Hi experts,
I am currently implementing various filter algorithms for AEC for my FYP onto
EZKIT BF533. However, I need to increase my tapsize to 1024 (currently,
it's at 64). my prof said that it might be due to the processing time being
shorter than the sampling period. Can anyone guide me along? I am quite bad at
processing time. Thanks
Increasing the tap size of FIR filters
Started by ●August 13, 2009
Reply by ●August 13, 20092009-08-13
On Thu, 13 Aug 2009, k...@gmail.com wrote:
> Hi experts,
>
> I am currently implementing various filter algorithms for AEC for my
>FYP onto EZKIT BF533. However, I need to increase my tapsize to 1024
>(currently, it's at 64). my prof said that it might be due to the
>processing time being shorter than the sampling period. Can anyone guide
>me along? I am quite bad at processing time. Thanks
That's a nice problem to have! All you need to do is run on interrupts -
every time you get a new sample reprocess the whole buffer and spit out a
new value. Use circular buffers so you don't have to worry about where
you are in the circle. The processor will spend most of its time
sleeping, but who cares.
It might be good to test this by using one I/O pin to flag when you start
and end your process. Put a scope on that and on you sampling clock. If
your measurment says you don't have enough time between samples to do the
whole buffer, then there is some other problem.
Patience, persistence, truth,
Dr. mike
> Hi experts,
>
> I am currently implementing various filter algorithms for AEC for my
>FYP onto EZKIT BF533. However, I need to increase my tapsize to 1024
>(currently, it's at 64). my prof said that it might be due to the
>processing time being shorter than the sampling period. Can anyone guide
>me along? I am quite bad at processing time. Thanks
That's a nice problem to have! All you need to do is run on interrupts -
every time you get a new sample reprocess the whole buffer and spit out a
new value. Use circular buffers so you don't have to worry about where
you are in the circle. The processor will spend most of its time
sleeping, but who cares.
It might be good to test this by using one I/O pin to flag when you start
and end your process. Put a scope on that and on you sampling clock. If
your measurment says you don't have enough time between samples to do the
whole buffer, then there is some other problem.
Patience, persistence, truth,
Dr. mike
