As an avid user of EQ with headphones for some time now, I’ve completed my ascent to at least stage 9 of the 10 Golden Stages.
(I wouldn’t know about stage 10 as I don’t attend parties unless they are full of other headphone enthusiasts)
And I’ve developed a few techniques to EQ headphones using online tools like the one from @listen_r. But I’m interested in how to further tune the EQ for my specific HRTF.
When I have raised this question in the past in the oratory1990 Reddit, I haven’t gotten anywhere. It seems oratory1990 does not believe there are any and he has specifically discounted doing frequency sweeps for this. Yet I have also seen @Resolve mention a few times recently that he uses this technique.
It seems to me that in ear microphones, even given their limitations, would get me at least closer. If so it would be well worth the investment to me. Again I’ve seen @Resolve mention using them in the past, but not recently. So I am eager to know if he still uses them and how.
So how about it? Does anyone here have experience tuning a headphone to their personal HRTF and of so, what did you learn? What works and what doesn’t?
6 Likes
So, Oratory’s conjecture that we don’t have a reference point for sweeps is IMO correct. I’ve also said similar things about not indexing for equal loudness, having tried to achieve it in the past. Things actually ended up sounding kind of weird in the treble.
With that said, when you hear a notable peak in music, you’ll typically hear that peak in the same spot with tone-gen. You may find additional peaks in tone-gen that you didn’t realize with music, and if that’s the case I wouldn’t worry too much about them.
And what I tend to find is that if I’m able to identify it with music, it’s also usually a stronger feature.
Bottom line is that I use it for big things but ignore it for small things. I think the question though becomes… what’s the relationship between this and our individual HRTF. And for that I’m not entirely sure, but I think we also need to remember that the DFHRTF isn’t based on a single angle of incidence.
Regardless… this is a good question IMO, and one that I don’t have a straightforward answer to as of yet.
3 Likes
My problem is that I can’t seem to identify peaks or troughs when listening to music. And maybe that’s ok. Maybe that means the EQ I am using is already good enough. And maybe that’s because my HRTF doesn’t deviate significantly from the measurement rig.
On the other hand, I’ve had a few experiences where I try something new and at first it just sounds different, but over time the music seems to gel better. It’s difficult to describe but it truly does sound more balanced. This is rare though, and typically I get the opposite effect. In which case I return to my previous EQ preset. But this technique is practically just trying random changes and each change takes a long time to evaluate with multiple pieces of music.
1 Like
Yeah, I think it’s worth doing a combination of music and/or sweeps. In my experience, on-head measurements can be quite useful in understanding how your anatomy may differ from that of the measurement ‘head’, although I’m not sure this is stuff folks need to get into.
Do you generate pink noise along with the sweep signal? I think that is the technique oratory1990 recommends, if you could call it a recommendation. 
1 Like
Yeah absolutely. Like… I don’t ever just use a sweep on its own. I use the cursor-based sweep function in REW, but it also has noise functions as well.
To be clear(er): I mean generating both the pink noise and the sweep signal at the same time.
Oh, yeah I mean if that works by all means. I haven’t personally tried that but worth a shot.
My guess is that when listening to music, your brain is doing the correction for you. The sweep is measuring something much simpler that you don’t normally hear.
My problem is finding the right EQ for the empty space between my ears.
5 Likes
I just use a nice single malt for that.
5 Likes
Very cool conceit for a thread and I’m interested to see how the discussion evolves.
So, my approach for IEMs is usually EQing using 5128 measurements to either our JM-1 baseline or the ISO 11904:1 DF HRTF, and then using pink noise and music to identify and change a few common features that I know will be a source of disagreement between the users’ HRTF and the rig’s, as well as the user’s in-situ response and the measurement in the rig:
-
The primary ear gain & canal resonances: while 5128 DF/JM-1 has a 3kHz primary resonance, the average human DF HRTF per ISO 11904:1 has a primary resonance closer to 2.8kHz, but this will vary between users, and thus should be done manually. I’ve found shifting the primary resonance of my equalization to be closer to 2.7kHz to be ideal for me, when checked with pink noise and music (as well as sweeps, wherein EQing to eg. JM-1 often has a rather annoying overemphasis at ~3.2kHz.) Additionally, your canal itself might vary such that you need to make rather dramatic changes in the 2-4kHz band for subjective smoothness. I can’t vouch for corrections here making it more congruent to your HRTF, but it can definitely make it sound better, so give that a shot.
