Serious distortion

Discussion in 'Measurement Techniques Discussion' started by Serious, Jul 29, 2021.

  1. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,955
    Trophy Points:
    113
    Location:
    near Munich, Germany
    What makes you think it's not microphone noise? Looks exactly like we are hitting the noise floor to me. Easy to confirm via an FFT anyway.

    For transducers I've seen which had high higher order distortion at lower levels, the level still didn't quite stay constant. Even if it just went up by 5dB between 70 and 90dB like on the HEDD Type 07 here.

    Would be fun/interesting to build a flat plate coupler with like 35 PUI 5024HD 10mm microphones arranged in a 5x7 grid (roughly ear-sized) lol. Combined noise floor could be in the 0dBA neighborhood. I'm sure the comb filtering and directivity would be interesting aswell, haha.
    Nowadays we have transducers with like -100dB THD at 100dB through DSP and acoustic feedback magic, so something crazy like that may even be necessary to gather additional insight in that case. It's something I've had in the back of my head as a project for some time now - might try to build it within the next months.
     
  2. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,662
    Trophy Points:
    93
    Because I can still get better results with HD650. If I hit the limit, HD650 wouldn't be any better / lower in the mids/highs, but it is.
    upload_2021-7-29_18-18-10.png
     
  3. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,955
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Looks identical to me. Level at 1kHz is the same, 2kHz is 10dB lower. So in the distortion plots we get -80dB at 1kHz and -70dB at 2kHz. Of course you'd get an inverse FR pattern for the noise floor. HD650 is just much flatter.

    What I don't really get is why the noise floor doesn't get lower (relatively) at higher SPLs like we see on Tyll's graphs for example. Perhaps it's something other than the microphone limiting SNR. Or maybe it's even a software limit you're hitting.
     
    Last edited: Jul 29, 2021
  4. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,662
    Trophy Points:
    93
    It's because of the way I am presenting. The distortion in db is relative to the highest level (104db). In a sense, it's absolute level of distortion. This is to get away from plots showing "higher" levels of distortion at lower volumes, which I think is more confusing.

    If you think HD650 and HE560 distortion look the same (with respect to noise floor), you are simply being obtuse.. Here's a different example, more drastic than HD650, against the IE900 IEM. Also, I never said drastically different. The IE900 is about as low as we can go here I think.

    upload_2021-7-29_22-10-27.png
     
    • Like / Agree Like / Agree x 1
    • List
    Last edited: Jul 29, 2021
  5. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,955
    Trophy Points:
    113
    Location:
    near Munich, Germany
    That makes very little sense to me. So you're saying distortion at 90dB is actually 14dB higher than plotted? If so, I'm surprised I just found out about this now. And if so it would make even less sense that it stays constant now if it is not noise, since what you're saying is distortion actually decreases with level. I've never, anywhere seen distortion decrease with higher levels with transducers in such a broadband fashion.

    I seriously don't think so and think you should either present them as absolute values (the whole plot with the fundamental above) or take an FFT to make sure.

    For the IE900 it makes sense aswell. Your measurements show more of an upper midrange emphasis, so why would you not expect a higher SNR in the 2kHz region? Same for the bass region. Take Tyll's SR-009 and NightHawk data I attached. Both look almost entirely noise limited to me. With the exception of the R channel in the Nighthawk measurement sub 100Hz. Just the SR-009 has an emphasis at 1kHz, the NightHawk a dip.
    Or are we talking about the roughly 5dB difference at 900Hz? In that case, no clue. My guess is the levels weren't exactly matched, but we may be seeing the HD650 distort here somehow? Different measurement sweep lengths?

    However if you're really getting -76dB D4 at 104dB* then something is likely wrong with the HD650. I'm getting -95dB D4 with my HD600 at 104dB and even then I'm surprised it's not lower. Easy to confirm with a quick 1kHz sine FFT anyway, no need to write another post and pull more existing data. The higher orders should be extremely low.
    *(and -62dB D4 at 90dB if I understood correctly!)

