How much does frequency response actually matter?

Discussion in 'General Audio Discussion' started by Lyander, Sep 20, 2024.

  1. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    91,905
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Good question. Science is fun. With typical sound reinforcement or DJ setups, it's almost always gonna be the woofers that start to sound poopy. Headphones could be different.
     
  2. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    8,613
    Trophy Points:
    113
    Location:
    Vancouver BC
    Would a quicker way to assess compression be how it's shown on the 'Linearity' graph here? Sweeps at set SPLs at some reference frequency, then overlay the curves to see any differences at other frequencies.
     
  3. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,724
    Trophy Points:
    113
    Location:
    Philippines, The
    LMAO right as I edited the above. Yeah Grados seem REALLY reactive to things which is part of what I love about how they sound, just my ears really dislike supra-aurals and the circum-aural Grados are NOT cheap.

    Possibly dumb question but you specifically mention heat which makes sense, temps would affect conductivity of things. Is this going to be similar to resistor ladder DAC warmup time where the effect may be negligible at first but more pronounced with longer time running, i.e. listening to the same song on loop for a while there *may* be perceptible compression later on?
     
  4. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    91,905
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Quicker, but taking FR sweeps may not reflect actual use of the VC's heating up.

    However, that linearity graph for HD620 seems fishy. I have never been able to get that kind of variance with any headphone measuring from 85db to 105db. Maybe it's specific to the HD620 (I'd still ask Solderdude to redo that measurement - it looks like he could be using an amp that changes volume via feedback, thus affecting output impedance). Any headphone that changes FR that much depending upon output level should be considered seriously defective.

    Here is R70X at 104, 100, 95 and 90db, but aligned at 1kHz.
    upload_2024-9-23_12-29-15.png

    upload_2024-9-23_12-30-22.png

    upload_2024-9-23_12-28-16.png

    Results seem consistent with the test tone method.

    I wouldn't say this is like R2R DAC warmup. When the voice coils get hot, things really start to sound like shit. Muffled, loud, loss of volume. It's pretty noticeable.

    BTW your question answered above - compression affects bass, at least on this dynamic driver.
     
  5. mitochondrium

    mitochondrium Friend

    Pyrate
    Joined:
    Sep 1, 2017
    Likes Received:
    1,151
    Trophy Points:
    93
    Location:
    A Cell
    I am not a headphone guy, but I use DRC in order to improve the SQ at the listening position. I started out using REW and now use acourate:

    https://www.audiovero.de/en/

    The guy behind this put a lot of effort into dealing with excess phase

    Mitch Barnett‘s book on DSP goes through the motions using acourate:
    IMG_5699.jpeg
    I am not even an idiot savant using acourate but the filters are better than anything I used before. Down side is it still is not a solution which requires only button pushing.
    I did not try another solution (like Dirac or Sonarworks) because the guy behind acourate is German and really helpful when you are a beginner. The mathematical manipulations acourate is capable of are probably also possible with FR derived from measurements of headphones, whether you can measure that FR using acourate or whether you need to import data measured with a different tool into acourate for headphones I do not know.
     
  6. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    8,613
    Trophy Points:
    113
    Location:
    Vancouver BC
    Here's a couple of even more extreme ones: Superlux HD681 AIR (semi-open), Sony MDR-7506 (closed-back): >2dB differences in the bass at 70dB vs 96dB.

    Might well have something to do with an impedance effect. The biggest linearity differences in all three are are around their impedance peaks (~100 Hz, ~35 Hz, ~55 Hz respectively). There might also be some resonance effect: that hump around 130 Hz in the HD 620S linearity (with the curves crossing over just right of it) is I'm pretty sure pad bounce.

    Doesn't seem to match with the (open) HD 560S though...
     
  7. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,724
    Trophy Points:
    113
    Location:
    Philippines, The
    Hmm these are all dynamic drivers so it makes sense cuz duh voice coils. Wonder if similar can be seen in planars, estats, even BAs and such for IEMs? Impossible to test cuz volume matching will be its own issue but that could be something worth taking into account when reading impressions, what folks' average listening levels are like. I'm mid-70dBA for regular stuff but when I want to be *immersed* in music I listen at 80-85dBA with some peaks nearly hitting 90 with high-DR music.
     
