B&K 5128 HATS measurement discussion

Discussion in 'Measurement Setups, Systems, and Standards' started by Serious, Jul 15, 2021.

  1. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,981
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I've been looking at the B&K 5128 measurements with great interest for some time to see what this new and supposedly much improved (compared to the previous generation) coupler has to offer for headphone measurements. So since Jude has only uploaded uncompensated graphs so far I just tried to come up with my own compensation curve. Since I am currently testing the AirPods Max and Jude's dataset is still very limited, I chose the HD800S and the APM for my target.
    Three tries later and my first somewhat rudimentary curve ends up producing measurements that look like this:

    [​IMG]

    So I think that's not that far off. I don't hear so much of a bump at 1.5kHz on the APM and I'd probably reduce the region between 200Hz and 1kHz a bit on the HD800S (but not the APM), but overall it could be indicative of what someone else hears.

    So I apply the same curve to the HD650 measurements:
    [​IMG]
    [​IMG]
    I think the results speak for themselves. I could of course make a 4th target which greatly reduces the upper midrange, but I don't hear the other headphones as that sucked out. Honestly, I wish I had done this earlier since now my interest in the 5128 is pretty much gone.
     
    • Like / Agree Like / Agree x 2
    • List
  2. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,981
    Trophy Points:
    113
    Location:
    near Munich, Germany
    For reference, here are my preliminary AirPods Max frequency response curves, taken with my in-ear microphone method:

    [​IMG]
    Subjectively I think the midrange is quite good, but otherwise sounds pretty U-shaped with some sharp sibilance. I like that the bass emphasis is centered on the subbass.

    Here's a link to my HD650 measurement with the same method (don't mind the upper treble dip, which is an artifact I didn't correct early on):
    https://www.superbestaudiofriends.o...area-measurements-by-serious.2518/#post-67054
    And here are comparable HD800S measurements:
    https://www.superbestaudiofriends.o...area-measurements-by-serious.2518/#post-67693

    EDIT: Upper treble dip is also not corrected with these AirPods Max measurements. These are completely uncompensated, preliminary measurements not even corected for the microphone FR (they are quite flat anyway anyway).
    Mostly to confirm the midrange tuning. And again, the in ear method tends to produce very uneven results past 4-5kHz so take that with a grain of salt.
     
    Last edited: Jul 15, 2021
  3. purr1n

    purr1n Super Friend

    Staff Member Friend BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    77,618
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Well, maybe the comp curve needs to push down 2-5kHz more? HD800S does not have as much 3-4kHz too.

    APM could be the anomaly. I've never met a coupler that gets every headphone right for every head.

    Maybe the 5128 is based on Ferengi ears with their massive pinna gain.
     
  4. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,981
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I'll see how much more data there is. There certainly aren't very many publicly available B&K 5128 measurements out there and Jude simply posted many of them to be buried somewhere in a long thread without properly linking to them in a database. Amir wasn't much better at categorizing and sorting them, either. Ugh. I know I saved some of them, but I don't have all of them yet. Gonna take a while.

    When aligned at 500Hz there was a roughly 3dB difference between the HD800S and HD650 from 1.5kHz to 3.5kHz in the InnerFidelity data. Here we now have up to a 6dB difference at 3.5kHz on the 5128. Subjectively I'm not sure the difference is that big.
    It's been a while since I heard both the HD800S and HD650, but extrapolating from how my modded HD800S compares to an HD800S and how my modded HD600 compares to an HD650 according to my measurements, I think the difference is closer to 3dB.

    Frankly, I still have hope for the 5128. I think it's quite unlikely the B&K engineers would release a turd, but from what I've read it also seems to be more difficult to work with and get a good seal with IEMs than earlier HATSs.

    Then again look at this! Here's a comparison between an HD650 and HD800S measurement on Jude's 5128 and on Tyll's HMSII:
    InnerFidelity with Serious 941 Target HD650 HD800S.jpg
    Jude 5128 HD650 HD800S custom target 05 2.jpg
    No, the target isn't perfect and I'd still like to tweak it here and there*, but...

    on the 5128 it almost looks like the HD650 is the brighter headphone! Yikes

    (Then again we are likely comparing an old black screen HD650 versus a new one straight from the factory here. But does that explain the big difference? IMO the HD800S should look brighter either way.)

    I don't like the look of the 3.5kHz dip on the HD800S and 4kHz dip on the APM with the 5128, either. Generally it looks more flat-plate-coupler-y to me somehow.

    *Notably for the 5128 both headphones should show more 1.5kHz and less 2kHz.
     
