Overlapping channels in polyphase channelizer

Is it possible to have overlapping channels in a polyphase channelizer?
say
you have 10MHz of bandwidth, and you want to break it up into 10
channels,
each 2MHz wide with the center frequencies being 1MHz apart.

would you just set the taps to have the right bw and lowpass filter
width
(in this case it would be 1MHz to create a 2MHz channel), then set the
number of outputs on the channelizer block in grc to be 10?

thanks,
-b

On Jun 18, 2015, at 6:14 PM, ben Gee
<[email protected]mailto:[email protected]> wrote:

Is it possible to have overlapping channels in a polyphase channelizer?
say you have 10MHz of bandwidth, and you want to break it up into 10
channels, each 2MHz wide with the center frequencies being 1MHz apart.
Hi Ben, perhaps this paper would be helpful.
http://static1.1.sqspcdn.com/static/f/679473/24573007/1395369129837/rondeau_gr_filtering.pdf?token=96KRK5gwTcSMhOfiIaaoFKDjZJk%3Dhttp://static1.1.sqspcdn.com/static/f/679473/24573007/1395369129837/rondeau_gr_filtering.pdf?token=96KRK5gwTcSMhOfiIaaoFKDjZJk=
It describes a way to rig the filter in order to split a signal into
channels using the polyphase channelizer so that no information is lost
at the channel boundaries. You could use this technique to split your
signal into 1 MHz segments and then join the 1 MHz segments together
into the overlapping 2 MHz channels. Another thing that might work is to
use two channelizer blocks, each generating five channels.

Hi Ben,

yes, that should work; as you’ve guessed, the polyphase channelizer
doesn’t really “care” what your filter looks like. That includes the
band edges, too, so your last channel will overlap with your first.

Best regards,
Marcus

On Fri, Jun 19, 2015 at 3:07 AM, Marcus Müller
[email protected]
wrote:

On 06/19/2015 03:14 AM, ben Gee wrote:
thanks,
-b

Also, look for the paper by harris and Rice, “Digital Receivers and
Transmitters Using Polyphase Filter Banks for Wireless Communications.”
There some good stuff in there about adjusting the bandwidth versus the
sampling rate.

Tom