  8. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,699
    Trophy Points:
    113
    Location:
    near Munich, Germany
    No, that's the point. The attack and decay is given by the frequency response, which consists of the magnitude and phase.
    When "fixing" the magnitude, the PEQ filter introduces the opposite phase shift and you get a linear response with faster attack and decay. This is because it's all minimum phase.*

    I currently do not have any proper examples - despite going through many measurements. But this is a discussion I had with @ultrabike back in the day when I couldn't quite believe it. But systems theory tells us that it works.

    *This is also why minimum phase DAC filters make no sense. The resulting FR is very far from linear - even if the magnitude looks fine, the phase is all wrong.



    Now this is all theory. In practice there's no relevance to it. But what I mean is that the burst decay really doesn't tell us anything special. That's despite me initially believing I had something when I saw the slow STAX rise back then.
     
    • Like Like x 1
    • Dislike Dislike x 1
    • List
    Last edited: Sep 23, 2024
  9. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,724
    Trophy Points:
    113
    Location:
    Philippines, The
  10. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    91,905
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Exactly. I'm not interested in HF'esqe pontification. I am interested in the now and what is doable. Just do it. Go for it! Go into Matlab and do some analysis or the IRs. Then send me the PEQ that will make the R70X have the same transient attack and decay as the ESP950.

    Speak for yourself. The correlation to perceived differences are not as strong or as clear as say FR, but there is some stuff there (I think @Lyander is right on showing the actual waveforms). For example, I don't think the faster 50Hz ramp up of the Audeze compared to the R70X is in accident. I also don't think the overshoots of the initial half-waves of the 4.8kHz bursts with planar types is an accident either. Frequency response doesn't explain "fast" planars (it's easy enough to make to mimic the FR of the ESP950 with a slow dynamic like the HD650). I also feel there is something to the fluidity of planars and some behaviors related to the decays bouncing for a long period of time, albeit at a very low level.

    I mean from certain point of view, we can say that FR doesn't tell us anything special either. We just need impulse responses at a variety of different output levels.

    Anyway, you have been saying the burst stuff is useless to you for years now. I think it would be polite if you just not partake in such conversations because it gets kind of old. It's like you are shitting in my yard again. You've already made up your mind. It's like not I am going to convince you and vice versa.
     
    Last edited: Sep 23, 2024
  11. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    8,613
    Trophy Points:
    113
    Location:
    Vancouver BC
    Hmm - doubtful. Given the impedance curves of those headphones, the variation in amp output impedance with volume would have to be massive to account for the observed variations. For example, here's the HD 620S:

    [​IMG]

    At 100 Hz, Solderdude's results at his 70 and 100 dB reference levels differ by ~1.25 dB. To account for that by amp output impedance that changes with volume, an amp with near zero impedance at 100 dB would have to be ~54 Ohms at 70 dB (red line vs blue line in the graph above).

    Seems unlikely. Back to lurking for me! :D
     
    • Like Like x 1
    • Epic Epic x 1
    • List
  12. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,724
    Trophy Points:
    113
    Location:
    Philippines, The
    I do think there may be merit to that though! What's the impedance bump but just the natural resonance point of the driver array, iirc? Makes full sense that that'd be the center of any great discrepancy from heat.

    Makes me wonder about planars even more. Will the same effect manifest there, but because the freq v impedance thing is flat it'll just be the whole FR trace being off from where it should be by a fragment of a dB? Fun stuff!

    Also please you're making me look shameless with all my posts here if you go to lurking haha. Honestly just think this could be a fun conversation cuz I'm open to the idea of it all being FR, just that time domain elements are more compelling and track with what I hear.
     
