FPGA Central

Hi All,
Does anyone have a suggestion for a fast down sampling filter?

km wrote:

> Hi All,
> Does anyone have a suggestion for a fast down sampling filter?

Do it yourself or hire a consultant?

VLV

On Dec 9, 8:54*am, km <kylei...@gmail.com> wrote:
> Hi All,
> * Does anyone have a suggestion for a fast down sampling filter?

Please be more specific. What is sample rate of your input data? What
is your required downsampling rate? Is this for a hardware or software
implementation?

Darol Klawetter

On 12/9/2009 7:54 AM, km wrote:
> Hi All,
> Does anyone have a suggestion for a fast down sampling filter?

For N:1 reduction, add N consecutive samples together and output that.
Move to the next window of N samples.

The response will not be great, but it is, in fact, a fast down sampling
filter.

Case closed.

--
Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

Eric Jacobsen wrote:
> On 12/9/2009 7:54 AM, km wrote:
>
> > Hi All,
> > * *Does anyone have a suggestion for a fast down sampling filter?
>
> For N:1 reduction, add N consecutive samples together and output that.
> * Move to the next window of N samples.
>
> The response will not be great, but it is, in fact, a fast down sampling
> filter.
>
> Case closed.

You think you are joking, but read this:

http://www.stereophile.com/features/104law/index1.html

BTW, an even faster downsampling algorithm is to simply sub-sample
(take every N-th sample).

Regards,
Andor

Eric Jacobsen wrote:
> On 12/9/2009 7:54 AM, km wrote:
>> Hi All,
>> Does anyone have a suggestion for a fast down sampling filter?
>
> For N:1 reduction, add N consecutive samples together and output that.
> Move to the next window of N samples.
>
> The response will not be great, but it is, in fact, a fast down sampling
> filter.
>
> Case closed.
>

Eric's suggestion is a very good one in terms of simplicity.
Conceptually, you can think of it as a moving average, followed by
Rune's suggestion of throwing out all but every Nth sample, aka,
downsampling.

However, if you want something fancier, you can use a decimation
algorithm. Decimation combines The ScopeFIR distribution file, which
contains an optimized and free/open decimation function in C, is
available at http://www.iowegian.com/download .

Also, decimation is a special case of the more general method of
resampling, which combines decimation with interpolation. Free
resampling software is online at
https://ccrma.stanford.edu/~jos/resa..._Software.html .

See dspGuru's Multirate FAQ at http://dspguru.com/dsp/faqs/multirate for
some introductory material on multirate topics like downsampling,
decimation, and resampling. <RTFFAQ>

Grant
--
__________________________________________________ ___________________

Grant R. Griffin
Publisher of dspGuru http://www.dspguru.com
Iowegian International Corporation http://www.iowegian.com
See http://www.iowegian.com/img/contact.gif for e-mail address

On 12/10/2009 2:51 AM, Andor wrote:
> Eric Jacobsen wrote:
>> On 12/9/2009 7:54 AM, km wrote:
>>
>>> Hi All,
>>> Does anyone have a suggestion for a fast down sampling filter?
>> For N:1 reduction, add N consecutive samples together and output that.
>> Move to the next window of N samples.
>>
>> The response will not be great, but it is, in fact, a fast down sampling
>> filter.
>>
>> Case closed.
>
> You think you are joking, but read this:
>
> http://www.stereophile.com/features/104law/index1.html
>
> BTW, an even faster downsampling algorithm is to simply sub-sample
> (take every N-th sample).
>
> Regards,
> Andor

I wasn't joking, it's just the simplest thing I could think of that
fully meets the OP's specifications. It's unlikely to be all that
useful in many, if not most, applications, but it meets the spec as stated.

Dropping all but every Nth sample doesn't "filter", which was part of
the OPs problem statement, so I didn't suggest that.

--
Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

km <[email protected]> wrote:

> Does anyone have a suggestion for a fast down sampling filter?

Good question. I have been wondering for a while now how HDTV
receivers do the down conversion. At the frame rate they run at
and the available processing I am sure they don't do the full
add zero sample/filter/select sample method, especially not in
two dimensions.

Maybe a simple low pass filter and linear interpolation.
As simple as it can be and give reasonable results.

-- glen