Active Device (op-amp/tube) coloring and distortion discussion

Discussion in 'Digital: DACs, USB converters, decrapifiers' started by Hrodulf, Oct 30, 2018.

  1. Hrodulf

    Hrodulf MOT - Sonarworks

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    2,436
    Dislikes Received:
    3
    Trophy Points:
    93
    This has to be the most boring and unaudiophile signal chain out there. Yet, it works fine, as it seems.
     
  2. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    47,268
    Dislikes Received:
    74
    Trophy Points:
    113
    Location:
    Antarctica
    Yes, transformers are colored.

    You don't think opamps are colored? That the THAT1646 doesn't contribute to the DM house sound? Or that JFETs and MOSFETS are colored to bipolar transistors? Or that the 300Bs, 2A3s, and 45s are colored? Or that caps are colored? Or that power supplies color the sound? Or that your headphones, and all other transducers, are SUPER colored.

    Just pointing out that you seem have an unnatural obsession over transformers being more colored than other piece of active or passive part. It just seems like a looserish attempt on your part to discredit my impressions, by bringing up transformer "distortion", and now "coloration", because you are in love with what you currently own: the DM Source DAC.

    Just stop now because I'm pretty sure my classic metal Judas Priest or Black Sabbath records, (or Ella Fitgerald, Louis Armstrong, Nat King Cole) went through through a ton of transformers during the recording and production process. And they happen to sound a lot better (more immediacy, more real) than anything made in the past 30 years.
     
    Last edited: Oct 30, 2018
    Lyer25, bazelio, taisserroots and 6 others like this.
  3. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    47,268
    Dislikes Received:
    74
    Trophy Points:
    113
    Location:
    Antarctica
    One of my little tidbits:

    Every time I hear some fucker say "wire-with-gain", e.g. the amp is "wire-with-gain", it always almost means this: commits few sins of commission, but many sins of omission. It's a loaded phrase. Don't trust anyone who ever says that. It's obvious they haven't listened to enough stuff, or just have bad ears.
     
    Last edited: Oct 30, 2018
  4. Psalmanazar

    Psalmanazar Non Voting Friend

    Slaytanic Cliff Clavin
    Joined:
    Sep 27, 2015
    Likes Received:
    2,786
    Dislikes Received:
    96
    Trophy Points:
    113
    The illusion is perceived as more real than reality up to the point of uncanniness.

    Everything has a sound but ultimately it comes down to design, design philosophy, and the results thereof. Different manufacturers have different philosophies. Do you want to try to recreate the illusion already there or create a new one? This is one of my favorite posts on SBAF:
    https://www.superbestaudiofriends.o...fat-2-at-schiit-blog.3507/page-42#post-196264

    Late 80s Slayer has that SSL sound everywhere that sounds cool. Wanting to cover up it up or make the drums sound less Phil Collins reverb wouldn't make everything gel as well; it would kill the illusion of Slayer, to make it so you're listening to Slayer through something or Slayer playing on something rather than Slayer themselves even if it's not SLAYER "Evil Has No Boundaries" and "Angel of Death" Slayer. Slayer were so cool. People should listen to Slayer on big speakers at least. LOUD! This is SLAYER, not Starbucks music. Drinking beer, not Slaytanic Blood Red Raspberry Frappachino with complimentary 7" eagle cookie.
     
    Last edited: Oct 30, 2018
  5. Hrodulf

    Hrodulf MOT - Sonarworks

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    2,436
    Dislikes Received:
    3
    Trophy Points:
    93
    Tube circuits have coloration mostly due to the fact that tubes don't have complementary devices. Depending wether it's SE or PP, a circuit using the same tube will have vastly different coloration due to it's distortion character. I would go as far as to say that PP circuits don't sound too tubey because they have lessened even order distortion. At the same time transistors can be used in SE circuits to get a rather tubey sound signature.

    As for transformers, they have inherent distortion due to hysteresis at all signal levels which is most apparent at low frequencies. This is the main reason why transformers are abundant for high frequency applications and less so for audio which is much lower frequency. In pro stuff transformers for line level signalling are used almost exclusively in coloured mic preamps and other instrument equipment where coloration is actually preferable. Modern line level equipment has moved to active stuff, unless one absolutely has no way of providing power to these circuits.

    Do all of the parts in the quoted portion colour the sound? Sure. Does a wobbly light volume knob colour the listening experience, it might, if you obsess about it. I'd say that an expertly designed circuit that uses active silicon based parts would colour the listening experience less than ruminating about the coloration.