    (For example I am getting about -105dB D4 with the AirPods Max at 104dB on an FFT. Possibly my HD600 mods increased distortion.)
    As for the plasticky timbre, what about the construction of the headphones? Maybe it's a less acoustically dead plastic they're using. You have an accelerometer from your speaker building days? Maybe it's just the diaphragm material after all.
     

    Attached Files:

    • Like / Agree Like / Agree x 1
    • List
    Last edited: Jul 30, 2021
  6. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,662
    Trophy Points:
    93
    Yes precisely. The microphone I have will not do any better than about -75db-80db at best for D4 (the noise floor). If you can get 100db SNR out of your microphone, then all the credit to you! Keep in mind that I do not adjust gain to max out SNR as part of workflow efficiency - I think max SPL is 120db so there is a lot wasted. I know, everything you have is the best. You have the best microphone that has 115db SNR. You have the best ambient levels in your place quieter than an anechoic chamber.

    The traditional way to plot would end up this way where distortion appears to be higher at lower levels in certain areas (when it really isn't). The distortion signal is higher as a percentage or ratio, but lower relative to the highest listening level / ambient.

    upload_2021-7-30_16-21-28.png

    I prefer the other way. There are pros and cons to each.
     
    Last edited by a moderator: Jul 30, 2021
  7. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,955
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Why make a big fuss out of it? Currently I am using microphones with worse SNR than yours to arrive at the figures above and we both know how to get measurements below the noise floor. Why you've gone to type up this response instead of taking 2 30 second FFTs, I don't know.
    Surely the HE560 will have way lower than -70dB D3 and D4 at 104dB at 2kHz. What I was trying to get across is that -90dB past D2 and D3 isn't all that special for headphones in the midrange even at insane levels.

    Either way I rest my case. I still think D3 and D4 are noise limited in the midrange for both the HE560 and HD650 plots until you've proven otherwise via an FFT or a FR and distortion plot.

    It's true I want to build something with lower noise for streamlined headphone distortion measurements, but that's only a project for the future. My LCT 550 nets me a similar SNR as my smaller capsules for headphones since the driver physically can't be put as close to the capsule and there's no seal. I use it because I still haven't fixed the distortion I get with my smaller capsules. Your rig is clearly better than mine. Anyway, the other thing is far more important to me...


    If you're obsessed with the graphs looking pretty to the point where you're throwing accuracy out of the window (why even put a scale if it is wrong for 3 out of 4 plots you're plotting?), then why not just use ARTA's automated multitone distortion plotter like Rin Choi used to? Combines accuracy and pretty graphs with low noise floor.

    To be honest I'm furious you're only mentioning this now, or maybe I overlooked a post where you explained what you do. This is the kind of stuff where a disclaimer has to be put under every plot. Lower levels lowered to match noise floor, something like that.
    You constantly make fun of Amir and Jude, yet you manipulate your graphs like that? I wouldn't be surprised if Amir would ridicule you for what you've done if he got wind of it. This is not a good showing and I'm really not sure what to think now. Judging power handling from the graphs is futile as is without mental gymnastics.

    I feel personally affected since I compared distortion at 90dB on your plots across different sources and had no idea I had to add 14dB! This way people end up thinking the HD650s bass distortion is way lower than it actually is, for example. And that things go really haywire when cranked up, when the reality is far less drastic.

    I know I need to calm down, but I think it's ridiculous. The fact that you've posted plots like this for months and nobody realized it is also concerning. I'm even disappointed in myself that I didn't question the weird looking plots myself until now.
     
    Last edited: Jul 30, 2021
  8. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,955
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Also why are you bringing the ambient noise discussion into this? It's like you want to stir up a debate when all I wanted was to see an FFT plot comparing the HD650 to the HE560.

