wrote on Tue, 28 Jun 2005 15:32:59 GMT in
rec.audio.tech:


Before we get into the specifics of your case, let me ask you,
in all seriousness, why do you want to do this?

Well, I got myself some speakers recently and they sounded like they
had too much bass. That's what I thought. I did find a review of them
on tomshardware which had a frequency response graph and it showed no
particular problem with the bass, it was roughly the same as the
treble. So I got curious and thought maybe I'll try something like
that myself and see what comes out. I knew that I wasn't going to get
the most accurate results, but I thought the results would be
interesting, particularly checking against other speakers I have
laying around here.

First problem: "several frequencies." Do you mean stepped sine
waves? Narrow band pink noise?

I used this: http://www.dogstar.dantimax.dk/testwavs/3stepoct.zip

Second problem, what kind of mic? This has a profound influence
over the usability of the results. For reliable results, just
any ol' microphone won't do.

Well, I knew of course that my mic is a cheap one, but I have to
admit that the recorded file sounded the same as what my ears were
telling me, and that's all that matters I guess. I used the same mic
on the other set of speakers, too.

and place it close to a speaker,

Third problem: depending upon what you mean by "close," your
results are going to be severely skewed by a number of proximity
effects that make such measurements suspect under a number of
conditions.

Why? What could those effects be?

Fourth problem: under what acoustical conditions are you
measuring? The existance of noise plus the influence of nearby
reflective boundaries and room-reletaed resonance can have a
tremendous influsence over such measurements.

Under my normal listening conditions, close to my computer. I did
turn off the AC though.

Fifth problem: such a display yells you nothing about frequency
response. Assuming the data you recorded is useable, a linear
display of amplitude is next to useless. FOr example, take the
differenence beween the loudest possible signal and one half as
loud: that's 1/4 of the height of the graph and corresponds
to a level difference of 6db between those to levels. But that
smae 1/4 height difference could correspond to a difference
of 10 db, if it's netween 25% and 75%, or it could correspond
to a difference of 34 dB if that 1/4 height is between 1% and
51%!

Ah, and this is one reason why I posted here, so I could get help.
What software would give more useful displays?

My understanding is that the recording should be as close as
possible to the test file that I use as input.

No, not necessarily.

Why?




Thanks for the reply, I'm not trying to say that my method is perfect
or anything, just to learn some stuff and maybe evaluate my speakers
(or my ears) better.