[Subjective Benchmark Part 1] Op Amps comparison in voltage gain stage


Disclaimer: This article represents my impressions on the different operational amplifiers tested, for a given topology.

The operational amplifiers, "Op Amps", were tested in a voltage gain stage of a DIY headphone amplifier / pre-amplifier, the WHAMMY. See the Diyaudio page for more information, including schematics.

My WHAMMY underwent several iterations of optimization until I reached a performance ceiling and felt it would represent a good host to evaluate various Op Amps.
See my WHAMMY build log for more details.


The OP Amps tested for this first round are:
  • Burson V6 Vivid*
  • LM833N
  • LME49720NA
  • OPA2107AU
  • OPA1642AID
  • OPA2156ID
*I wanted to thank Carlos at Burson Audio for sending a free sample in exchange for my honest review.

I swapped between all Op Amps in specific order to confirm the character of each and avoid bias, and allowed for sufficient burn-in time.
Directly to the point, see below a performance overview, summary of my extensive subjective evaluation:


The higher the score, the better.

Those ratings are the result of a 1 month comparison while keeping the same source/headphones and only switching Op Amps.
Note: While those ratings converged and my initial impressions got confirmed time after time, I am still doing some listening comparisons and might (lightly) update those graphs. 

If you ever wondered why some components, including Op Amps could sound noticeably different despite measurements and specs showing inaudible amounts of distortions and absolutely flat frequency response, my take on it is that music is of very different nature compared to static/generic test signals, and dynamics/transients are handled in very different ways depending on the circuit topology. Subjectively, one can focus on wide band attack/transient/leading edge of instruments (same for decay), and see how they are colored/reproduced. Acoustic guitars and drums are excellent "tools" for this. Note that even with MP3 320kbps, those differences are still noticeable.

I find it enlightening to look at below micro-dynamics = f(Hz) figure to understand the coloration/tonality/strength of each Op Amps:



How to read this graph: 
  • Best performance is as high and as flat as possible.
  • Too low = details are smoothed over: boring and uninvolving presentation.
  • Flat = neutral natural sound signature
  • Increasing curve = bright sounding; V shape = recessed mids, laid back
  • If a component adds extra artificial content/details, the rating will be lowered. (pretty subjective, so hard to position exactly).

Applied to the Op Amps tested, we get:
  • V6 vivid and OPA1642 are the closest to neutral (flat), with V6 vivid being more resolving across the board, especially more articulate in the Bass, and with a slightly darker sounding presentation.
  • OPA2107 has an emphasis on the upper treble, without too harsh discontinuities between Bass/Mids/Treble which still makes it sound pretty natural.
  • LME49720NA has a laid back presentation where the mids are recessed and the treble sounds itchy, yet lacks extension in the last octave.
  • LM833N had a pretty weak bass slam/details, but otherwise managed to keep a pretty neutral sounding signature. A bit on the warm side.

I find this graph to be a powerful tool: it allows to select component with specific tonality for best system synergy. It also shows that any component is full of compromise. For example, one should avoid associating two bright sounding components otherwise the combined system might sound thin and too bright, lacking body.

For example, this graph tells me that I should pick OPA2107 if my headphones/speakers lack top end extension / are too dark sounding.


I put below my raw notes taken during my listening sessions, in case you wanted more details:

  • OPA2107AU sounds laid back, and is very impressive at soundstage and treble reproduction, but fall short of sweetness/delicacy in High Frequencies, due to its unsmoothing nature. Still, it's a non fatiguing Op Amp thanks to its recessed mids. Mates well with Bifrost MB as it preserves its excellent low level micro dynamics and attacks/decays. OPA2107 slitghly lacks bite in mids to have then a bit more forward presentation and more at the level of treble in terms of raw definition/micro-dynamic (it's basically a laid-back V6 vivid). OPA2107 sound signature reminds me a bit of the sabre DAC presentation, very fresh and a touch bright sounding, but without the usually associated treble glare (over-exaggeration of treble details). Despite its V-shaped tonality, the OPA2107 remains very coherent throughout the whole spectrum without any harsh transition, giving a quite natural presentation. OPA2107 has a great sound stage with one of the best depth, but its imaging is a tad unfocused, as if there were some phase issue.
    I rated OPA2107's treble a bit above the V6 Vivid because there were times where I could hear more treble details (despite OPA's coloration), and there was also a bit more air and delicacy in its delivery, while V6 vivid was brutally honest, a touch rolloed of in the top octave to my taste, and with a slight hint of harshness in comparison. Might be related to the great mids/ low treble micro dynamics of the V6 vivid.

