Before I begin this post, I would like to address the elephant in the room:
”Why not diffuse field?”
This is something I’ve been asking myself every time I EQ’d my headphones (Audio Technica ATH-M50x) to a particular target (Harman 2018 AE/OE, SoundGuys, etc.), because without failure there was always excess energy in the sibilance region, and as someone with somewhat mild hyperacusis, this drove me up the wall. Songs with bright mixes such as some Drake songs, for example, were practically unlistenable for me regardless of the volume I listened at. Quiet enough to tame the sibilance/harshness, not enough bass impact. Loud enough to get some oomph from the bass in the song, and the treble would be attacking my ears with daggers (perhaps a slight exaggeration).
Now, of course, as with everyone who develops somewhat of an obsessive dive into the headphone rabbit hole, I was undoubtedly questioning the validity of everything. Why was the KEMAR used so widely yet sounded terrible with every target I EQ’d to? How does the 5128 cost so much yet targets developed with it sound astoundingly subpar? Aren’t the Harman filters an astounding oversimplification of the complexity of hearing? These are all thoughts I’ve had, but as I’ve come to learn with my time endlessly tinkering with target curves, there is always plenty of room at the table for subjective experience and opinion.
One thing I tended to notice was that target curves rarely, if at all, specified if they were intended for or developed on open-back or closed-back headphones. This may be an irrelevant point, but it is what lead me to exploring the idea of potentially using the 5128’s free field as a baseline to forming a target as opposed to the tried and true diffuse field. Queue the head shakes of disapproval. I am aware that the free field has been tested in research, and produced a very convincing last place position. Regardless, I wanted to give it a try. Perhaps it could solve the gripes I had with treble, and maybe solve the issues I had with that pesky 5-10kHz region, after all, it looks like this:
And out of sheer coincidence, the free field (on paper) appeared to exactly solve this gripe! However, EQing my headphones to this response curve went about as expected. No bueno. Not enough bass, very mid forward, and the sound was very cramped overall. Not a great experience. But nevertheless, I could apply the 2018 Harman filters and get a good sound, right? Err, no. Now I get bloated bass, an airy yet dull treble, and just an awkward tuning overall.
Alright… well now I was stuck between a rock and a hard place. The DF with Harman filters sounded bad, and the FF with Harman filters also sounded bad. So, now what? Well, like anyone who slowly drives themselves insane by staring at a bajillion headphone measurement graphs, targets, EQs, threads, forums, listening to all kinds of wonky tunings and self-made targets with questionable methodologies, I turned to the ever-elusive ~ Personal HRTF ~… with the only caveat being I don’t have the money nor the knowledge on the subject to pursue this avenue (sigh… one day). Luckily, I found this post by @listener which helped provide some insight into speaker characteristics within a room, and how it may relate to the perception of a “flat” headphone (or rather, how the in-room behaviour of speakers can be utilised to make a headphone tuning sound more “correct”). More importantly, I found this image in the post quite intriguing:
This graph, which shows the difference between the measurement of headphones on the 5128 (green) vs. a group of people (blue), appears to outline the very issues I have with both the DF and FF curves on the 5128, regardless of filters applied. That being:
- Not enough bass extension
- Too much 5kHz - 10kHz / sibilance & harshness
- Weird relationship between mids and treble
I also spent far too long racking my brain over how to use this data to modify a 5128 DF/FF. Do I add the extra treble and subtract the bass from the baseline target so that a person would hear the baseline? Or do I do the opposite, so that the target represents what a person hears? Honestly, I still can’t wrap my head around the logic, so I just tested both. Well, it goes without saying, I absolutely did not enjoy either the DF or FF with the extra treble and less bass, regardless of any extra preference shelf adjustments. With less treble and more bass however, things started to look promising. Both the DF and FF targets sound much more listenable with just this simple transform adjustment, albeit both still sounding a tad too bright for my liking. My preference also very much started to lean towards the free field in this regard, as the mids seemed to integrate much better with the treble.