    My narrative was never that my ambient levels were especially low. My narrative was always that generally ambient levels in normal suburban living conditions and with closed windows are lower than what can be easily measured with small omni microphones. As in less than 30-35dBA.
    Put another way, if the sound of the click of a smartphone power button isn't masked by ambient noise, you're good. That sound is roughly 10dBA according to my measurements, depending on the phone. The noise floor of the Vali with an HD600 should be in the 10dBA neighborhood aswell. Same goes for rubbing two fingers together with your arms stretched out. You've mentioned you can hear the Vali noise floor, so your ambient noise floor can't be as high as you said it is.
    Besides the noise floor in my room in the summer is limited by my running AC anyway. I get a lower noise floor wearing the AirPods Max in noise cancelling mode :)
     
  9. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,662
    Trophy Points:
    93
    Here, you are proven otherwise:

    RD modded HE560v4
    4kHz tone at 102dbSPL
    FFT size 32768
    upload_2021-7-30_19-8-32.png

    HD650
    4kHz tone at 103dbSPL
    FFT size 32768
    upload_2021-7-30_19-48-9.png

    Hint: just politely simply directly ask for an FFT for confirmation instead of being an obtuse ass.

    No one else has a problem with asking me directly for additional measurements or clarification without extraneous nonsense (e.g. wow, my microphone is so much better, you are wrong - prove it otherwise, your D4 results are shit, I can get better results, my room is as as good as an anechoic chamber except when the AC is running, Tyll's HATs is better than yours, etc.) It's like you have small dick syndrome. It's hard to really understand your request when it's wrapped with a brocade of shit which is your big ego.

    I'm here to serve. I've always been agreeable to going beyond when additional questions or measurements are asked for. Many times going the extra mile. I don't need to put up with your egoistical shit (this is how you come off if you are not aware). Remember, I've been doing this shit and see a lot more data from a lot more angles and a lot longer than you have*. You cherry pick measurements that suit what you want to believe and come to premature conclusions.

    One week off for being childish.

    P.S.

    *I already knew the answer. I admit to being a dickhead and stringing you along for the ride to see how you would react.
     
    Last edited: Jul 30, 2021
  10. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,662
    Trophy Points:
    93
    I actually wrote something about this a while back on how these graphs were produced and what they meant. I admit I just sort of rolled with it without further thought without realization that a note was necessary for each one of them. I think in my mind it was obvious because relative distortion (%) will not always be lower at lower volumes, and most cases exceed distortion (%) at higher volumes. But yes, I need to own it and I am owning it now.

    Yes, you need to calm down, hence your one week off. And yes, you should have realized this - because it if was pretty obvious to me, then it surely should have been to someone as smart and astute you. You don't need the histrionics and compare me to Jude or Amir.

    Perhaps a clearer presentation would be like this where absolute SPL is used on the Y-axis.
    upload_2021-7-30_19-35-4.png

    First, stop being a dickhead. I'm obsessed with graphs making sense. If I wanted my graphs to look good, I would present them like Oratory or even better RTings. We can discuss the merits or downsides of this approach.

    Just so you know Mr. Know-It-All, the multi-order distortion plotter in ARTA (LIMP) takes forever to run. I have a life. My house isn't quiet for hours on end. As I mentioned, workflow efficiency is a consideration. I do not do SBAF all day, and even if I could, this is not an efficient use of time.
     
    Last edited: Jul 30, 2021
  11. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,955
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I really hate to bring this up again and I really don't give a crap about the measurements now anymore anway.
    Still, now this plot shows roughly 10dB more D2 and 20dB more D3 and 20 to 26dB more D4 than your other plot for the HE560v4** :confused:

    Tbh I think the absolute SPL graphs for the distortion are the least confusing. I quite like them and they're the ones I would roll with in your case. I would not use the graphs with the relative scale for 104dB where the scale is wrong for the other SPLs.

    I see you've gone and uploaded other measurements in the same way, so thank you for that. **I haven't gone back to re-check those numbers for the updated graph you've posted of the HE-560v4 and I won't bother as I don't really care anymore. It's likely the new numbers would be different.

    You mention I cherry pick measurements. Which ones do you mean specifically? I uploaded the AirPods Max measurements despite the treble looking rougher than it should, same for my Focal Utopia measurements. I do tend to position headphones in a way I would use them when measuring them on my head instead of averaging multiple positions, however.
     

Share This Page