[amsat-bb] Theseus Cores PFB Channelizer - update!

[Michelle Thompson]

Hee! One of my all time favorites.

A channelizer splits up a section of bandwidth into channels.

Now, there's several ways to do this. Some are simple. Like, one filter
that is switched to a new center frequency. It only lets you tune to one
channel at a time, but it's cheap and reliable. Some are very high
performance, but are expensive in all sorts of ways. There is of course all
sorts of trade-offs.

A filter bank is a set of filters that each pick out a channel from the
spectrum you're interested in. Filter banks range from simple sets of
hardware filters to complex software filters.

Polyphase filter banks are a digital signal processing technique where you
get sharply-defined channels, with flat response, and very little signal
leaking out to mess up the adjacent channels.

This bank of receivers comes to life by calculating a sequence of fast
fourier transforms (FFT) on overlapping segments of the input data stream.

OK so that means you need an input data stream. This assumes you have a
good digital sample of your spectrum. The samples are flying at you and you
are going to take overlapping subsets of this data stream and you are going
to do an FFT on each of those segments. The overlapping part is super
important.

An FFT is a math technique that converts time series into frequency series.
You can invert it and get the original time series back again from the
frequency.

So we have overlapping segments of data in time, and we zap them with FFTs
into segments of frequency. But! We can't just do this however we want,
though. We need the segments to line up so that the resulting channels
automatically cancel out parts of the signal that would otherwise interfere
with each other. Energy is really uncooperative.

It's like trying to direct a parade. You don't want any gaps but you can't
have floats banging into each other.

So once we've got the number of channels and the overlapping and the
weighting figured out, and if we use fast enough circuits, then we have
very fast efficient channelizers that can be implemented in FPGAs and ASICs.

Usually, these types of channelizers are implemented in powers of two (2,
4, 8, 16, 32 etc.). But, you don't have to do it that way. You can have an
arbitrary number of channels, it's just less efficient to not use a power
of two.

Usually, the channels are all the same size. That's the simplest and
easiest way to do it. But, there is at least one method to set up a
channelizer with different channel sizes in the same bank! Some can be
large and some small.

Since polyphase filter banks are digital and can be built using
reconfigurable devices like FPGAs, they can be changed in the field. Say
you have more users than expected, or need the channels to be larger. Or
the FCC says to move over. Or, you have a system that can sense incumbent
users of the spectrum and adjust or adapt to those users, no matter where
in the world your radio is turned on.

For any system that receives channels separated into different frequencies,
and is converting all those channels into something like digital frames or
digital data, a polyphase filter bank channelizer is almost always the best
choice.

-Michelle W5NYV




On Tue, Aug 6, 2019 at 9:29 PM Greg D <ko6th.greg at gmail.com> wrote:

> Great work, Michelle!
>
> But tell me, is a "Theseus Cores RFNoC polyphase filter bank
> channelizer" anything like an "Illudium Q-36 Explosive Space Modulator"?
>
> What does this thing do, for those who don't speak Martian?
>
> Greg  KO6TH
>
>
>
> Michelle Thompson via AMSAT-BB wrote:
> > Good news. The Theseus Cores RFNoC polyphase filter bank channelizer is
> > working.
> >
> > The channel 0 timeouts, which were a vexing problem, seem to be gone
> thanks
> > to hard work from EJ Kreinar and Phillip Vallance.
> >
> > This particular part is vital for FDMA to TDM type payloads and
> groundsats.
> > Having a fast, efficient, open source implementation that works in GNU
> > Radio is a big step forward.
> >
> > We are taking the entire setup with us to DEFCON. We are using a linux
> > computer and a USRP. Experiments so far have been on the VHF ham bands
> and
> > FM broadcast. Next step, move it up to microwave and demonstrate Phase 4
> > Ground channelization.
> >
> > If you're at DEFCON, come by and see us in ham radio village! There's a
> > license exam session, demonstrations, presentations, and lots of
> > socializing.
> >
> > More soon!
> > -Michelle W5NYV
> > _______________________________________________
> > Sent via AMSAT-BB at amsat.org. AMSAT-NA makes this open forum available
> > to all interested persons worldwide without requiring membership.
> Opinions expressed
> > are solely those of the author, and do not reflect the official views of
> AMSAT-NA.
> > Not an AMSAT-NA member? Join now to support the amateur satellite
> program!
> > Subscription settings: https://www.amsat.org/mailman/listinfo/amsat-bb
>
>