A (hopefully) new way of presenting frequency response differences between headphones

Discussion in 'Measurement Techniques Discussion' started by Koloth, Dec 13, 2015.

  1. Koloth

    Koloth Klingon SBAF Ambassador - Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    120
    Trophy Points:
    43
    Location:
    Zürich, Switzerland
    Hey guys,

    I'm really interested to get your thoughts on this. For some time now I have been thinking about the way 'we' graphically present frequency response differences between headphones. The main issue in my opinion is the 'readability' of the charts as employed on SBAF, Innerfidelity etc for 'noobs' (but also for me tbh): If one wishes to compare two headphones one has to constantly switch back and forth between two graphs which are somewhat complex to understand and even more difficult to relate to subjective listening experiences.
    I think there might be a complementary way (which would not and could not replace the current system) of presenting the information, that might be much more noob-friendly, by focusing on direct representation of the relative differences in frequency response (tonality) between two headphones. My idea basically boils down to using the frequency response curve of model A as a correction curve for the frequency response curve of model B. That way you get a curve of the differences of B to A.

    The scenario would look like this:
    A user, say he goes by the nickname Koloth and has difficulties 'reading' frequency response charts, wants to know how a certain headphone B, that he intends to buy, compares to headphone A, which he is very familiar with. He inputs the two models - the comparison baseline A and the headphone in question B - into a web-application, which then generates a curve that looks like this:

    [​IMG]

    This allows him to directly and easily 'read' the relative differences between the two. He does not have to compare the frequency response curve to an ideal frequency response (flat/Olive-Welti/whatever) which he is not familiar with or use intermediary steps of other models he is familiar with, but he can instead precisely see how the model in question (in this case the Audioquest Nighthawk) compares to the model he is already familiar with (in this example Sennheiser HD600). This requires much less technical understanding and tedious back-and-forth-switching.

    In the above example - which I painted by hand and whose values are derived from Tyll's measurements on Innerfidelity - the HD600 serves as the baseline for the Audioquest Nighthawk. Using Tyll's measurements I have calculated the relative frequency differences at each frequency mark and drawn the above curve. (Note: This is only illustrative; Since I've done it by hand it's prudent to look at this as a 'ballpark estimate' lacking the precision we are used to from measurements.)

    The interested user could easily perceive that the Nighthawk has much stronger sub bass, somewhat thicker lower mids, but recessed mids/upper-mids/presence region. With very basic knowledge of psychoacoustics he could deduce the assumption, that the Nighthawk would IN COMPARISON TO HIS HD600 make voices sound more distant with much more bloated and deeper reaching but not especially more punchy bass (Note the comparative and non-absolute wording).

    I am sure people far more accomplished than I am with web design and software programming could come up with a software that automatically generates curves such as the above for any two user-specified headphones in a database. I feel that this might prove to be a very valuable mode of representation - not only, but especially for newcomers to the headphone world.

    I'm truly interested to hear what you think about this. I hope that its not obviously dumb or defective in some way or that this has not already been done on some other website I am unaware of. :)

    Best regards!

    PS: I take the freedom to mention @Tyll Hertsens , @purrin and @ultrabike in this post, as even a short feedback of these resident measurement experts would be much appreciated!
     
  2. Koloth

    Koloth Klingon SBAF Ambassador - Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    120
    Trophy Points:
    43
    Location:
    Zürich, Switzerland
    Another example I just put together:

    How does the Sony MDR-Z7 differ from the Mr. Speakers Alpha Dog which serves as a baseline?

    [​IMG]

    The Z7 has less subbass, quite a bit more recessed mids, thus the bass and upper mids are comparatively more pronounced. The presence region is a bit reduced so it will most likely sound a little less sibiliant than the AD.


    Also a general note: If a system could be built on top of a database of measurements to generate such curves per user request, one could of course include all kinds of standard compensation curves as well. For example: How does my AKG K701 compare to the Olive-Welti ideal curve? Or the ideal curve that Philips is advocating? And how does the Harman curve compare to the Philips curve? etc etc.
     
    Last edited: Dec 13, 2015
  3. Boops

    Boops Friend

    Pyrate
    Joined:
    Nov 8, 2015
    Likes Received:
    3,179
    Trophy Points:
    93
    Location:
    New York
    Really interesting stuff. I could see this kind of dynamically normalized visualization of FR being really useful as a toggle you could flip on a chart rendered in the browser.

    From an accuracy perspective, it would probably be worthwhile to keep the "true" FR as the default, with this as a secondary option, esp for the specific case you mention of comparing a known headphone to an unknown.
     
  4. Koloth

    Koloth Klingon SBAF Ambassador - Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    120
    Trophy Points:
    43
    Location:
    Zürich, Switzerland
    I agree absolutely. This cant replace the absolute measurement. And thats not what it's designed to do. Its simply a tool to visualize differences in tonality between headphones. As you say, a 'secondary option'. But one I would love to see available somewhere.
     