    Last edited: Jul 16, 2021
  5. Lurker

    Lurker Facebook Friend

    Joined:
    Oct 10, 2015
    Likes Received:
    119
    Trophy Points:
    33
    I mean, if you were to match them at 3.2 kHz then the 800S would be 7 dB above the 650 at 6k. Turns out the 800 is indeed the greatest headphone of all time, you guys are just listening too loud.
    Wer viel misst misst Mist.
    IMO the 5128 looks great. Better than the Kemar that has a huge gap in the skull that makes most over-ears not seal well. However I need a proper database to form an informed opinion. Would be cool if crinacle could spend his burger money on the base version without the torso. ZMF also has one.
     
  6. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,679
    Trophy Points:
    93
    HD650 is a bad reference since it has undergone revisions since Tyll last measured it. Also, there is the Jude factor. Do you actually trust that he knows what he's doing? I don't even think he runs good CanJams.

    We don't have enough data points from Jude. This is the quintessential problem and it will never be solved.

    The only reason Jude has the 5128, the latest and greatest, is so he can be taken more seriously and be an authority so he can shake down vendors who want to post at his site. Not because he wants to get the bottom of things. The 5128 is wasted on him because he doesn't do anything with it other than use if as a Head-Fi / Jude marketing device
     
    Last edited by a moderator: Jul 16, 2021
  7. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,981
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I started adding the Jude measurements to my database in my OneDrive folder that I linked in the Ambient Noise thread.
    https://1drv.ms/u/s!AlMZTB6OPpUMgZ9fo_iUOiFIzPvRlA?e=hbL98U
    I don't think many people realized this, but I uploaded traced versions of the InnerFidelity graphs there last year. The quality is pretty bad, though.

    Yes and that is pretty much the reason Amir had one aswell. But those two are AFAIK the only ones who have publicly posted 5128 measurements of multiple headphones. I know Jude has posted more than just a few measurements (from memory I'd guess roughly a dozen headphones or so) with the 5128, but I can't find all of them at the moment. Many are in his YouTube videos or somewhere in an appreciation thread on page 789.

    Fair point, but overall the curve just looks too warm to me.

    I know it doesn't help that I keep changing compensation curves, but I attached 5 individual curves now that I think are somewhat relevant. Clear Mg is like a clear with less of an 8kHz peak. Clear graph is available aswell, but more of a PITA to trace.
    I like how the Utopia curve turned out, but do not agree with the HD800S curve. Utopia curve looks quite similar to my own measurement actually, maybe a bit too warm. HD650 looks too lean to me and the seal seems too leaky. Clear looks about right, AirPods Max look too warm to me. I find them brighter than my modded HD800. Then again Apple could've changed the FR in a software update.
    I think I should raise the bass by 3dB or so on all of these plots. Doesn't look right to me.

    Honestly I still feel these couplers don't get the ear material right. Skin is actually somewhat absorptive and may be better approximated by a foam material. Which would ruin long-term reliability.

    EDIT: FWIW here's what the current target looks like:
    [​IMG]
    Doesn't look like it should look like that. The 4.5kHz dip is weird and the hump towards 8kHz also doesn't look very nice.
     
    Last edited: Jul 16, 2021
  8. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,981
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I realized I screwed up and the picture I posted above was actually Tyll's HD650 and HD800 measurements, not HD800S. Sorry!
    Here's an update with the InnerFidelity HD6XX graph instead of the HD650 graph to make sure it's at least a silver screen and modern production HD650. It looks like Tyll didn't get a good seal with the HD6XX however.

    Tyll HA HMSII HD6XX HD800S 941 Target.jpg

    Jude BK 5128 HD650 HD800S 06 Target.jpg

    I've mentioned that before, but I personally don't think 4kHz is recessed much on the HD800S and I think the dip is centered more around 2kHz rather than 4kHz like on a flat plate coupler.

    That the difference at 3-4kHz is so big and so small in the treble region between the two headphones is what I mentioned makes it look more flat-plate coupler-y to me. I'll see which other measurements I can find and which headphones. Honestly at this point what gets closest to what I hear is probaly the in-ear method up to 3-5kHz merged with the flat plate coupler or free air method past that.
    (Besides the bass emphasis and added sibilance the AirPods Max sound quite similar to my HD800 tonality-wise, much more so than the graphs would indicate.)
     
    Last edited: Jul 17, 2021
  9. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,679
    Trophy Points:
    93
    TBH, the $20 DIY flat coupler looks better than either IF's HATS (where there is too much difference at 6-7kHz, not enough difference in mid-treble) and the 5128 HATS (too much difference from 1-4kHz and, not enough difference in the overall treble):

    upload_2021-7-17_13-38-5.png

    We can argue the 4kHz dip on the HD800S measurement (ORA) should be 3kHz, but it's doesn't seem anywhere as non-representative as the comp'd plots with the IF and 5128 HATS.