  • LM833N has a bit raspy/forward mids, which got almost completely solved with burn-in. Otherwise quite (too) relaxed bass and treble, lacking soundstage and definition/texture. Its treble is a touch sweet, smoothing over details and loosing in transparency/definition by lack of extension, creating a slight veil. Actually after further listenning, this lack of extension mostly appears on the transients, which makes it a bit uninvolving and muted. Independently of the dynamic aspect, I also noticed low level treble content tends to get muted/dulled out, while higher levels are comparatively better rendered. This changing sound signature depending on level/dynamics makes it difficult to give ratings as it depends on the material played. Overall, warmer sound signature than OPA2107 due to rolled off top end extension/details. Sounstage is LM833N's weak point, imaging is nto too precise, but quite stable, but soundstage depth is pretty flat. Involvement with this Op Amp is pretty poor due to lack of dynamics in the extremes of the spectrum. It also tends to get messy with too complex/busy songs.

  • LME49720NA. Treble immediately feels more extended than LM833, almost as much as OPA2107, but the mids seem recessed, close to Artera+ signature where treble (here, low treble) had more micro dynamics than the rest of the spectrum, leading to lack of texture/meat in the mids. A bit too dry for my taste. complete lack of P.R.A.T / emotion. Stimela trumpets sound much less textured/ dynamic than OPA2107. Does nt sound very delicate nor intimate. Bifrost mids micro dynamics and ability to pop small details is getting wasted with this Op Amp. High mids / low treble is a bit harsh/itching and lacks subtely, as if it was ringing (it's not). Transients, especially low level ones, sound too bright, and lack the overall leading edge (too smoothed), leading to uninvolving presentation. LME49720NA's relatively high composure score is due to this recessed and poorly-contrasted/dynamic mids, which then naturally gives an unmessy presentation, a small silver linning for this unnatural presentation. I had the same observation when putting this Op Amp in a unity-gain preamp part of an integrated amplifier years ago.

  • OPA1642AID's treble is less extended than OPA2107, but is less bright and has a more natural/neutral tonal balance (up to system synergy). Its micro dynamics are more consistent across spectrum than say OPA2107/LME49720NA, (trumpet sounds especially good) which makes it sound a touch wet in comparison to other cooler sounding Op Amps. Sounds like a less-extended/defined on top end, a touch more scratchy/itchy than OPA2107, more so compared to V6 vivid. Imaging is extremely good, with very precise instrument location and contouring, but somehow sound stage is too forward/flat, a bit claustrophobic. Norah Jones voice is a bit too much in your face on "Painter song", and also a bit harsh. Treble (cymbals for example) can sound harsh/metallic at times (still smoother than LME49720NA). I attribute this to the treble / low treble being a bit too present/dynamic compared to the mids. Overall, the OPA1642 gives decent micro dynamics, and close to neutrality thanks to its good homogeneity across whole spectrum. That being said, it lacks the Bass slam that OPA2107 and V6 vivid have.
    OPA1642 and LM833N are the most most difficult to differentiate, with OPA1642 being a bit more analytical.