  5. velvetx

    velvetx Gear Master West/Vendor Spotlight Moderator

    Staff Member Pyrate
    Joined:
    Sep 26, 2015
    Likes Received:
    1,067
    Trophy Points:
    113
    Location:
    Bay Area, CA
    You sir are a gentleman and a scholar now I can read these things :)
     
  6. Koloth

    Koloth Klingon SBAF Ambassador - Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    120
    Trophy Points:
    43
    Location:
    Zürich, Switzerland
    Unfortunately I dont have the programming skills to build such a system myself. And of course I dont have a database of standardized measurements...
    Come to think of it, if I had the measurement data I could probably build something in Excel that would get the job done. But I have 0 knowledge or understanding of web programming.
     
  7. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,771
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    I'll make an iPhone app and make you guys pay $5 for it. Or you guys tell me.
     
  8. Armaegis

    Armaegis Friend

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    7,462
    Trophy Points:
    113
    Location:
    Winnipeg
    Um... you've been to Headphone.com right? They already have this...
    http://www.headphone.com/pages/build-a-graph

    And I seem to recall Tyll saying that he was looking into building something similar for his data.


    Here's an example:
    [​IMG]


    Here's an interesting doohicky though... grab the image url:
    http://graphs.headphone.com/graphCo...[]=853&graphID[]=2751&graphID[]=4163&scale=30

    and reduce each number by 1 like so
    http://graphs.headphone.com/graphCo...[]=852&graphID[]=2750&graphID[]=4162&scale=30

    And voila you've got some non-smoothed data
    [​IMG]
     
  9. Armaegis

    Armaegis Friend

    Pyrate BWC
    Joined:
    Sep 27, 2015
    Likes Received:
    7,462
    Trophy Points:
    113
    Location:
    Winnipeg
    Reduce it by two and get the raw/uncompensated (I think) data...

    [​IMG]

    Have fun with that.
     
  10. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Wouldn't be too hard, depending on your software. Just a matter of using a measured frequency response and using the data output as a frequency response compensator in ARTA.

    Hmm, actually...let me whip that up for you! Let's compare the HD650 and TH-X00. Left channels. Ignore THD due to wife doing stuff in same room (MMO, loud keyboard, chatting, etc.) and heater being on, also loud. These are also just quick, one-off takes. No averages from multiple takes or anything.

    Let's start with a standard HD650 measurement - well, HD650 with slight mods as seen on the forums plus the addition of some drawer liner under the front foam, also with quarter mod hole cutout. (I still don't know why people hear a lack of bass extension, but I digress.)

    hd650 normal.PNG

    Now a standard TH-X00 measurement.

    th-x00 left.PNG

    HD650 with it's own compensation applied. Yep, pretty much flat as you'd expect. Will never be perfect due to each measurement having slight variations.

    hd650 flat compensate.PNG

    TH-X00 with HD650 compensation applied. Ah, yeah, I can see why some hear these as brighter or rougher. Definitely more bass below 100Hz. Similar midrange. I am not 100% sure about the data above 10KHz.

    th-x00 w hd650 compensation.PNG

    These are pretty easy to crank out once you build the compensation file.
     
  11. Luckbad

    Luckbad Traded in a unicorn for a Corolla

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    3,408
    Trophy Points:
    113
    Location:
    Holly Springs, NC
    Awesome idea and execution! The TH-X00 proves slightly too fatiguing for long sessions for me, and this graph shows why.

    Love the bass, but the treble is still slightly high. The HD650 has the most treble I can handle at high volumes for several hours.

    This is a clever way to describe a headphone objectively compared to over many have heard.
     
  12. Koloth

    Koloth Klingon SBAF Ambassador - Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    120
    Trophy Points:
    43
    Location:
    Zürich, Switzerland
    Amazing responses in this thread so far! Unfortunately I'm almost passing out from tiredness atm, so I'll go to bed and take a good look at all of this tomorrow :)
     
  13. Boops

    Boops Friend

    Pyrate
    Joined:
    Nov 8, 2015
    Likes Received:
    3,179
    Trophy Points:
    93
    Location:
    New York
    It should be free but with a $.99 in-app purchase every time you render an FR curve.
     
  14. Ninja Pirate

    Ninja Pirate Friend

    Pyrate
    Joined:
    Sep 28, 2015
    Likes Received:
    163
    Trophy Points:
    43
    Location:
    NH
    Home Page:
    How do you typically go about creating the compensation files? Do you just manually edit via a spreadsheet or is there a better way?
     
  15. Hands

    Hands Overzealous Auto Flusher - Measurbator

    Staff Member Pyrate MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    12,285
    Trophy Points:
    113
    Location:
    Colorado
    Home Page:
    Well, I already have a compensation file in place for the bass roll-off on the ADC and mic phantom PSU. Measure with ARTA, save response results as text file, load back into ARTA as compensation.

    Do the same with the HD650. Save response data. Then I had to add that in with the ADC+mic phantom PSU compensation file to again catch the inherent hardware bass-roll off there. Load that all up as a compensation file.

    Take a measurement, and there you go! Now the results are all relative to whatever headphone you used for the compensation file.
     

Share This Page