Measurements on the speakers

Mitch Bradley wmb at
Fri May 11 01:17:56 EDT 2007

I have done some speaker measurements with an external mic.  It's just 
an ordinary desktop mic, not my fancy measurement mic, but I checked it 
with a good speaker and the desktop mic is fine for this purpose.  It 
measures less than 1% total harmonic distortion with the good speaker, 
and that is a typical value for a speaker.

There is a *lot* of unit-to-unit variation in the XO speakers.  Some of 
them have very audible distortion at fairly low listening levels, and I 
have measured upwards of 25% THD on some of them, at normal levels.  If 
you drop the levels way down, most of them sound okay.

There is also a substantial effect from the XO case itself.  You can 
markedly affect the sound, for better or worse, by holding the case in 
various ways.  In some circumstances you get very annoying buzzes from 
case parts vibrating against one another.

The unit-to-unit variation is so large that it is difficult to draw firm 
conclusions about the "average" speaker performance.

Jean-Marc Valin wrote:
> Jim Gettys wrote:
>> (Some of) the microphones in B2 are really lousy: B3 should improve the
>> situation on the mikes quite a bit.
>>                          - Jim
> From my experience, even cheap electret microphones (<<$1) tend to
> perform much better than loudspeakers, both in terms of frequency
> response and harmonic distortion. The only thing I think is important --
> if it's not the case already -- would be to use figure-of-eight
> directional microphones so that the loudspeakers end up being in the
> (theoretical) null of the microphone. That would minimise the direct
> speaker->mic path.
> 	Jean-Marc
>> On Fri, 2007-05-11 at 11:19 +1000, Jean-Marc Valin wrote:
>>> Hi,
>>> I just made some impulse response and distortion measurements on the
>>> speaker->microphone path. The measurements are shown in figures
>>> Both figues have the same data. The only difference is the frequency
>>> axis: first figure has the frequency of the harmonic on the horizontal
>>> scale, while the second one has the frequency of the fundamental that
>>> caused it. e.g. for a 1 kHz tone, you need to read the harmonics at 2,
>>> 3, 4, 5 kHz in the first figure, while for they're all aligned at 1 kHz
>>> in the second one. (second figure is easier to read)
>>> The main things that strike me from these figures is that there are
>>> resonances in the fundamental around 3-6 kHz and which cause quite a bit
>>> of harmonic distortion. The distortion is also especially significant
>>> because it happens at a "normal" volume. It's measured using a sine
>>> sweep from 40 Hz to 24 kHz as explained in:
>>> The sine waves were nearly full amplitude and the alsamixer volume for
>>> Master and PCM were both set to 61.
>>> In case anyone's interested, the raw data is available at
>>> and requires the paper above to
>>> interpret. I'm still not sure what can (or can't) be done to improve the
>>> situation. Any thoughts? Oh, and I was just wondering, what kind of
>>> microphone is the BT2 using (omni, figure-of-eight, cardioid, ...)?
>>> Cheers,
>>> 	Jean-Marc
>>> _______________________________________________
>>> Devel mailing list
>>> Devel at
> _______________________________________________
> Devel mailing list
> Devel at

More information about the Devel mailing list