Hi, 
I work on a convolution project  on a ADSP-TS101...
I convolute my audio samples by different impulse responses...
I have to compare these different impulse responses but the problem is
that I must have a Constant factor to scale all my audio outputs...

for the moment, I'm scaling my audio outputs by a factor from the
impulse response I'm convoluting...
Thus, the scale factor is different for each impulse response... But
this way, This is impossible to compare an impulse response with
another....

So how to have a constant factor....?

If you see what I'm trying to say..?
Thanks to show me the way..
Best regards
------------------------------------

On Thu, 17 Apr 2008, oli_renard wrote:

> Hi,
> I work on a convolution project  on a ADSP-TS101...
> I convolute my audio samples by different impulse responses...
> I have to compare these different impulse responses but the problem is
> that I must have a Constant factor to scale all my audio outputs...
>
> for the moment, I'm scaling my audio outputs by a factor from the
> impulse response I'm convoluting...
> Thus, the scale factor is different for each impulse response... But
> this way, This is impossible to compare an impulse response with
> another....
>
> So how to have a constant factor....?
>
> If you see what I'm trying to say..?

Not quite...

If the weights of all the coefficients of your convolution are summed up,
this should give you the correct scale factor for each result.  Taking the
same signal data and convolving it with the different impulse responses
and then multiplying by the corresponding scale factor does make the
comparisons correct.  You can not have a single constant.

But - what you can do is change all the coefficients so they sum up to 1.
If all the impulse responses have the same weight, then you can apply each
of them to the data and not have to work with a scale factor at all.  The
same data will then also give correct comparisons to all the convolutions.

Patience, persistence, truth,
Dr. mike
------------------------------------