[takato14]

I have quite a few rare (and common) headphones so I'd like a measuring system to assist in benchmarking their performance, without having to send them out to Tyll or someone else to test them for me.
 
A couple of members I know use REW V5 to do this but I don't know what other software is available, nor what microphones and/or dummy heads are available, etc. My searches on google have been baiscally fruitless so I thought it wouldn't hurt to ask here.
 
Thanks.

 

[anetode]

For relatively cheap kit you'll be looking at,
 

1. Calibrated mic - http://www.parts-express.com/pe/showdetl.cfm?partnumber=390-801
2. USB mic preamp - http://www.amazon.com/exec/obidos/ASIN/B001EW5YQS
3. homemade coupler/artificial ear - there's an art and a science to making these and the pro ones available on the market are prohibitively expensive

 
 
Here are a couple of threads which might be of some help:
http://www.head-fi.org/t/482386/building-a-headphone-measurement-lab
http://www.head-fi.org/t/636885/headphone-testing-to-pinna-or-not-to-pinna-that-is-the-question

 

[stv014]

What parameters of the headphones do you intend to measure exactly ? If you are using utilities from my package, I could post a detailed description of how to measure any of the following:
- frequency response using MLS
- frequency response using sine sweep
- THD vs. frequency
- THD vs. level
- impulse response from MLS
- group delay vs. frequency from impulse response
- CSD from impulse response
- impedance/phase vs. frequency

 

[takato14]

Quote:

What parameters of the headphones do you intend to measure exactly ? If you are using utilities from my package, I could post a detailed description of how to measure any of the following:
- frequency response using MLS
- frequency response using sine sweep
- THD vs. frequency
- THD vs. level
- impulse response from MLS
- group delay vs. frequency from impulse response
- CSD from impulse response
- impedance/phase vs. frequency

I was planning on doing a FR with a sine sweep, two square waves at 30 and 300 Hz, and a CSD plot. 

 

[jude]

Quote:

I was planning on doing a FR with a sine sweep, two square waves at 30 and 300 Hz, and a CSD plot. 

 
I made a post recently that outlined some of the measurement rigs from which graphs and data are routinely posted here. Here's the link: Headphone Measurements

 

[stv014]

Quote:

I was planning on doing a FR with a sine sweep, two square waves at 30 and 300 Hz, and a CSD plot. 

 
Actually, all of those (FR, square wave response, CSD) can be derived from the impulse response if you ignore non-linear distortion (which, with good headphones, is probably only a potential issue for the 30 Hz square wave at high SPL). Of course, it is easy to generate actual square waves, record them, and view the waveform in an audio editor. Using a sine sweep for frequency response testing has the advantage that it also gives a THD vs. frequency graph.
 
To measure the impulse response, you need an MLS signal, which you can generate with my 'testgen' utility, for example:
 
mls 0 25 1000010000000100010 0.3 0.3
 
This example works best at 48 or 44.1 kHz sample rate. You can actually delete the first few (<10) seconds of the resulting WAV file, because it will not be used in the impulse response output. The following command extracts the impulse response:
 
convolve.exe mls.wav ir.wav -inv=1 1000010000000100010
 
The second impulse - which will also have the highest amplitude if you truncate the beginning of the test signal - in the output file is the impulse response, you can use any audio editor to remove the unneeded parts of the file. To get a correct frequency response above 1-2 kHz, the recorded audio also needs to be equalized (this is not easy to get right with a DIY artificial ear). Note that the above command for "decoding" the MLS only works correctly if the original and the recorded file have exactly the same sample rate; that is, the DAC and ADC in the test setup should preferably be on the same physical device and share the same clock, otherwise pitch correction may need to be applied to the file.
 
The FR can be displayed with a simple FFT analysis of the impulse response. You can also convolve square waves with the IR to get the square wave response, if you did not record any actual square waves. For CSD analysis, you can use my 'csd' utility (I will explain the details of the usage and parameters later), or other programs.

 

[ultrabike]

Quote:

I have quite a few rare (and common) headphones so I'd like a measuring system to assist in benchmarking their performance, without having to send them out to Tyll or someone else to test them for me.
 
A couple of members I know use REW V5 to do this but I don't know what other software is available, nor what microphones and/or dummy heads are available, etc. My searches on google have been baiscally fruitless so I thought it wouldn't hurt to ask here.
 
