Innerfidelity FR target

Discussion in 'Measurement Techniques Discussion' started by Serious, Dec 19, 2017.

  1. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Tbh I much prefer the IF measurements to the flat plate coupler measurements, simply because I find the FR to be more accurate, but dummy heads are always a PITA to get a good compensation.
    I got annoyed that we still don't have a better compensation curve for the IF graphs and even though you start to learn how to read them after a while, things are much easier to compare when flat is a straight line. So that's what I tried to do, essentially make my own target. I'm now at a point where I feel comfortable sharing this with you guys. Tell me what you think! Here's what I came up with:

    0.941Target.jpg
    This is to use as an overlay for the regular graphs. It should fit with all of the more recent (2015 or 16 and newer I believe) graphs and you'll need to resize the older graphs.

    Remember - I'm an HD800 guy, so could be slightly too bright to you, or maybe it's even slightly warm to you. I know I've said a couple times how much I hate the Harman target with its 1-2kHz bump, but even I decided on a bit of a bump at 1kHz relative to Tyll's ID compensation. But remember, it's not final, this is just the first one that I quite like after working on it for a while.

    Here's what a few phones look like compensated with this target:

    Sennheiser HD650:
    Sennheiser HD650.jpg

    Sennheiser HD800:
    Sennheiser HD800.jpg

    Focal Utopia:
    Focal Utopia sn A1BEHF000812.jpg

    Focal Elear:
    Focal Elear sn 1BEBG004809.jpg

    Sonoma Model One:
    Sonoma Model One.jpg
    This one is interesting because we have this pic with a Harman target compensated version. Headphone system with built in EQ to almost perfectly match the Harman target, this is what it looks like on here.

    Stax SR007mk1:
    Stax SR-007 (late Mk1 sn SZ1-1023).jpg

    Stax SR009:
    Stax SR-009 SZ9-1278.jpg

    HFM HE-560:
    HiFiMAN HE-560 2014.jpg
    The HFM stuff tends to measure very bright on Tyll's dummy head, but I don't think they're that inaccurate tbh. The HFM stuff is bright.

    Funnily the Edition X almost perfectly matches my target:
    HiFiMAN Edition X.jpg

    Let me know what else you want to see. I also posted three Audeze phones here:
    http://www.superbestaudiofriends.or...ments-and-impressions.5556/page-5#post-181576


    EDIT: How-to here:
    http://www.superbestaudiofriends.org/index.php?threads/innerfidelity-fr-target.5560/#post-181909

    EDIT:
     
    • Like Like x 23
    • Epic Epic x 1
    • List
    Last edited: Dec 22, 2017
  2. bazelio

    bazelio Friend

    Pyrate
    Joined:
    Nov 16, 2015
    Likes Received:
    3,417
    Trophy Points:
    113
    That's cool. Thanks for the effort. Now, where can I get a HD800 compensated with a Focal Elear? ;-)
     
  3. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Actually...
    Sennheilser HD800X.jpg
    Not perfect, but pretty good. There are ways to get more upper mids and less treble than that. Also depends highly on pads.

    I was surprised to see it have more upper mids than the one with the SBAF and SD mod. Probably pads and a slightly smaller SBAF cutout. Especially the 1-2kHz region should be higher than that. The SD resonator actually lowers the 4kHz region in my measurements*, so no surprises there.
    *(compared to blocking the area completely, not compared to no resonator at all)
    Sennheiser HD800 sn 15001 SuperDupont and SuperBAF mods.jpg

    Sennheiser HD800SD:
    Sennheiser HD800 sn 15001 Superdupont Resonator.jpg


    The original MrSpeakers Ether was actually pretty damn neutral:
    MrSpeakers Ether.jpg
    Sounded a bit leaner and rougher to me, but I'm pretty sure the huge dip in the treble is just an artifact, so hard to say what it should really look like. Also you never know if the one Tyll measured sounds exactly like the one you heard.

    Some other phones:

    JPS Labs Abyss AB-1266:
    JPS Labs Abyss AB-1266.jpg

    Koss ESP950:
    Koss ESP950 Sample2.jpg

    Sennheiser HE60:
    Sennheiser HE60.jpg

    TH-X00:
    Massdrop TH-X00.jpg

    HiFiMAN HE1000 PreProduction:
    HiFiMAN HE1000 PreProduction.jpg
    I've heard both bright and smooth sounding HE1000 V1s. Both of Tyll's measurements seem to be of the bright sounding variant. I picked the pre production one because it seemed to have better channel matching.

    Sennheiser HD700:
    Sennheiser HD700.jpg
    Not much else to say about it. The FR looks like an even more extreme HD800 curve.
     
