spurious output on Bandpass filter

Hi all,

I am trying to make a series of bandpass filters each 100 Hz wide,

from 200Hz to 5000 hz. I am currently using butterworth 6th order.

I require sharp filters. ( suppressing better than 1% outofband)

My trouble is that any - all filters give spurious outputs for a short 

time for any abrupt signal.

for ex. a 500Hz filter gives output for a 200Hz signal (for a short time)
then settles down to zero. Otherwise for a 500hz signal it is real fine.

Is there sny way to stop this ? any change in type of filter ?

Please suggest,

Regards,

Devidas

"devidasbhonde" <devidasbhonde@hotmail.com> writes:

> Hi all,
>
> I am trying to make a series of bandpass filters each 100 Hz wide,
>
> from 200Hz to 5000 hz. I am currently using butterworth 6th order.
>
> I require sharp filters. ( suppressing better than 1% outofband)
>
> My trouble is that any - all filters give spurious outputs for a short 
>
> time for any abrupt signal.

The transients contain a wide range of frequencies.

> Is there sny way to stop this ? any change in type of filter ?

There is nothing wrong with the filters.  There is maybe no
easy way.  You could try to detect the transients and based on
that just ignore some of the outputs.  It probably depends on
the purpose of what you are doing.

gr.
Anton

On 2006-05-24, devidasbhonde <devidasbhonde@hotmail.com> wrote:
>
> for ex. a 500Hz filter gives output for a 200Hz signal (for a short time)
> then settles down to zero. Otherwise for a 500hz signal it is real fine.

Heh, over the weekend I was playing with an AD5300 (DAC) someone gave
me, and implementing a very low-speed DDS in a microcontroller to
experiment with.  Of course it quickly became necessary to add an output
filter, so I cobbled together a 2nd order Sallen-Key filter out of
readily available junkbox parts.  Later I decided to try some PSK-31
on the carrier, so I thew in a 31.25Hz phase modulation as a first
pass.  Imagine my surprise when there were big amplitude spikes in the
output at the phase changes!  I finally realized that my fudging on
the component values for the lp filter had moved the 3db point over
enough to attenuate my carrier, and the phase shifts injected a low
frequency component that the filter did not attenuate as much...
So it looks like there's a lesson here for all of us. :)

-- 
Ben Jackson
<ben@ben.com>
http://www.ben.com/

Ben Jackson wrote:

  ...

> So it looks like there's a lesson here for all of us. :)

Which is?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Not really. If it were an FIR you could consider establishing initial
conditions.

You're running up against a fundamental property of reality - a
narrowband filter requires a long time to respond but has great
frequency response. A wideband filter has a quick time
response but lousy frequency response. 

--Randy

Thank you all for the replies,

So it seems that IT is a basic property. And choosing some filter might
help.
Does the FIR filter in this respect ? That is, it has less spurious
output than IIR ? May be it has rather more delay.

Thanks,

Devidas

On 2006-05-25, Jerry Avins <jya@ieee.org> wrote:
> Ben Jackson wrote:
>> So it looks like there's a lesson here for all of us. :)
>
> Which is?

It's like Hamlet said, "There are more frequency components in your signal,
Horatio, than are dreamt of in your philosophy."

-- 
Ben Jackson
<ben@ben.com>
http://www.ben.com/