    I can't see an actual amp omitting signal. I.e. I'm not aware of the mechanics that would cause it to, unless we're talking bandwidth and/or amplitude limitations. Omission through commission is more like it, as the small stuff can get buried in random noise and/or harmonics-borne "noise" on spectrally complex signals. It's one of the reasons why I find jazz and other minimalistic music unsuitable or even detrimental to device playback transparency comparisons.
     
    Lyer25, SSL, Psalmanazar and 2 others like this.
  6. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    47,268
    Dislikes Received:
    74
    Trophy Points:
    113
    Location:
    Antarctica
    I disagree with you on everything.

    I've heard PP amps that sound worse than SE amps because of bad design or bad output transformers. It's really hard to compare because output power is different, therefore different applications.

    There is omission and there is commission. There is no omission through commission. We are talking about perception here. Just because you cannot see it on an averaged FFT, which is a simplification of the waveform for presentation / visualization reasons, doesn't mean you can't hear it. Our ears are more sensitive than decimated / re-presented data from measurements. Also, measurements only use steady-state signals. Perhaps measurements might work if we test our ears on steady-state signals - might be an interesting test to see if I can tell the difference between DACs with steady-state signals.

    This debate is existential in audio and why people still fuss about active devices vs. passive devices for the first stage of phono gain. Or (Nelson Pass / Jason Stoddard / Craig Uthus) vs. (Amir / Nwavguy / Doug Self) schools of thought regarding THD in amp design.

    All these DACs measure well below limits of human hearing, yet they sound different, with consistent agreement among owners and auditioners of their more obvious sonic qualities. Something's got to give, and that's going to be measurements, not repeatable human experience. The Convert-2 exhibits the same sound, whether I'm happy, pissed off, sleepy, drunk, etc.
     
    Last edited: Oct 31, 2018
  7. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    47,268
    Dislikes Received:
    74
    Trophy Points:
    113
    Location:
    Antarctica
    This is an easy test. I've heard some very high-end systems that make everything sort of sound the same, in a pleasant way. I've also heard systems that make different recordings sound drastically different. The advantages of the former are that less badly produced material sounds bad. The disadvantage is that everything sounds homogenized. For the latter, the huge disadvantage is that badly produced material sounds like shit (and shit could also mean sins of omission) - to the extent that you would just not want to listen to it. The advantage is that well-produced material sounds even better - they shine.
     
  8. Hrodulf

    Hrodulf MOT - Sonarworks

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    2,436
    Dislikes Received:
    3
    Trophy Points:
    93
    I'm not arguing that PP is better/worse than SE. My main point was that PP design cancels even order HD, so it ends up sounding in a certain way. If I would choose a tube amp, I would get/build a SET mostly because that's what I and imo many people associate with "tube sound". I don't think that it is useful to isolate parts from the circuit itself, when talking about the coloration. In other words - it can lead to imprecise conclusions.

    Is information loss due to noise floor or heavily summed distortion a sin of omission or commission? It is clear that the amp itself is adding something absent from the input and thus something gets rendered imperceptible due to that.

    Maybe we drink different things or amounts, but to me everything sounds really pleasant after a few cuba libres. Unless the sound is intolerably terrible in the first place. Hearing is easy to trick and I'll be the first to admit that. Back when I was with SW, we had a really vocal fan who had sent in his cans for individual calibration. After 6 months we release a new build and he's asking us what is this software for. Turns out he thought that we've modded his headphones and he could just use them like that. I mean - that's one of the whales audiophoolery stands on - ears are easy to trick.

    That's cool and the main reason I visit this place. Tell me if I get on your nerves.
     
  9. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    47,268
    Dislikes Received:
    74
    Trophy Points:
    113
    Location:
    Antarctica
    If we are talking about "adding something", that doesn't necessarily mean something is masked. That's what I call "FFT analyzer thinking": that bar which was music got covered by another bar which was distortion. How don't we know that the audio signal, the musical waveform, rides on top of the distortion waveform, rather than the distortion blotting out the signal? I'd say it's more the former. (This is also why a HUGE issue I have with measurements is that they rely only on steady-state signals).

    Let's do a thought experiment: instead of amps, let's think about even higher distortion devices in the chain: transducers. Consider the linear and non-linear distortion. The HD800 and Utopia are considered highly resolving headphones, but when we think about it, they are distorted messes in many aspects. Correcting linear distortion makes them sound less screwed up, but not more resolving. And in terms of non-linear distortion, they measure worse than most (less resolving) planars yet have more "plankton". Also, the HD800 has that screwy distortion peak related to the FR peak, but yet this doesn't seem to impede its ability to resolve low-level information in the mids.

