[time-nuts] decimation versus decimation

[Bob kb8tq]

Hi

If you take a look at how ADEV has been traditionally done for many 
decades, they *do* indeed “decimate without the filter”.  There is no
re-filtering process as you go from 1 to 2 to 10,000 second tau.

Bob

> On Feb 25, 2020, at 10:33 AM, Hal Murray via time-nuts <time-nuts at lists.febo.com> wrote:
> 
> 
>> One meaning is simply: "take every nth sample and discard the others without
>> regard to possible aliasing".
> 
>> The other meaning is: "take every nth sample, but first prefilter as
>> appropriate to (sensibly) eliminate aliasing".
> 
>> Which is it?
> 
> Unless you are doing something tricky, it doesn't make sense to decimate 
> without the filter.
> 
> I'll phrase your two cases differently.  The first is a box that just takes 
> every Nth sample.  Some other box did the filtering.  The second box includes 
> the filter because that is such a common case.  The "box" can be software or 
> hardware.
> 
> ---------
> 
> Sometimes the aliasing is a feature.  If you start with a 100 MHz signal, run 
> it through a band pass filter, now you can decimate and your signal will get 
> aliased down into baseband.  But that's not aliasing bad stuff on top of your 
> signal.  It's aliasing your signal on top of emptyness.
> 
> 
> 
> -- 
> These are my opinions.  I hate spam.
> 
> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.