    Last edited: Sep 23, 2024
  13. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,699
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Sorry, I was being an ass. And by saying that it's useless and immediately dismissing your work I did you a great disservice because in reality I do look at the burst responses you post. No, I don't take them as gospel that can perfectly describe what something sounds like, but it's another data point. Looking again your examples make a good point here.

    One reason why I was so mad is that I tried to dig in my measurements to make my point, but the truth is it didn't work and in the end I just didn't show any examples. In reality I could get the step response to improve by applying REW AutoEQ, but not to the extent that I thought would've been possible. And when I EQ by ear I would never use such steep filters in the top end, not even for speakers. That's just dumb, even if the measurement was perfectly accurate - move the microphone two millimeters and suddenly it's worse, not better.

    I don't have very many headphone measurements on my laptop, so I can really only give speaker measurement examples. And maybe the Hathor 3 way experiment measurements make for a good example:
    https://www.superbestaudiofriends.o...iy-cardioid-monitors.13292/page-3#post-426419

    Maybe there's also something to be learnt from looking at the Hathor vs Amon step responses, I'm not sure. The burst response will be similar, I think.
    There haven't been very many headphones that I found interesting since the Utopia or LCD-5. Maybe the MM-100 or the Moon Rover IEM.
    As far as measurements go I'm all about directivity maps and FR magnitude + excess phase graphs nowadays.


    As for the REW stuff: It only has the "ping" type of sounds, not the immediate start of the sine signal that you and me use for this kind of visualisation. In REW you can use IR and make a convolution with it with the sine burst files, but in the end that's as much work as just taking the audio recordings :)
     
  14. evanft

    evanft Acquaintance

    Joined:
    Oct 2, 2015
    Likes Received:
    51
    Trophy Points:
    18
    Location:
    Taylor, MI
    This is a really interesting measurement and I think we need more of this. I've had a theory for a while that one of the reasons why some headphones may sound more "dynamic" or "compressed" than others is down to differences in the level at which they're reproducing an input signal. My thinking is that more dynamic headphones are drifting more from the input than less dynamic ones, which may be causing an overall increase in perceived dynamic range.
     
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    91,905
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    It's probably these set of headphones that Solderdude found have severe amplitude linearity problems! I don't think he shows the linearity graph on all headphones right? Assuming he only shows them only for headphones that have issues.

    Of course we need more!

    I've had many ideas I haven't gone forward with! The APx555 in conjunction with a microphone and accurate microphone mic (still waiting for a small microphone preamp and ADC / DAC from @schiit), actually opens up an entire universe of measurements!

    Point is, frequency response measurement although almost everything, is also barely anything. It really depends upon your perspective.
     
    • Like Like x 3
    • heart heart x 1
    • List
    Last edited: Sep 24, 2024
  16. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,724
    Trophy Points:
    113
    Location:
    Philippines, The
    If this is a really subtle hint then please consider me unduly excited cuz I imagine that may make believers of folks in home studio circles yet. Everything from Arturia to Focusrite to Audient and even Topping now do have negligible changes to sound across mics, but I feel like if Schiit ever comes up with an interface that meets their standards of production it could be another PROPER market upset.

    I got my Arturia Minifuse 2 cuz it was super cheap on sale but when I tried it as an AIO and even DAC for my headphone rig... meh (despite me wanting to minimise number of boxes on desk). Nothing terrible per se, but meh. Too many SINAD/EIN-is-everything folks in that space.
     
  17. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    8,613
    Trophy Points:
    113
    Location:
    Vancouver BC
    It seems to be something he's been doing only recently, and then, yes, only on certain headphones. He suggests high D3 at low frequencies indicates compression at those frequencies with higher SPLs (eg in his Fiio FT1 review as well as the Superlux 681 and HD 620S), so he seems to focus on those ones.

    It might not correlate at higher frequencies and/or be an issue with planar drivers: in the Hifiman Edition XS, D3 is high around 4 kHz but he finds no compression there.

    An interesting test of this (for dynamic drivers) might be @Hands 's modified E-Mu Walnuts, in which he reduced the high D3 in the bass, vs stock. Or with different foam inserts in Custom Cans modded HD 6xx; hmm, maybe I should try that when I have some time...
     