-
The length mode resonance and subsequent resonances: This is something @Resolve and I speak about often in content, and its absolutely true that the length mode (half wave) resonance seen on measurements will rarely match with your individual in-situ response, as it is due to the length of the air volume between eardrum and IEM. You are not the measurement rig, so it’s likely you’ll need to find where it sits manually with a user-controlled sine-sweep if you’d like to equalize it. However, it should be noted that the difference between you and the rig also means the rest of the treble (which is also effected by the length of this air volume) may shift up up or down in frequency depending on the difference in length between your air volume and the rig’s.
-
Upper treble: this is a high uncertainty area, and unfortunately the best I’ve been able to do is identify resonances with pink noise and sweeps and slightly curtail them. I can’t guarantee this is the best way to handle this area, nor do I think its the best way to “match your HRTF,” but I’ve found it helpful in understanding specifically how my in-situ response + HRTF deviates (or doesn’t) from the 5128’s in-situ response compensated to its own HRTF, as well as of course making the sound subjectively more palatable.
This process should at least get you marginally closer to the goal, and then of course you will likely need to add preference adjustments as well (bass boost, treble cut, downward tilt, etc.)
I’m at my sister-in-law’s graduation so can’t speak more about this currently, but I’m excited to see where this discussion goes from here.
3 Likes
I think the idea is to simulate a music signal while you are listening for the peaks and troughs with the sweep signal.
Absolutely. But it can only correct so far, right?
1 Like
I suppose the question would be… if we perfectly equalize such that a tone-gen sweep has equal loudness to it, would that mean we’ve achieved agreement with our own HRTF? and if so… which HRTF?
My conjecture has been that there could be things we naturally hear that might not be meant to be perceived as equally loud across all bands. But I’m open to that being wrong, it’s just something that’s popped into my head when trying to do this.
2 Likes
It’s a good question. I don’t think equal loudness with a sweep should be the goal, but since we can’t really know in-situ response, I’m not yet sure how to more efficiently get a subjectively pleasing result than using pink noise, equalizing to a “reasonable” DF HRTF, making the adjustments that are bound to be necessary to address inconsistencies between the rig and the listener + lessen massive swings in perceived treble magnitude, and then start playing with preference filters.
1 Like
Are you saying that the HRTF would be for the specific headphone (if using a headphone) to generate the sweep? And that changing anything - position of the headphone on the head, clamping force, ear pads, possibly hair growth inside of the ear would create a different HRTF?
That it would need to be measured for each target headphone, ideally before listening to music?
And could not listening itself change the HRTF? For example listening an hour at 85 db vs listening after a quiet spell to 75 db? I know that after a rock concert - even with ear protection - I sometimes hear differently.
No like, ‘HRTF’ is not a proxy for just “how I hear things”, which I think maybe some people may use the term. So it’s a response at the ear drum based on a given condition - we use the diffuse field sound field for headphones because that’s the one that makes the most sense given headphones are worn on the head, and not like speakers at a distance. But that’s also a calculation from lots of different incidences.
So like, if you change the angle/position of a speaker, and you measure it at the ear drum, the FR result will change quite substantially as you change the position of the speaker. But at the same time, your head and ears are impacting the sound every time, and this could also be considered the “HRTF”, just not in a diffuse field, which would include all/no direction.
2 Likes
You are reinventing a whole bunch of perception science topics.
Testing simulated music is highly uncertain (dubious?) because of the inherent complexity, perceptual limits, known perceptual overrides/interactions/illusions. I’d test with simple signals OR real music and not bother with intermediates. For example:
Brain correction does have limits. See the very old research on this per just-noticeable differences. Few people would mistake the tone of a tuba for the tone of a flute, but given synesthesia, stranger things have happened.
You hear what you personally hear. Determining the precise physical/electrical causes and parameters versus the perceptual quirks of your body is a methodological bear…and then most of us habituate to whatever we hear pretty quickly. They sound “good” or “normal” and we often fail blind tests.
2 Likes
