DIY Headphone Measurement Rig

In this thread I want to share my “journey” towards my personal rig and I hope to get some guidance about things I may have missed, misunderstood or just general tips and bits. Since theres a ton of text I cut the up the topic for people to just read what they are interested in.

The Idea

Since making my own headphones I quickly realized that some kind of measurement system was needed to be able to quickly see how things “looked” without needing to spend multiple songs to get a proper grasp on changes that I made.
The first crutch was a crude “flat plate” style system based on my receivers room correction microphone stuck under some foam. That worked fairly well for the start but as we’ve learned the human head is not just a EQ profile added on that and you want an ear and an inner ear to more closely model the perceived sound.

The next step was just printing an ear model from the internet and sticking that on there. Looking at the influence of the different ear parts it seems to work fairly well to implement the Concha and Pinna effekt.
But being a 3D print it had the limitation of being a stiff piece of plastic with a fairly rough outline.

So a new System was “needed”. From what I read and gathered it seemed a clever idea to get a 711 Ali coupler and cast my own ear out of silicone, then calibrate that and have something alined to the current research.
I aproached @Resolve about the Idea a while ago and he had great insight why that would probably not be the best Idea for me, due to having no reference point and the inherent limitations of the 711 couplers. His recommendation for In-ear mics didnt quite convice me since I can see myself making Iems in the future.
So in the end I had a few weeks before the next semester started and curiosity got the better of me so I started the process.

The Process:

The Model

First I needed a 3D model of my ear / head to make a mold. For that I simply used a free app and had a friend “scan” my ear in good light to give the software a good chance for triangulating the depth of the various ear features. After importing it into Fusion I had to fix the holes in the surface and recalculate the faces to further process the model. Then I went into blender to smooth bumps it made from beard hair aswell as deepening some missed troughs inside the ear aswell as deepening the crease above the ear.


Having a good surface I then went back into Fusion to cut out a circular section for the final model. I made sure to have a good big chunk to also model the seating on my head and the seal issues derived from that. To attach the 711 I tried to model what the 5128s inner ear looked like and have the it screwable onto the coupler. That wasnt that clever but we’ll talk about that when we get there.
Lastly I added a lip to secure it to a rig and added the spout for filling.

Casting

To get the model castable I made a negative of it and cut it up into 3 pieces. On bit as the inner ear and the pinnae, then the outer shell split in half at the spout. After printing that mold I added some foam rubber at the parting line to seal better since my printing orientation was suboptimal. After clamping it shut and mixing the silicone rubber I poured slowly making sure to let it run down slowly to trap less air inside the mold. 24 hours later I was able to pull the mold apart and was greeted with a pretty nice casting. I had to clean up the parting line along the outer ear with a scalpel.

First Steps

To get an idea wether the Ear model was usable I tried sticking an Iem with solid squig backing into the rig in its normal form and into the ear. Ideally there should be little difference.


That mount doom is no little difference tho. Seeing how it resembles a very loosely inserted Iem with a low resonance peak I concluded that my attempt to mimic the 5128 was indeed stupid. I then cut off the orignal flange I molded and cast a quick sleeve to stick the coupler without the nozzle right on the back of the Ear model. That pushed the resonance back towards 8k where people are used to see it with Iems.
Lastly I quickly slapped together some wood and printed clamps aswell as a support for the headband.
That resulted in the Rig in its current form. The sizing is as closely as I could to my own head from width and height.

The Calibration

To be inline with the cool kids my goal was to get a DF target from that system. Looking online was a bit frustrating since I didnt find that much (not paywalled) information about how to do that. What I found was measuring the free field response at 0°,90°,180° and 270° with no height change and then averaging the responses. I didnt feel like the 270° ajacend measurements were that usefull and the resolution of 90° was fine enough on the ear side. Having recently finished a set of speakers with a calibrated measurment mic I decidec to use the known response of that speaker to measure from 0-180° in 10° steps with a gated measurement inside to negate reflections. That cut off the information below about 400hz but gave a clear response above that.

The Orange AVG already looks pretty promising but lets work on that. I then divided the response by the speaker on axis measurement made from the same spot by the calibrated microphone and then merging the Couplers calibration at 400hz to get a “linear” bass response.
This should yield something close to the DF target for the rig. Including a 8dB tilt for easier comparison to Resolves Gras Target gives us this comparison.


The green curve is taken from the paper where I took the DF measurement technique from. The “Resolve DF” Target is taken from one of the Sundara measurements. While the targets share a lot in common there are a few differences. Resolves shows a lot more lower and mid treble, and my rig shows some waviness from 400 to 1.5k which probably is due to the gated measurement limiting resolution. My Rig also shows a dip at 9k similar to the Gras rigs although its not visable due to the smoothing in the Papers response.

Hearing Tests

Now having a “Absolute Truth™” I EQ’d many of my headphones to the target and tried to figure out whether it actually gave a neutral result and sounded good to me.
What stood out mostly to me is that it always was a tad dark yet shouty to my ears. Reducing the ear gain Peak at 2.8k by about 2.5dB and then shelving the treble by about 2dB from about 5k was a lot nicer for me. Whether that sounded better was just preference or what I was used to and thus expected I am not sure.
Atleast it game me an indication why I liked one of my DIYs so much without having something to point to:


Otherwise it gave me an FR that sounded very clear without being unnatural or fatiguing to me. So judging from my point of view 10dB DF seems like a good way to go for neutrality.

I hope I havent forgotten anything here yet.
Curious what the experts here have to say about this endeavour.
And with that said I will now go to bed and apologize to Resolve for ignoring his advice :upside_down_face:

5 Likes