    In terms of noise, we can't think of noise as nice line at -xx db. It looks like that because we've averaged 4,8, or 16 successive FFT measurements, with each one of those FFT snapshots having super long windows sufficient to capture 20Hz. Noise by its nature is random, it's not all frequencies active at the same time per the "line" on the analyzer. A structured low-level signal near that "floor" can easily break through that randomness for the ear / brain to pick up some of it.

    Don't get me wrong, it's not black and white. There is a correlation between distortion and "plankton", but there is also a correlation between distortion and a "clean" CSD from the mids to the highs. Heck, we might as well throw in driver impedance in there too as having some correlation. Why do the 150-ohm Sennheisers sound like ass (in terms of microdynamics and plankton) compared to the 300-ohm models? Same thing for 8-ohm vs. 16-ohm compression drivers.

    This convo should be moved to another thread. Don't know if there is a suitable one that exists already.
     
    Last edited: Oct 31, 2018
    Hrodulf and direstraitsfan98 like this.
  10. Hrodulf

    Hrodulf MOT - Sonarworks

    Friend
    Joined:
    Sep 28, 2015
    Likes Received:
    2,436
    Dislikes Received:
    3
    Trophy Points:
    93
    upload_2018-10-31_22-6-40.png

    That’s a good question - it was interesting to think about it. Let’s take a look at this measurement of 32 equal magnitude tones. You see harmonics bunching up at higher frequencies, yet signal amplitude remains roughly constant - i.e. the signal tones don’t “ride” on harmonics. Let’s say, at 10k a -90dB signal would’ve been completely burried.

    I’ve done some RTA measurement and, as you say, noise is a bubling ocean of needles, rather than some frozen line. If the signal would “ride” on the noise floor, then there would’ve been random summing every now and then, which I haven’t really seen. As in - for such a measurement the grass is the shifty part, but signal stays constant. And when you turn it up, it doesn’t poke out instantly.

    As for the headphone distortion thought experiment - AFR correction, wether electronic or mechanical always increase headphone and speaker resolution, as dips and peaks introduce spectral masking.

    upload_2018-10-31_22-32-58.png

    Of course, not much can be done electronically about the decay characteristics of a driver. If there’s a resonant peak, one can attenuate the energy there and have both a more correct tonal response, less local HD (less energy being pumped in) and a faster decay (less energy to dissipate at the same rate). I haven’t looked at HD800/Utopia CSD’s vs say LCD-2/3/4. I’d recon decay has a lot to do with microdynamics. Naturally, I don’t claim to have all of the answers and you are right that identically equalized the HD800 has more detail than a LCD-2.

    As for low vs higher impedance drivers, the difference should be voice coil geometry. In theory this stuff should change TS params and for some speaker drivers they do change, but there’s a whole lot which basically stay the same.

    P.S. It would be cool if mod could move this convo to some other thread.
     
  11. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,138
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    Done.
     
  12. OJneg

    OJneg The Most Insufferable

    Friend BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,722
    Dislikes Received:
    28
    Trophy Points:
    113
    Location:
    Grand Rapids
    You're posting frequency domain graphs. Marv is referring to the time domain idea when he says waves riding on one another.

    For example, the small high frequency waves will ride on the higher amplitude low frequency waves when they are summed together.

    If the high amplitude wave pushes the signal into the non linear region of the amplifiers transfer function, the smaller HF wave will be similarly effected.

    Your graph shows high amplitude, high frequency components which will obviously be effected by nonlinearity on their own.
     
    taisserroots and Hrodulf like this.
  13. OJneg

    OJneg The Most Insufferable

    Friend BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,722
    Dislikes Received:
    28
    Trophy Points:
    113
    Location:
    Grand Rapids
    Imagine biasing the amplifier into the less linear region (using DC instead of LF) in the first place and then running just the HF signal into the input. Same effect in principle.
     
    taisserroots likes this.
  14. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,138
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    The fundamental and harmonics do overlap. But the picture is in dB scale, and indeed harmonics 90 dB down from the fundamentals should have a small impact on the fundamentals themselves.

    Non-linear distortion on a multi-tone measurements should add to the noise floor as shown in the above plot. Indeed noise floor could be viewed as non-linear distortion that is uncorrelated to the source signal.

    Technically, I agree that frequency response corrections should increase resolution as long as it does not increase distortion or noise floor relative to signal. Perceptually it is difficult to say for me what the outcome will be as long as things are not way out of whack.

