A few posts above I wrote about my experience with in-concha microphones as an interesting way to measure headphones' effective frequency response on someone in particular's head. I also mentioned how this approach is severely limited past 1khz and that other means are necessary to get an accurate picture of what is going on above that.
Well, there is a way. It costs several thousand of dollars and it's called "Etymotic ER-7C". Here's how that baby looks :
https://www.etymotic.com/auditory-research/microphones/er-7c.html
It's a probe / tube microphone. Basically a silicone tube that is meant to be inserted in someone's ear canal, at the end of which there's a microphone and a lot of calibrating stuff happening after that.
Probe microphones were used in that study to measure the effective FR of 6 headphones on 6 subjects, for example :
https://www.aes.org/e-lib/browse.cfm?elib=16877
In that thread Audeze mentions that they use them during the development phase : “We use in ear probes (Etymotic Research
ER-7C Probe Mic System, series B) to get a sense of the frequency response at ear drum and it also turns out to be the best way to capture one's HRTF” (
https://audiophilestyle.com/forums/...-lcd-4z-review/?do=findComment&comment=988783).
Basically, it’s mostly used for research purposes, or by at least a few companies to assist headphones development (I’d be surprised if Sennheiser isn’t using such a device at times).
There’s a good reason why it’s used so little : the tube has to be inserted at a specific depth into the ear canal and, in general, as close as possible to the eardrum (in the above study audiologists were tasked with positioning the tube’s orifice at around 2mm from the drum). Then, it has to stay put for the entirety of the test. The entire system has to be calibrated to compensate for the effect of the tube. In other words it’s a royal PITA apparently.
Well I’m dumb enough that this wasn’t going to deter me
. I’ve started on the path to make my own DIY probe microphone.
DIY probe microphones are nothing new : David Griesinger (
http://www.davidgriesinger.com) already made quite a few of those and I’m indebted to him having done most of the preliminary leg work. More on that here :
IMPORTANT NOTE : if you’re ever interested in trying your hand at it,
GO SEE AN AUDIOLOGIST FIRST TO EXAMINE YOUR EARS. It’s very important to
make sure that inserting tubes in your ears is
not going to cause detrimental problems. They can also assist with helping you determining the length of your ear canal and place markers on the insert tubes to help you hit the mark.
For now I’ve focused on a “proof of concept” probe (meaning that it’s a quick and dirty work - by quick and dirty I mean that it’s already taken me quite a few dozen hours, as I'm not making a straight copy of David's work as my ultimate goals are quite different) with one simple objective : if I select headphones which, once mounted
on my own head (and it will be different for you), exhibit, according to my ears, peaks past 4khz or so that are sufficiently audible and constant across multiple reseats that I can confidently note down their location when manually running sweeps, can I get FR measurements that will accurately
locate these peaks ?
In other words,
if I note down “peak at 7300hz”, can I get measurements that will show a peak there ?
This is where the HD560S comes in (*drum rolls* tadaaa !). It has a lot of qualities for that task, for me :
- On my head, it produces a number of very easily audible peaks past 4khz that are very constant across multiple reseats, for both of my ears. I can note them down with a lot of confidence.
- It’s a Sennheiser open HP, ie channel matching can be presumed to be very decent - and so far the limited measurements I’ve made seem to bear that presumption. Same applies for sample to sample variation, we can presume that it's decently low.
- None of the measurements I’ve seen of the HD560S, regardless of the test rig, locate peaks past 4khz in a way that’s consistent with what I hear (and therefore that makes all of them not that useful to EQ the trebles), with one exception (the 4300hz peak, which seems to be an intrinsic resonance of the HD560S), which could possibly make for an interesting “anchor”.
Unlike my previous post I’m not going to bore everyone with the methodology, because, well, unlike the in-concha mics where it’s an easily obtainable, off-the-shelf device, requiring only a few validation and calibration steps to fully characterise them and make sure that I’m not over-interpreting the graphs, this DIY probe is very much a work in progress. I’m also quite far from reducing tolerances to a level I’m satisfied with.