    Last edited: Dec 21, 2017
  4. Priidik

    Priidik MOT: Estelon

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    2,148
    Trophy Points:
    113
    Location:
    Estonia
    Your comp certainly makes modded HD800 look alright and close enough vs what I consider speaker neutral to be.
    Perhaps the overall brightness is a few dB under represented with your compensation.
    Difficult thing to nail down, as broadband tilt one way or another is easy to accommodate.
     
  5. Tyll Hertsens

    Tyll Hertsens Grandpappy of the hobby - Special Friend

    Pyrate
    Joined:
    Sep 26, 2015
    Likes Received:
    2,805
    Trophy Points:
    93
    Location:
    Bozeman Montana
    Home Page:
    Damned nice work. When I get my headphone measurement tool up and running, I'd be interested in a compensation curve that you author. Needs to be 512 points from 10Hz to 22kHz. Let me know if you're interested.
     
  6. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Hell yeah!


    When looking at the average level, especially above 10kHz I think you might be right. These curves had a lot more upper treble: https://www.innerfidelity.com/conte...rman-target-response-curve-various-headphones
    When looking at the stock HD800 curve for example the average level above 10kHz looks to be around the same as the 1.5-4kHz upper midrange level, but when looking at the peaks, it's about the same as the bass level. In that case I prefer to look at the peaks (although technically that doesn't make much sense). I think you get a lot of these narrow dips in the top octave with the dummy head that aren't necessarily as bothersome subjectively. Also the 12kHz dip shows up in a lot of these measurements (and in my own too) and I do hear it when listening to sweeps with both the HD800 and HD600, so I decided to leave it in.
    Maybe increasing the overall level above 7kHz would be a good idea, but then some phones like the stock HE-6 below look very bright already and I'm not sure if I hear it as that bright. The modded HD800, Utopia and Abyss graphs look about right to me where they are now, but the SR007 and SR009 and maybe the HD650 and HE-560 could maybe use some more upper treble. I also don't remember the TH-X00 as that extended up top. It's hard to say overall.

    I'll also say that I prefer a bit of a downward slope after 10kHz compared to flat with speakers, like the Accuton diamond tweeters have (5db down at 20kHz), but that's partly also to compensate for MC cart rising top end (most of my listening is still digital, but that'll change sooner or later). It's also pretty easy to tweak that with cables subjectively, so hard to set absolutes.

    HiFiMAN HE-6:
    HiFiMAN HE-6.jpg
    Maybe it should have a bit more upper treble relative to the 5-8kHz level?

    I do notice that peaks sometimes get shifted in frequency with the dummy heads relative to my own head. The spike that's just above 6kHz in the Utopia measurements was probably a bit lower in frequency on my head. At least it is in my measurements. Subjectively I can't say exactly, but I think it was slightly lower than 6kHz. The peak between 9 and 10kHz and the strong dip afterwards are both a bit lower in frequency.
    WIth the HD800 graphs the 9kHz dip (actually a peak in disguise) is at a higher frequency (I hear it at 8kHz), although I do hear 7, 9 and 11kHz peaks too.

    Some measurements also make me think that the 2kHz level is too high (HD650, HD800, HD800SD), but it seems fine with the others, or even slightly low with the modded HD800 (depends on mods and pads so hard to say). The 3-4kHz level looks slightly low with the stock HD800 and HD650, but it's just right with the Utopia and the HE1000, etc, so I think it's about right.
    The 5-8kHz level is highly variable with positioning I think and averaging a couple positions might broaden features here. I do hear a dip with the HD650 around 6kHz, but I don't think it's as wide as the measurements suggest. But increasing the level above 7kHz will make everything look super bright and I don't think the dip is that far off anyway (imagine the HFM, STAX or Focal curves with 3db more at 7-9kHz).

    Anyway, just my thoughts on the graphs. I spent way too much time on this the past couple of days so I'm a bit more protective of this than I might like. Definitely feel free to call me out on anything you feel is wrong. I need a bit of a reality check sometimes. It's not perfect and it never will be.
    I realize this post got pretty lengthy (almost 1 Bill-P post length), but I hope you have a better idea on how I read the graphs now.
     
    Last edited: Dec 20, 2017
  7. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Another thing to keep in mind is that right now the left and right channel compensation isn't exactly the same, so what looks like a bit of an imbalance in the measurements (mainly from 500Hz to 1kHz and 7-10kHz) in most of these measurements isn't actually there. The left channel is often a bit higher in that region.

    You can see the difference between the L and R target curves here:
    https://www.innerfidelity.com/content/new-compensation-curve-innerfidelity-measurements

    The curves look closer in both that article and this one, so I don't think the difference between left and right is needed.
     
  8. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    A few more headphones.

    Focal Clear
    Focal Clear sn A1BRQE000013.jpg

    Looks pretty good to me. Like the Utopia and Elear it still has the 8-10kHz bump. Some similarities to HD650, but a bit rougher and brighter it seems.

