Re: is sample pattern matching possible in any way?

On Dec 4, 12:34 pm, cincy...@xxxxxxxxx wrote:
On Dec 3, 8:55 pm, George Nychis <gnyc...@xxxxxxxxx> wrote:



On Dec 3, 8:04 pm, cincy...@xxxxxxxxx wrote:

On Dec 3, 3:58 pm, George Nychis <gnyc...@xxxxxxxxx> wrote:

Hi all,

This is not typical the level of radios I work at, so I'm not entirely
sure of the feasibility of this. I'm working on a MAC development
platform using a software-defined radio platform, GNU Radio and the
USRP.

The problem that I am trying to overcome is the latency between the
USRP and host, and the ability to meet the specifications of common
protocols. For example, most 802.11 timing requirements are on the
order of microseconds, but it takes on the order of milliseconds to
get samples to the host and pass them through the PHY layer (at the
host). Just as an example, the time to determine an incoming
transmission is destined for you based on the frame decoded by the
physical layer is too great to turnaround an ACK within 802.11 specs.

What I'm wondering is if there is any way to do sample pattern
matching to detect incoming frames (start of frame bits) without full
decoding of the PHY layer? This technique must be independent of the
PHY layer.

Once you've decoded a frame, you now have access to the raw samples
that were needed to decode it. Is there anything you can do with
these samples to detect frames without requiring the full PHY layer?
For example, passing samples back down to the USRP/FPGA which can use
them for detecting frames before the samples are sent to the host.

By dumping the raw samples used to decode the start of frame bits from
several transmissions, I am not finding any correlation in the
samples. I didn't entirely expect to find correlation, however it
would have been nice ;)

So, I'm wondering if anyone knows of any techniques that have been
published to do something similar. It would be extremely useful to my
research.

Thank you!
George

I'm not familiar with the 802.11 specifications, so I don't know
exactly what it is you're looking for, but, at a high level, pattern
matching in the form of matched filters or synch word detectors is a
very common tool in communications receivers. Judging by the magnitude
of the turnaround times you're talking about (microseconds), I would
guess that most implementations handle that level of processing in
hardware. I'm sure it would be difficult to implement something like
that in a software radio; you're going to be limited by how often you
sample your data and how fast you can process it.

Many communications protocols have a "start of packet" sequence (synch
word, preamble) that is fixed for all packets, so you can correlate
your input sample stream with a replica of the sequence (a matched
filter) and easily lock onto the packet header. I'm surprised that you
say they don't all look the same, but it is possible that there is
some level of variation introduced to the preamble for some reason.
Maybe someone more familiar with the guts of 802.11 can illuminate
this further.

Jason

Hi Jason,

Right, there is a sync word and preamble in terms of bits. But to get
this I must pass the raw samples through the physical layer. For
example, if I use GMSK I must pass the samples through a channel
filter and a gaussian filter to get the bit stream, and then look for
the sync word in the bitstream. I agree, this is extremely common
receivers, but this is not what I'm looking to try and do. GNU Radio
can in fact do this, I am able to successfully use a physical layer
and framing and detect incoming frames if I pass the samples through
the physical layer.

What I'm trying to find out if it can be done, is if I can some how
detect the frame *without* using the physical layer and decoding the
sample stream in to bits. I don't know if this is entirely possible.
I don't see correlation because I'm not talking about correlating the
decoded bits. If I correlate the decoded bits of the sync word, it
should be a positive correlation of 1 for all frames. What I'm trying
to correlate are the raw samples... I want to know if its possible to
do the frame detection in sample space.

See what I'm getting at? It's not specific to 802.11 at all... I want
a solution that works for any protocol/physical layer.

- George

If you have access to the raw samples, then yes, you can do these sort
of things. In fact, that's how you would employ a matched filter. The
filter could be matched to your pulse shape, to locate individual
bits, or you could shape it like a preamble waveform to locate the
position of an entire sequence. How you employ these techniques would
differ depending on the modulation format and data protocol.

Jason

Jason,

I think you're on to what I'm looking for now. Thank you very much
for the response.

What I am interested in doing is shaping it like a preamble waveform
to locate the position of an entire sequence. When you say this:
How you employ these techniques would
differ depending on the modulation format and data protocol.

Is it possible to implement a highly reconfigurable matched filter to
support any modulation format and data protocol? This would allow it
to been reconfigured (not reimplemented) if a new modulation technique
is used by giving it some information like "this is what the waveform
looks like in terms of the sequence I'm looking for."

Or is the matched filter performing specific modulation techniques to
detect the sequence? I would hope not, else you might as well just
use the full PHY layer.

I'm going to start reading up on matched filters now.

Thanks again!

- George
.

Relevant Pages

  • Re: How to align two pictures.
    ... set layer property to "Difference". ... Handy for those doing hand-held HDR frames, ... needing to have an automated way of aligning video frames. ... rotates, and scales frames automatically. ...
    (rec.photo.digital.slr-systems)
  • Re: Layers Content overlaps
    ... I would redesign: ... >> their browsers in small windows. ... >>> I have created a htm file with frames. ... >>> in one layer appears to overlap on to the next layer. ...
    (microsoft.public.frontpage.programming)
  • Keyboard Shortcuts
    ... What is with the keyboard shortcuts have "Ctrl" in virtually every ... Video Frames to Layers... ... Layer Comps to Files... ... Define Brush Preset... ...
    (alt.graphics.photoshop)
  • Re: Zero motion vector Inter prediction in H264
    ... I want to do an H264 encoder with minimal inter prediction. ... Any other suggestion on improving compression efficiency from Intra- ... Allow "Layer Clumps". ... results than trying to compress the whole series of frames "as is". ...
    (comp.compression)
  • Some mp3 and wma files wont play
    ... MPEG 1.0 layer 3 ... 32kbit, 25779 frames ... CRCs: No ... Emphasis: None ...
    (microsoft.public.windowsmedia.player)