Blind Virtual Audition of Several Headamps

Which amp provides the presentation that you like the best ?

  • A

    Votes: 1 5.0%
  • B

    Votes: 7 35.0%
  • C

    Votes: 1 5.0%
  • D

    Votes: 1 5.0%
  • E

    Votes: 2 10.0%
  • F

    Votes: 4 20.0%
  • G

    Votes: 2 10.0%
  • H

    Votes: 2 10.0%

  • Total voters
    20
  • Poll closed .
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Founder of XSA-Labs
Joined 2012
Paid Member
Update Jan 8, 2017 - person who identifies the most clips with its respective amp gets a free Pocket Class A Headamp as a prize. Details here.


I have collected over a dozen headamp builds over the past few months, always in search of the ultimate sounding one. I frequently get asked via PM, which one is my favorite, or which one do I think is the best. Following a similar approach that I took over in the Fullrange Speaker forum for blind testing of speaker drivers using mp3 files recorded with the setup, I present here, the recorded sound clips of 8 different headamps for you to listen to.

They were all recorded using the same setup, same source, level matched to within 0.5dB, and using the same 270ohm resistive load meant to simulate my DT880-250 headphones.

The signal was provided and recorded by a Focusrite Scarlett Solo 2G audio interface running at 48kHz and 24-bit wit the conversion to a 320kbit mp3 file done by Audacity.

The clip is an mp3 of Anne Bisson's September in Montreal (45 seconds long).

So the idea is that everything in the chain is constant except the DUT (device under test). So whatever you hear that sounds different has to be the result of the changes in the DUT.

Here is a photo of all the amps lined up on the bench getting tested.

597277d1486362657-blind-virtual-audition-several-headamps-amp-comparo-setup-photo-1.jpg


From left to right (far left box in red is Focusrite Solo 2G audio I/O unit):

1. Agdr's Super CMOY 1688,
2. my Pocket Class A,
3. a brand new Fiio A5 (Muses02/LME49600 based),
4. Aksa's CFP with BF862-2SA1837,
5. SRPP with 4 pairs of BF862,
6. Salas' DCG3,
7. Aksa's CFP with BF862-BD140, and
8. Juma's BF862 preamp paired with an LU1014D source follower with an IRFP240 cascode.

Most of the amps can be found in the following threads:

http://www.diyaudio.com/forums/solid-state/301164-mosfet-source-follower-headamp.html

http://www.diyaudio.com/forums/head...f862-based-se-class-headamp-without-heat.html

http://www.diyaudio.com/forums/head...y-2x-9v-real-ground-headphone-relay-pcbs.html

http://www.diyaudio.com/forums/head...llower-headphone-amplifier-3.html#post1130743

The sound clips are labeled only as A, B, C,.. H and will remain anonymous until the poll is over. The poll just asks you the simple question: "Which amp provides the presentation that you like the best?"

Also included is the reference mp3 source used for the test.

You need to change the extension .asc to .mp3 to listen.

Have fun...

EDIT Jan 7, 2017: Please use the sound clip for B found here - the one in Post 1 has an error (clipping artifacts).


Spoiler Alert: answer key here.
 

Attachments

  • Amp-Comparo-Setup-Photo-1.jpg
    Amp-Comparo-Setup-Photo-1.jpg
    85.6 KB · Views: 4,423
  • Amp-A-Clip-1.asc
    1.7 MB · Views: 318
  • Amp-B-Clip-1.asc
    1.7 MB · Views: 221
  • Amp-C-Clip-1.asc
    1.7 MB · Views: 212
  • Amp-D-Clip-1.asc
    1.7 MB · Views: 186
  • Amp-E-Clip-1.asc
    1.7 MB · Views: 176
  • Amp-F-Clip-1.asc
    1.7 MB · Views: 201
  • Amp-G-Clip-1.asc
    1.7 MB · Views: 173
  • Amp-H-Clip-1.asc
    1.7 MB · Views: 178
  • Headamp-Compare-Reference-Clip-1.asc
    1.7 MB · Views: 196
Last edited:
Inaccurate how, Weitian?
It doesn't say anything about how those amps would drive reactive loads;
It doesn't tell if the amps may oscillate or not;
If those amps are bass light and my amp is base heavy, it ends up being right, impacting music perception;

From the point of music, rather than sound, is it meaningful?
Definitely not for me, but we can disagree on this if you insist.
 
BTW, this is driving headphones, not speakers. The reactance issues, excepting electrostatics, are not so important.

Not true, there's quite a few VC headphones that have non-negligible reactive elements in their impedance. As it happens these models are also the ones that headphone audiophiles tend to say 'benefit' from a headphone amplifier and like to buy and do their system component rolling on - Sennheiser for example make a few models like this.