    Sennheiser HD600:
    Sennheiser HD600.jpg
    I hate posting this one because it's one of the only HD600s in existence with a channel imbalance.

    Sennheiser HD580:
    [​IMG]

    Sennheiser HD6XX:
    Sennheiser HD6XX.jpg

    Beyerdynamic DT880 600 Ohm:
    Beyerdynamic DT800 600 Ohm.jpg
    I never thought the DT880 was as laid back in the upper mids as the stock HD800 for example. Generally would be a very nice tonality without the peak.

    I do think they could be more accurate and in general the 1-3kHz area probably looks a bit too forward on these graphs, but I don't think that's true for every measurement and the HD600 does have some vocal forwardness that I tried to correct with mods, so it apparently did bother me enough. I do think it's not just a 3-4kHz bump, but I also think there's less 2kHz. Since most other phones look alright to me and the HD600 and HD580 right channels look about right too up to 5kHz I'm not too worried.


    Audio Zenith PMx2:
    Audio Zenith PMx2.jpg

    Emu Purpleheart:
    Massdrop x EMu Purpleheart.jpg

    ENIGMAcoustics Dharma D1000:
    ENIGMAcoustics Dharma D1000.jpg
    The 2kHz bump did bother me subjectively. Comparo to Hands and Marvey's measurements. Could be some variance between the units too.

    Klipsch HP-3:
    Klipsch Heritage HP3.jpg

    AKG K701:
    AKG K701 Sample B.jpg

    AKG K712:
    [​IMG]

    AKG K812:
    [​IMG]

    EDIT: Fixed HD580 alignement
     

    Attached Files:

    Last edited: Dec 22, 2017
  9. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Sennheiser HD660S:
    Sennheiser HD660S.jpg
    Not too bad it seems.


    I started out with just a few measurements and I'm creating these as we go, so I also don't really know how they will turn out until I've created them. Looks to me like the HD660S FR is more accurate to how people hear them than the 650 or especially HD600 measurement. It's interesting that the HD660S is so much brighter than the HD650 while Marv's measurements looked the same. Maybe it's those new pads?
     
  10. Drakkard

    Drakkard Facebook Friend

    Joined:
    Nov 24, 2015
    Likes Received:
    136
    Trophy Points:
    33
    Location:
    Saint-Petersburg
    Very cool. How do you create those modded graphs?
     
  11. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    I use the GIMP curve bend tool to make the graphs.
    First I take a screenshot at 500% in Chrome of the FR in the pdf file. Then if it's an older graph I resize it to match the template (the one with the green line in it). Then I select the curves and put them in a new layer. I always select an area of 1150x580 so it's consistent. It's important to always select the same area. I always select 2 pixels of the black bar on the left. There'll be a bit of a gap on the right side. After the curve bend transformation the cuve should now be 1150x1046 pixels. I then resize the bent curve to be 1088 pixels high to fit my less compressed y-scale. And then I add the name of the pdf file and export it.
    I'm sure there are better ways to do it, but this was the simplest for me. It was huge PITA to get a good curve with the bend tool as it's generally not meant for precision. Here's the curve that I use:
    https://www.dropbox.com/s/h1a60iex652p5yf/SeriousV0.941.points?dl=0
    (Even though the pics all say 0.94, they're all the same 0.941 file)


    The green line above is simply an inverted version of a bent straight line, scaled to match the regular graphs. It actually started with me drawing a line in paint and then recreating the line with the bend tool and them improving on that original line. That one already looked very different from the one I shared in this post.
    Here's that curve. I mainly used the Focal Utopia and modded HD800 curves to make draw the line and sort of averaged the results I got from drawing my target on both. The 0.86 target already looks very similar to the one I use now.
     
    Last edited: Dec 21, 2017
  12. bartzky

    bartzky Acquaintance

    Joined:
    Dec 22, 2016
    Likes Received:
    62
    Trophy Points:
    18
    Location:
    Germany
    Home Page:
    Nice work Serious!

    I was curious how your target would look like at the eardrum (DRP) as I am more used to raw measurements, so here it is:

    [​IMG]

    I used the red curve from this graph for my calculations. I think this is the curve @Tyll Hertsens usually uses for his compensations.


    Some comparisons, all aligned @ 1 kHz:

    Vs Harman 2013:
    [​IMG]

    Vs "New Innerfidelity Curve" from here:
    [​IMG]

    Vs "The Blue Curve" that had been discussed here:
    [​IMG]

    Vs my own target curve (HFX Target):
    [​IMG]

    Vs Warren TenBrook's Target from this graph and discussed here:
    [​IMG]
     
  13. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,621
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Are these really even comparable though? Do we know that Tyll's head / ears simulator produces the same results as the Sean Olive's head? If not, how far off are they? Comparing different head / ears simulators with a yet a different set of compensations seems to be a recipe for disaster / confusion.
     
