Sound accuracy

Every Dolby theatre that has received THX qualification has had the sound system adjusted so that the  "eardrum pressure" response on the KEMAR manikin falls within a tight tolerance of the "THX curve" here .

This curve contains the roughly 15 dB boost at 2.8 kHz characteristic of the external ear "horn" and resonance of the human ear.  Any high-fidelity earphone must reproduce the same boost at the human eardrum or KEMAR's eardrum-position microphone.

Sometimes it is useful to make a binaural recording of the sound of an auditorium with a KEMAR manikin.  The problem is that when you play that recording back over a good loudspeaker, the response at the listener's earphone contains a 30 dB boost:  15 dB from KEMAR's ear and 15 dB from the listener's ear.

This problem was addressed by the writer some 28 years ago  in a paper describing an extremely simple "diffuse-field inverse filter" which remove's KEMAR's boost leaving only the listener's natural 15 dB boost.

We have recently used the same filter to make 2nd- and 3rd-generation recordings of the output of various earphones placed on KEMAR.

The idea is simple.  One test of a good copy machine is that if you make a copy of a copy of a copy, the final copy should still look like the original.  The copies of a Monet seascape painting were made on the writer's inexpensive hp Officejet printer-copier set for maximum quality.  Clearly the copy of the copy did not reproduce the original very well.

In the 1960s, Edgar Villchur used a similar approach to compare the accuracy of loudspeaker responses.   By recording the output of a loudspeaker and then playing that recording back through the same loudspeaker, he demonstrated that a loudspeaker which had passed live-vs.-recorded tests (see future Posting) could "reproduce itself" quiet well:  The sound of the original recording and the copy were very similar.  A loudspeaker with a ragged frequency response, however, could not reproduce itself at all well, as he demonstrated.

(This was first demonstrated at a 1961 NY local AES meeting, and later described in Villchur E (1962) “A method of Testing Loudspeakers with Random Noise Input” J Audio Eng Soc, Oct 1962, 96-99.

For the 2006 HeadFi convention, I made use of a similar demonstration with earphones.  The frequency response of the original and 2nd- and 3rd-generation copies of the ER-4S is shown below.  The sound of the original and the copies can be heard here. Not surprisingly, listeners found that both copies sounded much like the original.

|

|

|

|

|

|

In contrast, a popular earphone with exaggerated bass and treble response did not copy well.  The exaggerated bass and treble can be heard.  Listeners heard almost no mid-range response in the second-generation copy.

For those with a KEMAR manikin, Etymotic makes the ER-11 microphone-preamplifier combination, with self-contained diffuse-field-inverse (DFI) filter.  This DFI filter is designed for sound-field pickup.  Because of the 5 dB boost at 10 kHz built into most CD recordings, the ER-4S earphones have a 5 dB rolloff*. The DFI filter used for the above demonstrations included an additional 5 dB  "CD/LP boost" in its response.  Make certain you talk to one of the engineers before purchasing a pair of these.

    *Note:  The original ER-4B earphones did not contain this rolloff, and many customers complained they were too bright.  For use with true binaural recordings, however, the ER-4B earphones are still appropriate.