PDA

View Full Version : Aggregate Beamformer?


Peter J. Kootsookos
10-20-2003, 05:03 AM
Hi All,

I was just flipping through the October issue of JASA and came across
the "Aggregate Beamformer" idea from David I. Havelock.

Has anyone looked at these?

Ciao,

Peter K.

PS: The key idea is that instead of synchronously sampling all
channels of an array, channels of the array are sampled randomly (and
then "time order sorted" into what Havelock calls an "alignment
buffer".

The gain is that the hardware is greatly simplified at little
performance expense.

--
Peter J. Kootsookos

"I will ignore all ideas for new works [..], the invention of which
has reached its limits and for whose improvement I see no further
hope."

- Julius Frontinus, c. AD 84

James K.
10-20-2003, 04:11 PM
"Peter J. Kootsookos" <[email protected]> wrote in message
news:[email protected]...
> Hi All,
>
> I was just flipping through the October issue of JASA and came across
> the "Aggregate Beamformer" idea from David I. Havelock.
>
> Has anyone looked at these?
>
> Ciao,
>
> Peter K.
>
> PS: The key idea is that instead of synchronously sampling all
> channels of an array, channels of the array are sampled randomly (and
> then "time order sorted" into what Havelock calls an "alignment
> buffer".
>
> The gain is that the hardware is greatly simplified at little
> performance expense.
>
> --
> Peter J. Kootsookos
>
> "I will ignore all ideas for new works [..], the invention of which
> has reached its limits and for whose improvement I see no further
> hope."
>
> - Julius Frontinus, c. AD 84

Very interesting topic to me, MIMO signal processing engineer.
I'll review this issue, and I'll attach my comments on this your article.

In addition, although I'm not sure my idea for MIMO beamforming is in-line
with this
article; I attached below for your comments and intresting.

<Abstract of my idea for MIMO beamforming>
In this paper, we propose the effective feedback beamforming scheme for
multiuser scheduling in multiple-input multiple-output (MIMO) wireless
broadcast channels. Beamforming is well known to be a solution to increase
the system throughput, which requires the channel state information (CSI)
fed back to the transmitter from the receiver. Thus, the accuracy of CSI at
the transmitter is also significant for performance improvement. On the
other hand, when the multi-user environment is considered in MIMO systems,
the scheduling algorithms need to be adopted in wireless networks. One of
those schemes is the MIMO scheduling with the antenna selection beamforming
that independently assigns the transmit antennas to the different receivers
instead of utilizing the full weight beamforming. However, in this scheme
the complexity is a disadvantage, because the nonlinear equalizers need to
be exploited at the receiver to approach optimal throughput which can be
achieved by the full weight beamforming. To reduce the complexity of the
above scheme, we propose the alternative method that delivers a unitary
matrix (e.g. the right singular matrix of the channel) to the transmitter by
feedback signalling. Numerical results demonstrate that our proposed scheme
performs close enough to the optimal throughput.
-- End --

More intrested reader may refer below to see in detail:
R1-031142 by ftp of www.3ggp.org (/WG1_RL1/TSGR1_34/Docs/Zips/),
or
news:[email protected] om

--
Best regards,
James K. ([email protected])
- Private opinions: These are not the opinions from my affiliation.

Stan Pawlukiewicz
10-20-2003, 06:29 PM
Peter J. Kootsookos wrote:
> Hi All,
>
> I was just flipping through the October issue of JASA and came across
> the "Aggregate Beamformer" idea from David I. Havelock.
>
> Has anyone looked at these?
>
> Ciao,
>
> Peter K.
>
> PS: The key idea is that instead of synchronously sampling all
> channels of an array, channels of the array are sampled randomly (and
> then "time order sorted" into what Havelock calls an "alignment
> buffer".

"Rebar Signal Processing" ? ;)
>
> The gain is that the hardware is greatly simplified at little
> performance expense.
>

Rune Allnor
10-20-2003, 10:59 PM
[email protected] (Peter J. Kootsookos) wrote in message news:<[email protected]>...
> Hi All,
>
> I was just flipping through the October issue of JASA and came across
> the "Aggregate Beamformer" idea from David I. Havelock.
>
> Has anyone looked at these?
>
> Ciao,
>
> Peter K.
>
> PS: The key idea is that instead of synchronously sampling all
> channels of an array, channels of the array are sampled randomly (and
> then "time order sorted" into what Havelock calls an "alignment
> buffer".
>
> The gain is that the hardware is greatly simplified at little
> performance expense.