Thanks.

 
For mic I use an EMM-6 which Anetode recommended above.
 
For USB audio interface I use a Focusrite 2i2: 
http://www.bhphotovideo.com/c/product/822508-REG/Focusrite_SCARLETT_2I2_USB_Scarlett_2i2_Portable.html

 

[yuriv]

Holy thread resurrection Batman! But I couldn’t find a more suitable place to post this info—the result of my frustration with the Jaybird X3 Bluetooth IEM and its equalizer app, MySound.
 
With most headphones it takes me only a couple of minutes with Sinegen and a parametric EQ to get it close to the tonal balance of my calibrated speaker systems. The Jaybird X3 has a built-in DSP and it remembers the last EQ settings, allowing it to work system wide with any paired device. But its EQ app, MySound, is fiddly and imprecise. It has almost useless markings for the horizontal (frequency) axis. Worse yet, the vertical axis doesn’t have units:
 

The Jaybird MySound app with 5-band parametric EQ
 
As it happens, I have this cheapo Dayton iMM-6 measurement mic with me and some ¼” inside diameter vinyl tubing lying around, so…
 
    
 
 
Lol. It kinda fits like a glove. This idea isn’t new. For example, some folks online have tried to fit vinyl tubing with the now-discontinued Panasonic WM61a mic capsule:
 

More information here: http://www.johncon.com/john/wm61a/
 
They mention that the average human ear canal is 7 mm = 0.276” in diameter. The vinyl tubing’s ¼” = 6.35 mm inside diameter might be close enough for quick and dirty measurements.
 
I used an iPad Pro and an iOS app, FFT Plot – Real Time Sound Frequency Analyzer by ONYX Apps. It’s $5 but it also comes as part of ONYX Apps’s $15 Spectrum Analyzer bundle. Here’s what white noise looks like in FFT Plot with the Etymotic ER4PT and the iMM-6:
 

Etymotic ER4PT with small frost tips (ER38-15SM) raw response on iPhone 5S
 

Etymotic ER4PT with small frost tips (ER38-15SM) raw response on iPad
 
 
As you can see from the raw measurements, this ain’t no IEC 711 ear simulator. When a cheap Chinese unit becomes easy to order, maybe I’ll spring for one. But for now, this setup might be useful enough for some things. The Dayton iMM-6 is only $14.50 at Parts Express, and if you don’t want to buy a $5 app, you can use Room EQ Wizard (REW):
 

REW measurement of the same thing – logarithmic sine sweep
 

Real-time spectrum in REW using white noise
 

Some of the equipment used to make the measurements in REW
 
 
 
Now for the Jaybird X3 with no EQ:
 

Jaybird X3 raw measured response, medium silicone tips
 

Jaybird X3 raw measured response, medium Comply foam tips
 
It looks like the Comply foam flattens the big peak around 5 kHz—one less thing to EQ, which probably isn’t a good idea anyway because the frequency of those resonances often depends on the insertion depth.
With white noise playing, you can tweak the EQ in the MySound app and see right away the resulting spectrum on the screen. It’s like using the EQ to “sculpt” the sound to the frequency response that you want. Despite the fiddly controls in MySound, I managed this measured response:
 

Jaybird X3 raw measured response,  medium Comply foam tips, EQ’d, first attempt
 
If this response were a from a 711 coupler after the Real-Ear-To-Coupler Difference (RECD) had been taken into account, this EQ might make more sense. So the current rig has more high-frequency roll-off. With music playing, the result, indeed, sounded too bright.
 
Take two:
 

Jaybird X3 raw measured response,  medium Comply foam tips, EQ’d, second attempt
 
This is starting to sound pretty good. I’ll double-check with Sinegen later to see if it comes close to my speakers.
 
Until those Chinese couplers become more readily available, this inexpensive setup can be useful for making comparisons or for coming up with good starting points for further equalization. Just be aware of its limitations.
 
Possible improvements: Heat up and stretch one end of the tube to shape it like a funnel. Or maybe you can use a 3D printer to print that 7mm ID tube.