  14. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    That's super interesting. What bugs me is how it looks like there's a bit of a dip at 4.5-5kHz, when looking at the graphs if anything I'd want even less 4-5kHz. The HD650 graph looks flat at 4kHz when it should maybe be a bit forward. In that sense I agree with your target the most, but the 1.5-3kHz region looks pretty forward. I can see why one would want a bit more 2kHz, but another way of looking at it is that it's like the new IF curve, but with more energy below 1kHz and less concha gain.

    Heck, this compensation curve is based on my "this is the shit that I hear" curve which is probably different from yours.

    When looking at the graphs like this my curve isn't all that different, which is reassuring in that the general direction we're heading doesn't look too bad. I'm surprised by how close it is to the Harman 2013 curve (aside from the bass boost). Well, there's 2.5db more upper midrange energy with the Harman curve, which is not insignificant, but the general shape isn't too far off.
    Not saying that 2.5db too much upper midrange wouldn't be annoying. The 2db (according to my measurements) bump at 2kHz on the Elear drove me nuts within seconds!
     
  15. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,621
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The compensations are excellent and seem accurate. Am I correct to assume that your neutral ltarget is a straight line?

    I'd probably prefer something less bright, but that's EASY to figure out because I can just lift the line 2db past 5kHz in my head. I've got ears-on-the stick coming in, go going forward I might provide two sets of measurements.
     
    Last edited: Dec 21, 2017
  16. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Yes, essentially. Although some graphs aren't quite like how I expected them to be. I expected the HD600 FR to look more like this.
    My hunch is that the dummy head ears aren't as flexible as real ears and that's causing some of the differences. The HD6xx graphs look especially wrong to me. The HD600 actually touches my ears ever so slightly and if the dummy head ears need more force to move the driver will be closer to the ears increasing the upper mids and lowering the treble. Or maybe the pads were old on the HD600 and HD580 or Sennheiser changed the headphones slightly since then. Those measurements are pretty old by now.

    I think I'm going to leave it like this for now. With a 2db increase in treble the Audeze curves would be dangerously close to neutral. They already look similar to the one Hands just posted for the 2C (except Hands doesn't compensate for the upper treble rolloff like I do in my own measurements and I probably like 2-3db more 5kHz than his lines show).

    Here's an old LCD3 curve:
    Audeze LCD3 sn 2312260.jpg
    I like that the general shape of the treble from 5-7kHz on is similar to the old V1 changstar (or even pre changstar) era FR and CSD measurements.
    Compare here: http://www.changstar.com/www.changstar.com/index.php/topic,204.0.html
     
    Last edited: Dec 21, 2017
  17. bartzky

    bartzky Acquaintance

    Joined:
    Dec 22, 2016
    Likes Received:
    62
    Trophy Points:
    18
    Location:
    Germany
    Home Page:
    The HFX isn't actually a "this-is-the-shit-I-hear" curve, so I didn't really want the 1.5-3 kHz region to be like it is. Essentially I mixed diffuse and freefield according to the directivity of a good loudspeaker and combined it with a room-slope. And that curve came out.


    I find it remarkable how similar all those curves are considering the different approaches.
     
  18. Mayor D

    Mayor D New

    Joined:
    Nov 8, 2015
    Likes Received:
    15
    Trophy Points:
    8
    Location:
    Brighton
    Can you Do NightHawks?
     
  19. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,589
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Here you go:

    AudioQuest NightHawk (stock pads):
    AudioQuest NightHawk.jpg

    AudioQuest NightOwl (stock pads):
    AudioQuest NightOwl.jpg

    Was also interested to see some other phones...

    ZMF Atticus:
    ZMF Atticus.jpg

    ZMF Eikon:
    ZMF Eikon.jpg
    Hmm, those measurements are interesting. Not sure if that's what I heard. My measurements looked pretty different.
    *Actually rereading my impressions and looking at the measurements again, these FR plots might be more indicative of what I heard, except that I broke the seal on purpose to get less bass. I knew it sounded brighter than my measurements showed!

    Stax SR-L300:
    Stax SR-L300.jpg

    Audeze Sine:
    Audeze Sine.jpg
    Consistent with Marvey's impressions I would say

    Focal Spirit Pro:
    Focal Spirit Professional.jpg

    NAD VISO HP50:
    NAD VISO HP50.jpg
     
    Last edited: Dec 22, 2017
  20. metal571

    metal571 Not a friend

    Annoying cousin
    Joined:
    Oct 24, 2015
    Likes Received:
    54
    Trophy Points:
    13
    Home Page:
    Woah. Just chiming in here to say that's a great compensation - indeed very close to my experience with all of these cans. I still agree with yours and Tyll's approach of keeping the bass flat to visualize our own personal bass preferences.
     

Share This Page