  • Burson V6 Vivid - fresh out of the box, sounds very defined, great micro dynamics especially from low treble and above. Mids lack a but of lush/fluidity, upper bass is tight and precise, but does not dig as deep as OPA2107. SoundStage depth is already great, as is imaging and textures. High notes on trumpets/ violins tend to be a touch harsh. All issues mentioned are now fixed after sufficient burn-in. Low mids sound ever so slightly recessed, but very close to neutral. Feels like all notes start and stop faster, giving very defined attack and decay with better control yet still a more natural / fluid sound, which is impressive as those two aspects are usually a compromise. While V6 vivid imaging is great, its biggest strength is the sound stage depth that is noticeably deeper, and seems almost to be floating on some recordings. V6 vivid has better SoundStage compared to OPA2107 which sounds a bit flat in comparison. Micro dynamics is definitely more present across spectrum, when OPA2107 was at similar level only in treble. Imaging is great, especially on the extreme left/right. V6 vivid takes the great treble micro dynamics of the OPA2107 and extends it to the mids and even upper bass. This gives textures and micro dynamics and much more natural character as the frequency balance remains consistent whatever the material played, which helps to keep a natural tone. Bass is tighter and less bloomy/woolly than OPA2107. Composure is so well rated because of the fact that even when lots of instruments are played at the same time, their own tonality and positioning/contouring on the sound stage are preserved. After more than 2 weeks of listening, what shock me the most is the V6 ability to render a tone/atmosphere specific to each albums/recordings, while there are other Op Amps that impose their signature and make all track sound similarly colored. Odd things were noticed with the V6 vivid: nasal-like tonality on some songs that I notice only now with the V6 vivid. Switching back to the (yet) mids-recessed OPA2107, this tonality was indeed there, but I never noticed it as much. Same happened with different sonority on different songs.
    After all this praise, it's also fair to mention some downside: the low treble gets just a bit aggressive at times, which is not the most relaxing experience.

  • OPA2156ID - getting too hot! Listened 10 seconds before shutting down. Probably oscillating or too much voltage (40V slew; running at +/-16.2V)? Sounded like an OPA1642 with with a litle more mid-treble micro dynamics. SoundStage was laid back. After a second try, micro-dynamics were clearly sub-par, sounded rolled off. I think this amp is either broken or oscillating. To be checked again with second sample. No conclusion yet for this one then.

________________________________
CONCLUSION
________________________________

Ranking based on performances only:
  1. Burson V6 vivid
  2. OPA2107 
  3. OPA1642
  4. LM833
  5. LME49720

When price/performance ratio is taken into account:

  1. OPA1642 (2.2€)
  2. LM833 (0.8€)
  3. OPA2107 (21€)
  4. Burson V6 vivid (70€)
  5. LME49720 (2.4€)
It is actually disappointing that the performance follows the price tag.
I was hoping to get great performance from low cost and greatly specc'd IC op amps (to lower the B.O.M on my future projects). Maybe in next round.

The Burson V6 vivid is simply put a terrific performer (within this circuit at least), and should be tested if one wants to max out his system quality.

Potential candidates for the second round are:

  • Burson V6 Classic
  • AD8599ARZ
  • OPA2156
  • OPA1612
  • OPA1656
  • LME49990MA 
  • SparKos SS3062
  • LME49860MAX/NOPB
  • 2X LT1358 or LT1128ACN8

Feel free to suggest other Op Amps by comment/email.




Comments

  1. would be interested in OPA1656 performance, it's what I currently use. cheap and very good performing to my ear, and on paper.

    seller "mindofyou" on eBay also has rare LME49990MA for sale. supposedly very good.
    bought two of these to roll in my burson swing amp.

    ReplyDelete
  2. Thank you, Verifonix.
    I updated the list of candidates for the second round.
    I'll also update the page with a table of specs and topology for each Op Amps, which might help to draw some conclusions after sufficient data.

    ReplyDelete
    Replies
    1. cool!

      PS I got mail about an update https://martigane.blogspot.com/2020/02/objective-benchmark-part-1-op-amps.html

      this was by mistake? :p

      Delete
    2. Hi,
      No, it was not a mistake :D, I was just building the blog with this (incomplete) article, but still lack the measurements.
      More important, I first want to perform the blind test, but I need to wait for my quiet lab construction to be done, and my colleague to swap Op Amps for me :).

      Delete
    3. The Measurement of each Op Amp is now online!
      https://martigane.blogspot.com/2020/02/objective-benchmark-part-1-op-amps.html

      Delete

Post a Comment