 

[WhiteKnite]

Oooh this just gave me an idea.  I have been trying to figure out how to use my UMIK-1 calibrated measurement mic to test iems.  I have a set of CIEMS on the way and I am thinking I could cover the canal section and the end of the UMIK (positioned properly) in thin film like seran wrap and then recreate my canal with 'green stuff' epoxy putty.  A thin layer of sticky tack would help to make sure I get a good seal but not sure if that would even be necessary, maybe just to make sure the outer part of mic end is perfectly sealed.  Worth a try anyway, I have plenty of spare putty and sticky tack from modeling.

 

[yuriv]

@WhiteKnite: Do it! It’d be nice if you and others with measurement mics would try the same thing to see how well the results can be duplicated.

---

Here’s a follow up to my last post. A few of my friends and I decided to measure a few of our IEMs using the cheap DIY rig built around the Dayton Audio iMM-6.
 
The ER4PT is easy to fit into the ¼” inside diameter tube with the small frost tips. But many IEMs have trouble getting in. So I fitted a 3/8” inside diameter, ½” outside diameter clear vinyl tube at the end of the smaller tube. It’s an airtight fit. This was described in the article I linked in my previous post. But I went further: I fitted a ½” inside diameter tube, 5/8” outside diameter tube at the end to accommodate IEMs with larger sleeves like the Bose SoundTrue Ultra In-Ear and its StayHear+ tips.
 

The Dayton Audio iMM-6 and clear vinyl tubing coupler

The DIY coupler and the Bose SoundTrue Ultra In-Ear

 
 
Here’s the result:
 

Bose SoundTrue Ultra In-Ear frequency response measured by DIY coupler
 
The peak resulting from the half-wavelength resonance occurs at 5 kHz. Compare this with the measurements at rtings.com and Clarity Fidelity, where the peaks happen near 6 kHz and 7 kHz, respectively. So I must be using a longer tube than theirs. With Singen, I found that I hear that peak at 6 kHz when using the Bose SoundTrue Ultra, which has a consistent insertion depth, if you can call it that. The SoundTrue Ultras don’t really go inside the ear canal. Their StayHear+ tips just cover up the opening, making them the shallowest-insertion sealed earphone I’ve ever used. So I shortened the coupler’s tube to match my hearing.
 

The DIY coupler after shortening the tube.
 
Here’s the new measurement:
 

Bose SoundTrue Ultra In-Ear, measurement after shortening the tube.
 
It doesn’t take long to take a measurement. Just play white noise and wait for the FFT to average the readings. It took less than an hour to produce the following raw measurements:
 

Brainwavz Jive, medium silicone tips
 

Brainwavz Jive, medium Comply S400 tips
 

LG Quadbeat 3, AKG-tuned version
 

LG Tone Ultra HBS-800
 

Monoprice MEP-933 (8320)
 

Panasonic TCM-125
 

Philips SHE3905, Comply S400 tips, small foam plug damper
 

Sony MH1. I hear Way. Too. Much. Bass. I wrote about how I dealt with this a few years ago.
 

Sony XBA-C10
 

Zero Audio Carbo Tenore ZH-DX200
 
The rig can also be used for quick and dirty distortion measurements. For example, the 3^rd harmonic distortion of the Apple IEM at 3 kHz is well documented—and now I can verify it:
 

Apple ME186LL/A, 1 kHz tone and harmonic distortion
 

Apple ME186LL/A, 3 kHz tone and harmonic distortion
 
As for the Jaybird X3, I suppose I could have EQ’ed it using my usual parametric equalizers, measured the result, then copied that using the unwieldy MySound app. For $99 street price, Jaybird should have done better with the app. The results are worth it though. I think I’ve got the EQ to the point where I have the preferred balance anywhere I go and with any paired device. So far I’m liking the sound better than any unequalized IEM mentioned above.
 
Sometime later, maybe I can figure out how to tweak this thing to get decent measurements from earbuds:
 

 

[castleofargh]

I've made a "tube" out of what I had left of some Rubson bathroom silicon thing because I wished to check if there were serious variations when the canal was more flexible. but there are just so much imperfections(little bubbles from the stuff being a few month old since I used it) that the variations could come from pretty much everything. inconclusive stuff is inconclusive.
I once tried to use an ear print I had(last time I made custom earplugs I ask to get the mold back) to mold part of my own ear canal and hope to use it for measurements, but the 2 materials just stuck to each other(silicon sticks to silicon... you don't say?) and I ruined everything because I'm stupid. then I tried with some leftover acrylic product for nails and it seemed to work pretty well but I didn't have enough, made a giant mess in the kitchen and broke the mold when trying to cut a hole in it to place the mic. so now I'm back to tubes because I'm still curious, but mostly I'm more about being cheap than curious.
 