    The Drop iFi specialty amp for the HD6XX leveraged Tyll's IF measurements and ended up with w/ 6kHz boost and major bass boost. It was awful. (The amp was good, the EQ was not). I had no idea how Drop even released it. Oh yes I do. They didn't have @CEE TEE there anymore.
     
    Last edited: Jul 17, 2021
  10. Serious

    Serious Inquisitive Frequency Response Plot

    Friend BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    1,981
    Trophy Points:
    113
    Location:
    near Munich, Germany
    If we're going to go there, I sort of prefer crinacle's GRAS 45AG graphs. In fact even @Crinacle himself wrote "while something like a B&K Type 5128 is arguably more accurate (though I would argue otherwise personally)".
    crinacle HD800S, HD650.jpg

    Heck, I may even prefer Hands' and my in-ear method measurements to both HATS and the FPC above (but not the GRAS). The AirPods Max result still puzzles me however. I really don't hear the midrange as an LCD-2C/LCD-4 or SR-007 tonality (compare crinacle's graphs, they're very close) and I hear it rather as an HD650 or modded HD800 midrange tonality. Also the B&K 5128 was the only coupler to perfectly mimic the minor 350Hz dip I get on my head. It doesn't even show up on the other HATSs.

    In the meantime I did this:
    Foam open cell ear.jpg
    I tried to cut out an ear from the open cell foam I use for my "v1"-style coupler. It even sort of fits my UERMs. I then stuck this ear onto my v2 type closed cell foam coupler since the other foam doesn't give a good seal with anything but open headphones with leaky pads.

    The result was a very rolled off curve, but moderately promising when compensated for the resulting roll-off. When matched at 500Hz the V2 coupler on its own nets a 7dB difference at 4kHz, the V1 coupler a 5dB difference and this weird contraption gets a 4dB difference. My in-ear measurement method gets a 2.5dB difference between the two headphones here, the closest to what I hear personally. But again the results are almost unusable past 5kHz.
    Foam Ear on v2 closed cell foam coupler test 1 modded HD800 and modded HD600.jpg
    In this case the foam seems to deaden resonances too much. I guess I'll order some silicone ears from AliExpress. Maybe an IEC 711 type coupler aswell to get an ear canal and middle ear simulation.
     

    Attached Files:

    Last edited: Jul 18, 2021
  11. Lurker

    Lurker Facebook Friend

    Joined:
    Oct 10, 2015
    Likes Received:
    119
    Trophy Points:
    33
    Yeah, I'm not sure about the foam ear.
    The best DIY thing you could try is the Sony method: https://www.sony.jp/headphone/special/park/history/index.html
    They take negatives of the ear using silicone to make a replica. The replica can be fitted with a coupler to measure earphones:
    [​IMG]
    Here is a more comprehensive interview with the "ear type craftsmen": https://av.watch.impress.co.jp/docs/topic/600821.html
    I feel like pouring silicone in there could wreck your eardrum if you're not careful so I wouldn't recommend this. Instead of random hardware store silicone you could probably use the Otoform from McEar.

    Nonetheless, unless you only measure open headphone you won't get around a coupler.
    A month ago I built a rig using the aliexpress 60318-4 ear simulator ("Type E610A") and the aliexpress pinna. It actually works quite well despite not mounting flush:
    [​IMG]

    To evaluate accuracy, here is the Audio Technica ATH-MSR7b with the HD 650 (stock, new pads) as compensation:
    [​IMG]
    Crinacle's G.R.A.S-system:
    [​IMG]

    The point of the above graphs is to demonstrate
    1. The accuracy of coupler's input impedance
    2. Error induced by weirdly shaped Aliexpress pinna and resonant 3d-printed plate
    For 1. it is important to recognize that the input impedance can ruin comparability even on the same rig.
    Generally open headphones have a low acoustic output impedance and sealed headphones have a high acoustic output impedance. This is why the sealed type typically looks weird on flat plates. The impedance match is not accurate. This gets especially bad on sealed insert earphones which is why the tube couplers are at best a guesstimation.
    For simple DIY this accuracy is good enough for me. And it costs less than the MiniDSP ears (sans cost of 3d-Printer, pins, screws, Scarlett2i2, Rode VXLR+, my precious time).
     
    • Like / Agree Like / Agree x 2
    • List
    Last edited: Jul 19, 2021
  12. Marvey

    Marvey Super Friend

    Staff Member Friend BWC
    Joined:
    Jun 6, 2021
    Likes Received:
    1,679
    Trophy Points:
    93
    The latest from GRAS is best if you want accuracy IMO. They've been at it the longest, with a ton of data from Oratory (43AC and 45BC) - I like how he is offering a better compensation away from Olive's consumer preference crap.

    TBH, the 5128 sounds like e-peening. On the part of both Jude and B&K. Could care less about the "tech". Ultimately, it's where's the beef?
     

Share This Page