    • Like Like x 1
    • Epic Epic x 1
    • List
  18. Lyander

    Lyander Official SBAF Equitable Empathizer

    Pyrate Contributor
    Joined:
    Feb 25, 2017
    Likes Received:
    11,724
    Trophy Points:
    113
    Location:
    Philippines, The
    I just had a random thought since I'd been on the Magni Piety for a long while but recently switched to the Unity for a bit to try and recalibrate to a "cleaner" sound cuz I've been working on some random mixes for a friend.

    The main thing that really stands out to me about the Unity remains how clean and refined-sounding the treble is which is genuinely shocking at this price point. Granted I just have the two headphones (HD600, YH-100) and two pairs of IEMs (Andromeda 2020, Wan'er Studio) I really use so mayhaps with significantly different gear I'd be thinking elsewise. After the treble though the next thing that comes to mind is how much more bass the Piety really adds to the Sennheisers (which is to its benefit, per my preferences).

    Immediate thought, ah the Piety measures horribly compared to the Unity so of course there'll be a mess of distortion down low making things sound richer! Well... yes and no?

    Magni PIETY Distortion Surface:
    https://www.superbestaudiofriends.o...urement-new-approach.13754/page-8#post-419485

    Magni UNITY Distortion Surface:
    https://www.superbestaudiofriends.o...iew-and-measurements.13965/page-2#post-420213

    NOTE: 300R load for the Piety measurements and scale up to 6dBu; 29R load for the Unity with scale "only" up to 0dBu so the Unity's performing very well even at a disadvantage here.

    Granted you'd think from eyeballing the graphs of the Piety that that's a crapton of distortion esp at D2, but that's mainly evident at lower-level signals. When you listen to really compressed junk like I do fairly often (I actually really like that new Linkin Park song, their second single with the new vocalist is turning my opinion around) -24dBu signal is pretty low, and that's not even considering how transducer distortion seems to effectively overwrite amplifier distortion (I need to turn my brain on to fully digest that, will be a few months :rolleyes:).

    Considering that, where the heck is all the extra low end on the Piety coming from? OI of the Piety is sub-0.6 ohms while the Unity's sub-0.1 ohms, but is that enough of a difference to matter? The Unity's also a spicier amplifier dunking 415mW into 300R whereas the Piety *only* dunks 400mW into the same load, but you'd think that'd mean the Unity'd kick more, no?
     
  19. Priidik

    Priidik MOT: Estelon

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    2,384
    Trophy Points:
    113
    Location:
    Estonia
    Here a typical alu-mg driver step response with it's signature breakup behaviour. One raw, other with PEQ notch.
    Kuvatõmmis 2024-09-30 152626.png
    Screenshot taken from the awesome HifiCompass review page of a BliesMa driver review.
    By minimizing an oscillation source, you will have modified nonlinear distortion mechanisms linked to that oscillation source.
     
    Last edited: Sep 30, 2024
  20. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    91,905
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    You are actually correcting linear distortion there. A 5.5kHz peak in FR means there will be oscillation at 5.5kHz (in the step or impulse response). If there is no 5.5kHz oscillation, than a 5.5kHz peak in the FR would not be possible.

    upload_2024-9-30_12-30-50.png

    The oscillation that is seen as corrected in the above step-response is actually the linear distortion aspect, not non-linear. Linear distortion doesn't mean it won't show up over time in an impulse or step response. In fact, it must!

    ---

    Also keep in mind the results are with a notch filter - that means eliminating all frequencies at 5.5kHz! And even with such a notch filter, the vestiges of the ringing at 5.5kHz are still there! upload_2024-9-30_12-38-53.png
    The driver just has a resonance point there.

    ---

    Finally, this shit for sure cannot be corrected with EQ. This is what is meant by non-linear distortion. Not easy to see this in an impulse or step response, hence spectrum.

    [​IMG]
     
    Last edited: Sep 30, 2024

Share This Page