I can only get to the abstract, not the paper itself, via JASA Online.
Here is the abstract, just so we know what we're talking about:

================================================== =========================
Sensor array beamforming using random channel sampling:
The aggregate beamformer
David I. Havelock
Institute for Microstructural Sciences, National Research Council,
Ottawa, Ontario K1A 0R6, Canada

(Received 15 February 2003; revised 16 July 2003; accepted 22 July 2003)

The aggregate (AGG) beamformer can provide significant savings in hardware
and software cost and complexity relative to conventional beamformers while
retaining equivalent directional performance. The key to achieving the
savings is collecting data from array channels in a random sequence, rather
than simultaneously or sequentially. This allows the AGG beamformer to
convert unwanted off-beam signals into wideband noise that can then be
reduced by filtering. The total sampling rate is chosen to obtain the
desired residual noise level and signal bandwidth. It is independent of
the number of sensors. Analog anti-alias filters normally are not required,
since the Nyquist frequency is determined by the total sampling rate, not
the (lower) per-channel sampling rate. Beamformer delay quantization and
steering resolution are greatly improved relative to conventional
beamforming without the need for data interpolation. The beamformed signal,
prior to decimation filtering, is obtained without arithmetic operations
on the data. Low front-end hardware-complexity makes the AGG beamformer
suitable for highly integrated systems. The number of sensors in the array
can be altered without re-configuring buffers or altering the sampling rate.
The principles of the AGG beamformer are introduced, and simulation and
experimental results demonstrate performance.
================================================== =========================

Based on the abstract alone, it's just too much. This seems to be too good
to be true: No requirements to temporal presicion, no dependence on the
number of sensors (what about aperture?), no arithmetics on the data...

At the outset I would approach this article with the ambition of finding
out where the flaw is. IF I by trying that, do not find a flaw, IF I find
some obscure sequence of causes and effects that actually gives the desired
result, IF other people attempt the same and succeed, then I will MAYBE
start to believe this guy has a case. But until then... caution, caution
and then some more caution, all of it applied very cautiously...

Rune

Rune Allnor
10-21-2003, 09:53 AM
[email protected] (Rune Allnor) wrote in message news:<[email protected]>...
> [email protected] (Peter J. Kootsookos) wrote in message news:<[email protected]>...
> > Hi All,
> >
> > I was just flipping through the October issue of JASA and came across
> > the "Aggregate Beamformer" idea from David I. Havelock.
....
> At the outset I would approach this article with the ambition of finding
> out where the flaw is.

I've got a copy of the paper and have browsed very quickly through it.
It did take some time until I understood what the idea is, the article
would be much more readable if figures 9 and 10 came directly after
figures 1 and 2.

Anyway, the case in question is that an N sensor array is sampled using
an analog channel switch and one ADC. Thus, the output signal from the ADC
is effectively a time multiplexed mixture of the individual channels.
Havelock's argument is that if this mixture is systematic (fig. 9), one
have to demultiplex all the channels before processing them (which is what
regular beamformers do), but if the time mux scheme is "random", the mux'ed
time sequence itself resembles a random one (fig. 10) and can be processed
in its own right. So Havelock designs a "random" multiplexing scheme that
scrambles the multiplexed sequence, processes this scrambled sequenci in
the computer, which effectively substitutes SW decimation filters for HW
antialias filters, and finds a discrete representation of the input signal.

So far so good.

The processing scheme relies extensively on intimate knowledge of
non-uniform sampling theory and multirate signal processing. I don't think
I have enough knowledge to comment on these things. What I would like to
see, though, is a more detailed description of the sample scrambling scheme.
The C code segment at the end of section II in the paper, mentions an index
buffer S that keeps information about the channel sampling sequence. Is this
sequence truly "random", i.e. updated and randomized between each array
sampling cycle, or is it merely "arbitrary", in that the channel sampling
sequence is non-sequential but fixed once determined? Apparently, the paper
does not answer this question.

The second unclear question is how rigid the requirements are to the timing
of the sampling of channels. The simulation results in the figures indicate
that the "snapshot sampling rate" SSR and "ADC sampling rate" ASR are
related as

ASR = N*SSR

where N is the number of channels. I could imagine a system being specified
from the required SSR and the *maximum* number of channels, such that one
installation with fewer channels actually show an "ADC Duty Cycle" < 100%.
I don't know how important such questions are, but could a short ADC duty
cycle counter the effects of scrambling the mux sequence? Somehow I think
it could.

Last, the examples shown of the beamformer performance only shows one
signal present. So can see no indications about how well the beamformer
suppresses interference from other directions than the steering direction.

In summary, this could be a very interesting approach to beamforming. It
appears that Havelock chooses and approach to the signal processing that
seems to be somewhat related to dithering that has been discussed in another
thread. The main idea is to reduce the cost of the system hardware related
to signal sampling, possibly at the expence of increased demands on
processing management SW and HW.

Rune

Jerry Avins
10-21-2003, 04:14 PM
Rune Allnor quoted with skepticism:

> ... Analog anti-alias filters normally are not required,
> since the Nyquist frequency is determined by the total sampling rate, not
> the (lower) per-channel sampling rate. ...

Telephony people would like to have that too!

Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ

Mindsport
01-16-2005, 01:12 PM
Would ne body mind giving the link of the paper for a critical
evaluation

TomC
01-26-2005, 09:56 AM
Maybe not exactly what you are looking for but US patent 6,747,584 gives
reasonable detail about the technique.


"Mindsport" <[email protected]> wrote in message
news:[email protected] oups.com...
> Would ne body mind giving the link of the paper for a critical
> evaluation
>