• To: AUDITORY@xxxxxxxxxxxxxxx
• Subject: cascade filtering usign FFT
• From: sherin tech <sherin.tech@xxxxxxxxx>
• Date: Fri, 7 Apr 2006 18:26:46 +0530
• Delivery-date: Fri Apr 7 09:12:27 2006
• Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=tfSauIosrbxgDr2zxFjgibJncO+QnOpJO2amLnBoVN0PIIwCv1Icsq3L6L86JXUCVC+jL0QQFywo0Lwsmyyq5XZHDdv0leteIa37TuWGpUYpmgj97Gf/F1tC0/GUtM7TkTtwTeF+yIN3UOEWkbYCFi+19FWLM+uI9zIwDbBBHow=
• Sender: AUDITORY Research in Auditory Perception <AUDITORY@xxxxxxxxxxxxxxx>

```Hi,

I had posted some queries on filtering using FFT (in frequency domain).
I could implement it properly.

my working code is doing ...

// doing one filtering

ie,           input -> Filter_1 -> output

while(1)
{
1) Take 512 samples input (no windowing, so, hop size is 512 again).
Zero pad it to make it 1024.
Take 1024 point FFT.
2) Take 128 filter-1 coeff ,
zero pad to make it 1024.
Take 1024 FFT.
3) Multiply the spectrum to get 1024 bins.
4) Take 1024 point IFFT.
5) do a 50 % overlap and add to get back 512 out sample..
}

Now, I want to do the 2nd filtering in a cascaded manner ..
ie,
input -> Filter_1 -> Filter_2 -> output

what is the best way to do this ??

coming back to time domain and going for another filtering using FFT
won't be efficient .. i guess .

conceptually multiplying the Transfer function of these two filters
will give the resultant Transfer function...

whats the practical approach ?

need

regards,
sherin

```