at about 500Hz (2ms between peaks), this makes a blip in the measurements and only at 500Hz
    2. The Jay's Audio LPS has significantly more ripple at 60Hz (16.7ms between peaks), yet has as low 60Hz measured in the output as the other LPS.
    3. The U8001A LPS has almost the same 60Hz hum as (2) despite no ripple at the frequency.

    Of course ground hum isn't a pure 60Hz tone https://en.wikipedia.org/wiki/Mains_hum
    But it's interesting that the PS-III coupled with the U8001A puts out much less of the upper harmonics of the 60Hz hum than with the other two PSes. But do you guys see the relation between this and the actual measured ripple waveform of the PSes? I don't.

    I just suspect that taking measures to directly address ground hum will result in better measured results using any of the PSes here than measured so far.
     

Share This Page