But today I’ve arrived at a point where I’m starting to see measurements that are
starting to reflect what my ears are actually hearing, and learnt / confirmed a few interesting things, so why not share ?
So, here’s the location of the (obvious) peaks or dips I’m hearing from the HD560S, while manually running sweeps, as I noted them down (give or take a hundred hz or so) :
- Peak at 4200-4400hz
- Peak at 7200-7400hz
- Dip at 8000hz or so
- Peak at 9100-9400hz
- Dip somewhere around 10000hz
- Peak somewhere around 11000-11500hz
And here’s a few quick “proof of concept” measurements I’ve taken with my DIY probe wearing the HD560S, on the right ear, progressively pushing the probe deeper into my ear canal (
while not moving the headphones - the probe I'm making is meant to that effect).
IMPORTANT NOTE : just like with in-concha mics,
THESE RESULTS ARE NOT VALID FOR YOU. They’re just my attempt at illustrating how the HD560S actually interact with
my own anatomy and the resulting FR curve.
It's important to note that the relative calibration of the probe isn't as good as I'd like it to be. I'm not happy with the current results vs. some of my other mics, but I think that the most egregious peaks / dips have been neutered.
The brown trace is quite close to the entrance of the ear canal. The pattern that emerges is not too dissimilar from the sort of measurements you can do with in-ear mics albeit with more energy around the ear canal gain area (2000-4000hz).
The red trace is deeper into the ear canal. I don’t know exactly at which distance it was from the eardrum, but I suspect : not close enough
. I stopped after a while as I was getting tired and didn’t want to risk going further for now.
What I like to see however, is a pattern that corresponds pretty well to what I was expecting, even if not all peaks and dips manifested themselves quite as sharply as I’d like (but I think I know why).
The reason probe microphones have to be inserted deep into the ear canal is, I believe, because of standing waves inside the canal. I’m not sure but I believe that the reason you see wild swings in the response the deeper you go, and not a consistent, smooth progression of the FR curves, is because of them. This is why, for example, while four of the five traces show the 9100-9400hz peak quite accurately, one of them suddenly shows a dip in that area (blue color). It’s also possibly why only the red trace starts to show an accurately located hint of the (normally very audible) 7200-7400hz peak (or maybe it's the perfectible probe calibration, at least to some degree).
Another positive and surprising aspect is how decently representative the probe is
below 2khz or so, compared to my in-concha mics. I was expecting plenty of troubles in that area and a poor match, similar to the one experienced by that user with the ER-7C :
https://translate.google.com/translate?hl=en&sl=ja&tl=en&u=http://sonove.angry.jp/ER7Ctest.html&sandbox=1
It's probably not accurate, but it's good enough that it shows similar features as the in-concha mics (for example the small kink in the FR curve of my HD560S
on my head around 630-750hz).
To make sure that this wasn’t a fluke I’ve also made quick measurements on other headphones. So far, not quite right, but not so bad either.
So what’s next ? Well, either
going shallow, or
going deeper .
The “near ear canal entrance” measurements seem to show quite a few peaks accurately located, particularly past 8000hz or so. This is quite consistent with some of the papers I’ve read about such measurements. But they also don’t represent well what happens in the 4000-8000hz range, and I’m not sure that a simple,
constant transfer function can really work (this will be an interesting thing to investigate - there's a reason I'm designing the probe to be slide-able).
So, I’m going deep. But it will take months to refine the probe and the methodology to ensure decent repeatability, decent positioning accuracy, and effective
relative comparisons.
What's the end goal ? Well, I'd like to learn a bit more about the way my own ears modulate the FR curve and find a way to get a guiding hand for equalisation past a few kHz, something I nearly always want to do as it's quite obvious that most headphones don't interact with my ears in a way that's similar to speakers,
at all. I find third party measurements not superbly useful in that range to help me EQ headphones (as said in my previous post, even after EQing headphones to the same target, they still show easily audible differences in FR, and still vary in very inconsistent ways - particularly past a few kHz), EQing by ear is a b*tch, and I'm not fully convinced that microphones placed at the entrance of my ear canal will suffice. Let's see where it goes
.
In any which way, the HD560S now officially are my favourite headphones for experiments !