I wouldn't know really whether you'd classify this as a problem of headphone design or a problem of amplifier design, but nevertheless it keeps the never-ending whirlpool of headphone audiophilia going.
 
BTW, this is driving headphones, not speakers. The reactance issues, excepting electrostatics, are not so important.
Cheers,HD
Yes and no.
Better load would be reactive load network/headphone simulator or pick off record signal across actual headphones fitted to some kind of dummy head.
I am more concerned about level control post processing at limited bit depth causing errors obscuring any actual differences.

Dan.
 
Founder of XSA-Labs
Joined 2012
Paid Member
What I will be hearing, is the effect of your amps on a pure resistive (270R) load, super-impose on my headphone amp driving my headphones.

Rating amps by such virtual audition is inaccurate, to say the least :rolleyes:.

When I did the first (of 6 separate Subjective Blind tests) of full range drivers, I got many many negative comments like this. "Useless", "it won't work", "my headphones or speakers impact the sound", "my audio chain is flawed", "my audio chain doesn't use gold played contacts", "my audio chain doesn't use the oil-skin of virgin goddesses from Mt Olympia as the capacitor dielectric", etc.

If you look through this thread and the 5 others after it (different rounds for over two dozen drivers), you will see very similar negative comments from the typical naysayer with nothing to contribute but negativity. If that's the case you are most welcome to not participate in the poll.

http://www.diyaudio.com/forums/full...nd-comparison-3in-5in-full-range-drivers.html

I agree that it doesn't show you reactive loads or even different impedances. I could repeat test with 30R load resistor. I could add a 20uH inductor in series or a 10nF cap in parallel, etc.

But the fact remains that the sound you hear the difference to is purely the result of changes of the DUT has to remain (barring a file processing error on my part - perhaps file B is victim to). So think of it as what you are getting if you were sitting there and auditioning a purely resistive head phone (if one existed). I can think of several technologies that if made into headphones would be purely resistive.
 
Founder of XSA-Labs
Joined 2012
Paid Member
Yes and no.
Better load would be reactive load network/headphone simulator or pick off record signal across actual headphones fitted to some kind of dummy head.
I am more concerned about level control post processing at limited bit depth causing errors obscuring any actual differences.

Dan.

Before recording I used the live VU meter function in RightMark Audio Analyzer to set channel balance (left to right) to within 0.1dB, then I would play the clip and adjust level to make sure max signal is not clipping (using VU meter in Audacity), and is within 1.5 to 2dB of 0dB but never greater to ensure not clipping. I also monitor red LED clip indicator on Focusrite when recording. I set Audacity to 48kHz and push record button. I then play the track from iTunes and monitor the VU meter again as it plays and records to make sure no clipping and almost maximum extent of bit depth is utilized by not letting max signal fall below -3dB. I then select 45 seconds from known markers on track and clip to remove front and back tails not needed. Then use amplify function and set max peak level to -0.5dB. Then convert to 320kbit MP3 in true stereo. Then rename file with .asc extension. This is same process I used on my previous threads with addition of a better interface (Focusrite Solo 2G) and ability to precisely adjust left and right channel balance prior to recording.
 
Founder of XSA-Labs
Joined 2012
Paid Member
I looked at B and indeed there is a problem - it looks like I must have uploaded the wrong file but cannot change that until i get home later tonight. The file was supposed to be the discarded one due to clipping errors. So please don't vote yet.

Sorry for the inconvenience but it was late when I did this. Thanks MaxHeadroom for pointing this out.
 
Hmmm.
Right, so you are playing an MP3 file and recording it as Wav file and then using normalise to -0.5dB, then converting to MP3 again.
Your method is so wrong that I can't be bothered to continue, and explains why on quick listen that I did not hear any readily identifiable differences except for B file level problem.

Dan.
 
Founder of XSA-Labs
Joined 2012
Paid Member
Given 1.8MB max file size limitation, I could provide 8 seconds of lossless FLAC files and play from FLAC, save to FLAC etc but I think the differences between amps are indeed more discernible than the differences between 320kbit MP3 and FLAC. What are you using to listen - you need good headphones and a good headphone amp and a quiet space and time to do this. It takes concentration and is not something a casual listen on PC speakers while you do something else will work.

Have you ever taken the Klippel distortion audibility test? If you did, what rating did you get? If you scored -21dB like the majority of people, indeed, you may not be able to tell the differences as they are at the -40dB to -60dB levels. However, harmonic profile is another matter - it can be way below distortion audibility threshold, but is apparent in our enjoyment of the music and sensation of pleasantness vs fatiguing sound.

Listening Test
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.