One octave spectrum analysis

[Mybabejc]

Hi. I need to produce a one octave spectrum analysis of white noise and speech babble transduced by a loudspeaker and headphones. I need to find a way to keep the input levels as close to a constant as possible so that any differences between the transducers will. E related to frequency response and not to input levels.
Please help. Thanks!

 

[crutschow]

Not enough info.

So what is the required amplitude accuracy and frequency resolution (bandwidth) of this analysis?

 

[Mybabejc]

There is no required amplitude or frequency resolution..

 

[crutschow]

There is no required amplitude or frequency resolution..

Then just use a one-octave band-pass filter.

 

[audioguru]

An AGC circuit (automatic gain control) circuit keeps levels at almost the same level. Then you will not see any differences betweeen different transducers.

One octave of speech will not show much. 80Hz to 1kHz will show vowels and 1kHz to 15kHz will show consonants.
The one octave from 80Hz to 160Hz or from 160Hz to 320Hz will just be a grunt or a groan.
The one octave from 5khz to 10kHz will just be whenever there is a consonant which might be an "SSSSS".
There are much better ways to see the differences between transducers.