Evaluating the accuracy of my new simulator [UPDATED]

[Updated on 09/27/2012: Updated with a brand-new Phonak PFE 112's frequency response plot, super-imposed with data from the manufacturer and Tyll. My calibration effort with the new simulator is finally complete here]

[Updated on 09/23/2012: Added a time-domain comparison with data measured with B&K 4128C dummyhead due to much confusion arose with AKG K3003's waterfall plots.]


Frequency response comparison:

I've previously discussed the acoustic input impedance deviation of my dummy head's ear canals. They were under my tolerance limit, but I had to break open EURI in order to get a direct access to the occluded ear simulators. That is why I got an IEM-measurement only IEC 60318-4 simulator, and measured K3003 with it.  As suggested by Inks, I successfully insulated all the acoustic leak with a hot glue, and the result seems quite promising.

NO PRE/POST EQUALIZATION APPLIED

Compared to Tyll's IEC 60711 simulator (HA HMS II.3) using Grado GR10:

Green/Red plots are mine 

IEC 60711 and IEC 60318-4 are pretty much identical in terms of acoustic input impedance, so there is nothing special to comment on. You may ask, "why do you use Tyll's data when you don't even like Innerfidelity?" Hey, I just do not approve Head Acoustics' compensation figure, and it does not mean I hate Tyll and his works at IF. I respect a visionary, and he's a visionary, who first made the measurement data available to the public while running Headroom.

Compared to Etymotic Research's ASA/ANSI S3.25 simulator (Knowles DB-100) using ER-4P

Green/Red plots are mine

As expected, a Zwislocki coupler demonstrates less ear canal resonance @ 3~4 kHz. Not quite sure why I am getting a sub-bass response deviation in contrast to the Zwislocki data. There is absolutely no leak AFAIK. All of these should have something to do with either Zwislocki's smaller centre volume, or Knowles' manufacturing deviation criticized by Per V. Brüel, the founder of B&K.

Compared to Tyll's IEC 60711 simulator (magenta) and Phonak's IEC 60711 simulator (gray):

black plot is mine

Again, my data and Tyll's data are almost identical, but Phonak's has a slight level increment in the sub-bass frequency range. Same type of bass boost has been previously observed with Etymotic Research's official measurements as well, so I am afraid this is a manufacturer-manipulated equalization, in order to make their products look good on a paper. (Or could it be possible that Phonak is actually using a Zwislocki coupler? Possibly so, due to the particular ear canal resonance characteristic shown above. A Zwislocki coupler has less ear canal resonance.)


Waterfall comparision:

B&K 4128C data:

 EURI's data:

The plots pretty much speak for themselves, and EURI's time-domain characteristics are up to industrial grade. It is also important to note that the charts on the right are analyzed with the reference parameters used for my general analysis data. (On REW, plots are normalized to 90 dB @ 1 kHz, 25 dB top-down range) Compared to the balanced-armature driven ER-4P, dynamic earphones, such as CX500 and K3003(woofer unit only), have a lot more residual sub-bass decay.