anyway, we learn from failures, and boy do I learn! ^_^

 

[WhiteKnite]

While waiting for my CIEMs I figured I'd rig something up.  I used a piece of 3/8" ID tubing left over from PC water cooling.  Inside I fitted a metal sleeve to slightly reduce the ID, probably bringing it to around 8mm.  I also tried a 5mm sleeve which gave some very different results.  These tests only proved a couple of conclusive results, for one, there is absolutely no frequency response difference between my Modi -> Little dot I+ setup and my X-fi card -> O2 amp.  None... At all... despite having the "warmer" Mullard tubes in the LDI+.  Also ear canal shape and depth make a HUGE change to the sound.  I knew that was true to some degree, but seeing what a difference slight changes made to the measurements was a bit shocking.  I didn't have enough time tonight to experiment too much.  Here's a series of measurements that I threw together quick just for the heck of it.  THESE ARE HIGHLY INACCURATE AND EXPERIMENTAL.  I'm pretty sure the H3 had a poor seal, I need to change tips and see if I can get something that looks more like what I perceive from them.  Also these were all measured from the left channel only and smoothed 1/48.  
 
 
 
 
 

 

[zareliman]

Wow, so far loving this thread.
Now I'm encouraged to make my own measuring rig, they're not near as expensive as I thought.

 

[yuriv]

@castleofargh: Lol. Cheap and curious is good. So you’re back to using tubes. I’d love to see your setup and some of the measurements you’ve made with it. More data is good. Maybe there’s an idea there that I’d like to steal.
 
 
@WhiteKnite: That’s awesome! Yep, the electronics make a minuscule difference once you’ve taken care of things like impedance interactions. More interesting are the things that make a huge difference, like, as you’ve discovered, the geometry of the setup. Highly inaccurate you say, but I’ve heard some of those IEMs, and from what I remember, your graphs aren’t a complete lie.
 
BTW, your photo insipired me to dig up my old Behringer ECM8000 to see if I can get similar results. See my update below.
 
 
@zareliman: Yes, dooo eeet. $14.50 for the iMM-6, a few bucks for tubing (they have it at Home Depot if you’re in the U.S.), and no more than $5 for an iOS app. The iMM-6 also works on Android, and there are several free spectrum analyzer apps. This one called Advanced Spectrum Analyzer PRO shows up at the top of the search:

 
As you can see, the mic input on my 2013 Nexus 7’s headset jack has a lot of roll-off. Maybe it’ll work better on a flagship phone. Or another app might have customizable compensation settings. I didn’t look into it any further.
 
If you have a notebook computer with a decent headset jack, you can use REW, which is a free download. That setup works ok with a MacBook or a Windows PC, partly because you can compensate for the low-end roll-off with REW itself.

Dayton Audio iMM-6 coupler with Lenovo Thinkpad Carbon X1. Please excuse the poor lighting.
 

---

I decided to see if I could get the same results with an old Behringer ECM8000 measurement mic. Its barrel is much thicker than the iMM-6's but the mic capsule itself should be the same size. The ½” = 12.7 mm inside diameter (ID) tubing fits it well. I fitted a  ⅜” ID, ½” OD tube and a ¼” ID, ⅜” OD tube inside it, so that one end of the two inside tubes is flush with the front of the mic. The other end of the coupler looks like the one I made for the iMM-6. It looks like this:
 

 
Here’s the first measurements with the ER4PT:

 
There’s a deep dip around 10 kHz that I don't get with the iMM-6. These nulls can sometimes be caused by destructive interference—a reflection of the wave that cancels the original where it’s 180° out of phase. At 10 kHz, sound has a wavelength of approximately 34 mm. A quarter of that is about 8 mm; if I were to guess, it could be the diagonal distance from the edge of the inner edge of the ¼” ID tube tube to the inner wall of the 5/8” ID tube along the notches surrounding the ECM8000’s tip. I’ll have to double-check this to see if it makes sense.
 

