What fixed point format should i chose ?

Hi all,
I am working on an Noise Adaptive Playback project to eliminate the noise from a microphone of the USB HeadSet.
My project is working ok with floating point number, now i am converting it to fixed point number to run
on an ARM7 processor.
Most of calculation in my source code is from -1 to 1, but there are some value very big , the biggest value is
around 23000. I don't know which fixed point format should i use to cover the range of all value in my source code.
Any body can help me on this.

Thanks in advanced !
Tung Start a New Thread

Based on your actual project, you can choose some scale to convert float to fix.For example , when the value is between -1 and 1 ,you choose Q15,but around 23000's value, you use 32bit format with Q15.




2008-08-21 18:05:38=A...@yahoo.com д
Hi all,
I am working on an Noise Adaptive Playback project to eliminate the noise from a microphone of the USB HeadSet.
My project is working ok with floating point number, now i am converting it to fixed point number to run
on an ARM7 processor.
Most of calculation in my source code is from -1 to 1, but there are some value very big , the biggest value is
around 23000. I don't know which fixed point format should i use to cover the range of all value in my source code.
Any body can help me on this.

Thanks in advanced !
Tung Start a New Thread