I also tried to slap on the Harman 2018 filters onto both of these modified targets, and quickly discovered that for once in my life (I am a basshead), I did not enjoy the amount of bass. The overall signature of both became much too boomy and dark for even my tastes, as someone who tends to prefer some extra bottom end and a smooth, potentially warm signature. I then remembered that the post that I had got the population transform from was primarily about relating headphones to the behaviour of speakers in-room. Now, I will depart from what listener used for the high shelf, and I opted for a HS that is 4500Hz, -4dB, 0.4Q. This very closely resembles an almost linear downward slope from 1kHz to 20kHz, a behaviour I have measured from my unequalised speakers in my listening space, and one I have seen mentioned in several forums across the web as being the general in-room response that is preferred by mixing engineers and whatnot, though this may be a baseless claim. I also decided to tack on a -1.8dB peak filter at 3000Hz, just to tame some of the ear gain that I don’t necessarily need.
Overall, this leads me to my target for my (closed-back) headphones: A B&K 5128 free field, modified to represent a population’s in-ear measurement, and applying the natural timbre roll-off of speakers in a room:
“What happened to 8/9kHz!?!?”
Well, to me, at least, this curve sounds damn near perfect, or at least the very best tuning I have tried on my headphones. I have checked by going to several songs and videos which I had previously perceived as having harsh sound, whether through the mids or in the treble, and that harshness is completely eliminated. Songs that I already loved the mix on, got an immediate upgrade. I have tried applying the same adjustments to the diffuse field curve, however there is just still too much 7-10kHz, many youtube videos become difficult to watch with how harsh sibilance becomes, and I find this to be true for every target I’ve tried that uses the diffuse field as a baseline.
This target curve also very closely mimics how a good set of stereo speakers sounds to my ears, despite the free field baseline being measured from a 0 degree incidence. I also find it particularly interesting that everything I perceive as harsh and unnatural in my headphones seems to correlate to anything that does not sound to me like listening to stereo speakers.
Now, to clear up some concerns you may have:
- “This is just confirmation bias of a very particular preference response”
- Potentially, it would be interesting to see if anyone else likes how this target sounds on their closed-backs. I find it to be very natural sounding, and although the “soundstage” is narrowed, most sounds seem placed in front as if listening to speakers.
- “You don’t prefer free field, you prefer diffuse field with a heavy cut from 7kHz to 10kHz”
- Maybe, however I find the mids and bass to sound much more natural and accurate to me with the free field target.
- “Why just for closed-backs?”
- Alright, that part is just pure speculation. I don’t own and have never owned a pair of open-backs, but I have noticed that almost all headphone-related discussion seems to revolve around open-back headphones, and from my shallow digging into Harman research, it seems to me that there is a heavy emphasis on open-backs there too, though I may be wrong about that. Another reason why I state this to be for closed-backs, whether for myself or for others, is that I predict that the diffuse field baseline may actually be the ideal baseline to use for open-backs, as they interact directly with the external environment. Since closed-backs are a sealed environment, it makes sense to me to treat them as an anechoic chamber, logically reinforcing my perceptual preference for free field. Perhaps some confirmation bias at play here.
- “No headphones have ever been tuned like that”
- I am well aware, as I’ve compared it to every headphone available on listener’s 5128 squig.link database, and despite some potentially coincidental similarities in some frequencies to certain models, there is absolutely nothing that resembles this curve remotely, especially in that 9kHz zone. You may see this as a sign that my target is just incredibly personalized. I’m not refuting the idea, my hopes are that the methodology I used could be helpful to some, or perhaps spark some productive discourse. Of course, it would be amazing if somehow I’ve created the magical closed-back target that everyone loves, but of course I am working with a sample size of one and expecting such a result is baseless.
Well, there you have it. Is this a massive nonsensical dribble post? Did I create the world’s highest effort sh-tpost? Who knows. I’m interested to hear anyone’s thoughts on this, if anyone decides to spend their afternoon reading my very professional thesis (/s). Please let me know if I missed anything or didn’t quite explain something properly.
I’m also new to this forum, so forgive me for any improper formatting/etiquette. I look forward to chatting with you all.