Tip of ECM8000. Photo from http://www.coutant.org/behringer/
 
In any case, I decided to cut out a ring of earbud foam to place around the mic’s opening. Measurements with this in place lessened the depth of the notch around 10 kHz. So I tried it again with earplugs, which are made of a similar material as Comply foam.
 

Ring of earplug foam surrounding the mic opening.
 
Here’s the new measurement:

New measurement with foam ring around the ECM8000’s opening.
 
Now I’m getting something similar to the setup with the Dayton Audio iMM-6. Here are some sample measurements with the current ECM8000 setup:
 

Bose SoundTrue Ultra In-Ear, LG Quadbeat 3 tuned by AKG
I adjusted the length of the tube so that the frequency of the peak on the Bose matches where I hear it—around 6 kHz. The measurement above is similar to this one by Clarity Fidelity/Speakerphone:

Bose SoundTrue Ultra In-Ear and LG Quadbeat 3 Tuned by AKG measurement at clarityfidelity.blogspot.com
 
 
I’m getting taller peaks, though. Maybe there’s a way to adjust the acoustic impedance at the mic end so that it’s closer to that of a typical human eardrum. Also, the hump around 3kHz doesn't get as high. Obviously this cheap DIY rig is nowhere close to an IEC711 coupler. But it can be useful for making quick and dirty measurements for comparisons, and can be used to double-check what I'm getting with the iMM-6.
 
 
 

Brainwavz Jive, medium silicone tips, medium Comply S400 foam tips
 
 

Philips SHE3905, medium silicone tips, medium Comply S400 foam tips and small foam plug damper
 
 

Jaybird X3, medium Comply foam tips, no EQ
Sony XBA-C10
Zero Audio Carbo Tenore ZH-DX200
 
 
So far they’re pretty similar to the results with the Dayton iMM-6. Maybe the next project will be to get a silicone mold of a pinna to attach at the end of the coupler.

 

[castleofargh]

I use the vibro veritas for IEM measurements, instead of the IMM6. not because I find it to be a superior coupler, it's not. the IMM6 actually rolls off later on in the trebles. on the other hand my IMM6 rolls off in the low end(I tried to seal everything, it didn't improve) when the veritas is almost right without any compensation in the low end. overall it would be better to compensate the low end of the IMM6 in my case and use it for measurements. the plug doesn't work in my inputs but I get it to work with small crocodile plugs on the jack ^_^.
why I don't do that? I started with the veritas and have some measurements of IEMs I no longer have with me, so I decided to stick to the vibro veritas for consistency.
 
now if I was looking for more than very basic FR graphs, I would get an actual microphone I could use with my scarlett2i2. neither the veritas nor the IMM6 work with it and I'm as frustrated of it not working as I am of not understanding why. ^_^ it's too bad because the input could be so much cleaner than the little startech crap I use as input.
so in practice the IMM6 really helped me to set up a pair of speakers(and check if my original pair was ok). having it plugged into my cellphone and moving around in the room, now that's next level practical stuff. I loved it.
 
so here is an example of what I get with the vibro veritas and REW, I show er4sr because it should be very close to your er4, and xba-c10 because we have that one in common. and the hf5 because I always felt the C10 was a close enough parent to EQ one into the other(super cheap hf5 without isolation).

this one is real uncalibrated, uncompensated, non smoothed data
 
 
 
 
 
 

 this one is the same data compensated to get the er4sr to follow the graph on the certificate that came with it. it actually shows too much trebles based on my hearing for other IEMs, maybe if I spent more time testing for insertion depth I could get that right, but the er4 is a special kid when it comes to going deep and I'm not sure I go deep enough compared to the rest in an actual ear. I also struggle when it comes to comparing vented and sealed stuff. they don't measure the same variations I feel in my ears(often hearing way more low end than what I measure). 
 
 
 

and this is still the same data, but compensated to better resemble online stuff in general. to my ear at least, the xba-c10 really drops like a stone after 10khz, and that compensation seems to best represent how I feel(still raw though).
 
in the end what matters are the variations between 2 measurements and I don't bother with compensation when I'm testing stuff for myself(tips, filters, EQ, source impedance variations...) but when I show stuff online I usually try to have a general look similar to online stuff so that people who don't know how to properly read graphs end up being only slightly wrong ^_^.  I could warn about not making comparisons from graph of 2 different provenances 10 times a day, and still have people who will do it because it's simple.