    As far as the decay characteristics of a driver, I believe a lot can be done electronically. What is not possible to correct is the inherent bandwidth limitations of a driver. If the driver can't play smoothly a certain frequency range, then there is not much to work with.

    Those are good points IMO. However, I believe those issues will also show in the frequency response plot of a particular device tested using the same high power LF + low power HF signal stimulus.
     
    Last edited: Nov 29, 2018
    taisserroots likes this.
  15. maverickronin

    maverickronin Friend

    Friend
    Joined:
    Mar 17, 2018
    Likes Received:
    314
    Dislikes Received:
    2
    Trophy Points:
    63
    Location:
    Midwest, USA
    How complicated are we talking about? PEQ and simple convolution, won't do much for that. You need an earside mic like an ANC 'phone to make real progress with that sort of thing.
     
  16. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,138
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    PEQ and simple convolution will correct CSD (again, as long as things are not way out of whack).

    Whether the target response is "correct" or not, that's a different story.
     
  17. OJneg

    OJneg The Most Insufferable

    Friend BWC
    Joined:
    Sep 25, 2015
    Likes Received:
    3,722
    Dislikes Received:
    28
    Trophy Points:
    113
    Location:
    Grand Rapids
    I think it is erroneous to consider non linear distortion components as adding to the noise floor in any way. Noise floor of electronic systems will be psuedo-random wideband energy, while non linear distortion will either be related to our input signal via a harmonic multiple or intermodulation product.

    I would daresay our ears/ brains are better at ignoring the unrelated energy content as opposed to the related content, even if they are similarly low in level
     
    taisserroots likes this.
  18. purr1n

    purr1n Finding his inner redneck

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    47,268
    Dislikes Received:
    74
    Trophy Points:
    113
    Location:
    Antarctica
    I think we are talking about different things. Think waveforms. Say there is a low level 2kHz sine that is part of the recording. The gear produces D2 at 1kHz. For the sake for argument, this D2 is low, moderate or whatever. Will this small level of D2 obliterate the 2kHz which is part of the recording, or will the 2kHz which is part of the recording ride on top of the D2? Or maybe a little bit of both?

    I'm not suggesting the signal rides on the noise floor in the case. I believe the random needles do mask signal - they are they actually there. It's just what happens when the random needles are not there at certain points in time and over time? Are we able to hear signal just poking around this area past this on and off noise?

    Bingo! Non steady-state, transients, energy storage and dissipation. This at least would be a start.

    Recorded bursts at 50, 120, 500, 1200, and 4800Hz. Prototype software. Possible errors. Waveforms not normalized. Sinc interpolation filter not in place yet for 1200Hz, 4800Hz (code is there, too lazy to paste and integrate). The devil is in the details. Small things here make a big difference perceptually. There is more to measurements than FFTs than Amirites would have us believe.

    Ether burst.png
    hd800 burst.png
     
    Last edited: Oct 31, 2018
    Hrodulf likes this.
  19. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Friend MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,138
    Dislikes Received:
    16
    Trophy Points:
    113
    Location:
    Irvine CA
    I agree.

    It depends a little on the context. In DSP, LTI systems have a precise definition for what "linear" means. And both harmonic non-linear distortion, and noise fall into what is considered non-linear under that context.

    However, correlated and uncorrelated (to the source signal) distortion don't have the same properties. And further classification is used. Usually uncorrelated (to the source signal) distortion is thrown into the "noise" bucket.

    Furthermore, "noise" can be colored or uncolored. Colored noise may not be correlated to the source signal, but it would be correlated to itself, and that one may be easier to pick by our brains. A filter can fix that too though, but not w/o affecting the source signal, so compromises may be made.

    As far as what gets added and what does not get added, signal power adds.
     
    Last edited: Oct 31, 2018
  20. Mr.Sneis

    Mr.Sneis Friend

    Friend
    Joined:
    Oct 2, 2015
    Likes Received:
    853
    Dislikes Received:
    0
    Trophy Points:
    93
    Laymans take!

    The NE5532(4)'s and OPA(2)134's are about as standard as they come. I think all the pro gear I have seen use the opamps to great effect.

    My personal experience with opamp when rolling is that the circuit has to be precisely matched to the specific opamp and at my level, that's just not quite in my bag of tricks. I don't have a scope or take measurements, it's all blind rolling.

    When you get it wrong the opamps tend to oscillate and sound like crap or maybe they come close to sounding pretty decent but will fail in some other way. What kills it for me is when you can hear that crunchy ass treble or bloated midbass from the distortion.
     

Share This Page