[Scott Dorsey]

Adrian Tuddenham wrote:
Make up a stack of thick hardcover books about the same thickness (and
density!) as the space between my ears. Allow the headphones to clamp
themselves over the stack and slip a couple of electret pressure
microphone capsules on thin wires into the cavities. Connect the
headphones to an audio signal generator, connect the mic capsules to a
small recorder with suitable phantom power (Tascam DR-04) and read the
sound level off the recording level display. The measurements are good
to a dB or two - and are going to be far more accurate than the accuracy
of most budget headphones (and some expensive ones).

The thing is, you don't want flat response in this environment.

First of all, this doesn't necessarily give you an accurate measurement
because the shape of the ear canal dramatically affects response above
around 1kc or so. Which is why we have things like the Zwislocki coupler
for measuring earphones.

Secondly, for sealed headphones, the volume inside the cups (which most
unfortunately includes the volume inside your ear canal) changes the
response on the low end substantially. (Unfortunately this means that the
measured response of headphones on my head will not be the same as on your
head.)

But the real big deal is that you don't _want_ flat on-axis response from
a headphone driver. In normal life, you're listening to sources that are
in front of your head, 90 degrees off-axis, so you want the response to
simulate that of an off-axis source. Which.... means an upper midrange
dip somewhere. But exactly _where_ depends on a person's head.

And this is why headphones that measure and seem flat to one person may not
to another.

If there are any acoustic artefacts that need investigation, they will
change if the mic capsule positions are changed or the air cavity shape
is altered by compressing the earpads. If they don't change, then they
can reasonably be attributed to the headphones themselves

This is a fair way of identifying issues that result from the first two
effects I described... but the third one is the killer.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."