Mimo beam forming on USRP

GNU Radio
1

Hi Guys,

I am planning to build a mimo system which supports beam forming. Is
USRP a
good platform for my target? I do not want to use WARP because I need to
operate at frequencies lower than 2.4GHz.

A potential problem of USRP is phase alignment. As the following doc
suggests, phase alignment can be done by MANUALLY tuning the phase. This
process can be labor intensive when the number of antennas increases. I
am
also worried about the problem that the manual phase calibration might
need
to be done each time you use the mimo system. Is there any approach that
can make this process much easier?

http://www.ettus.com/content/files/kb/mimo_and_sync_with_usrp_updated.pdf

Would be great if you can give me some suggestions about what devices
should I use for building a mimo system for beam forming. Thanks.

Pengyu

2

If you use the phase-resynch feature of the SBX card, and you have
common refclock and 1PPS across all your USRPs in the array,
then you should be OK.

In general, synthesized downconverters use fractional-N synthesizers
where their starting phase after a re-tune is random with respect to any
other synthesizer using the same reference clock. That’s just the way
fractional- N synthesis works. It happens that SBX uses an ADF4351
synthesizer, which has a (rare!) hardware feature to allow phase-resynch
after tuning. In order to make this work, all SBXes in your “array” must
be tuned at exactly the same time, so the timed_commands feature must be
used to do the tuning.

See:

http://files.ettus.com/manual/page_sync.html

On 2015-03-03 11:05, Pengyu Z. wrote:

Pengyu

Discuss-gnuradio mailing list
[email protected]
Discuss-gnuradio Info Page [1]

Links:

3

Can you give us some details (frequency, number of antennas) that you
want to have?

A lot of people have done MIMO with USRPs, up to 100x100 antennas, so
there is some precedent. For beam forming, you’ll need very good phase
alignment, depending on what you’re planning to set up. This is always
difficult, because all sorts of things affect phase. You should, in any
case, have a way to test phase alignment.

Cheers,
Martin

4

Martin - I am thinking about doing beam forming around 900MHz. The
number
of antennas that I want to use is around 8x8. I am thinking about using
USRP N210+SBX+OctoClock-G suggested by ettus. But not sure how much
phase
mis-alignment will I observe.

Marcus - Thanks for the tips. Maybe I should look at how other folks do
phase calibration on USRP. Any recommendation of sample projects?
Ideally,
I hope that this process can be done automatically in software or in
hardware.

Pengyu

5

Indeed, I should have emphasized that despite having features like
“phase resynch”, in any practical phased-array, you’ll still need some
mechanism for calibrating the phase centers of all your antenna.
Differential environmental factors can, for example, affect the phasing
of the antenna feed paths, even with your LOs all agreeing quite-well on
phase.

In Radio Astronomy, they do this all the time–there’s always a
phase-calibration process that is undertaken prior to observations,
because despite careful layout of common LOs, carefully-measured
feedlines from the same batch of cable, etc, differentials creep in.

On 2015-03-03 11:26, Martin B. wrote:

Can you give us some details (frequency, number of antennas) that you want to
have?

A lot of people have done MIMO with USRPs, up to 100x100 antennas, so there is
some precedent. For beam forming, you’ll need very good phase alignment, depending
on what you’re planning to set up. This is always difficult, because all sorts of
things affect phase. You should, in any case, have a way to test phase alignment.

Cheers,
Martin

On 03.03.2015 08:18, [email protected] wrote:
If you use the phase-resynch feature of the SBX card, and you have common
refclock and 1PPS across all your USRPs in the array, then you should be OK. In
general, synthesized downconverters use fractional-N synthesizers where their
starting phase after a re-tune is random with respect to any other synthesizer
using the same reference clock. That’s just the way fractional- N synthesis works.
It happens that SBX uses an ADF4351 synthesizer, which has a (rare!) hardware
feature to allow phase-resynch after tuning. In order to make this work, all SBXes
in your “array” must be tuned at exactly the same time, so the timed_commands
feature must be used to do the tuning. See:
USRP Hardware Driver and USRP Manual: Device Synchronization [1] On 2015-03-03 11:05, Pengyu Z.
wrote: Hi Guys, I am planning to build a mimo system which supports beam forming.
Is USRP a good platform for my target? I do not want to use WARP because I need to
operate at frequencies lower than 2.4GHz. A potential problem of USRP
is phase alignment. As the following doc suggests, phase alignment can
be done by MANUALLY tuning the phase. This process can be labor
intensive when the number of antennas increases. I am also worried about
the problem that the manual phase calibration might need to be done each
time you use the mimo system. Is there any approach that can make this
process much easier?
http://www.ettus.com/content/files/kb/mimo_and_sync_with_usrp_updated.pdf
[2] Would be great if you can give me some suggestions about what
devices should I use for building a mimo system for beam forming.
Thanks. Pengyu _______________________________________________
Discuss-gnuradio mailing list [email protected]
mailto:[email protected]
Discuss-gnuradio Info Page [3]
_______________________________________________ Discuss-gnuradio mailing
list [email protected]
Discuss-gnuradio Info Page [3]

Discuss-gnuradio mailing list
[email protected]
Discuss-gnuradio Info Page [3]

Links:

6

Pengyu:

I think people tend to do it differently, depending on the particular
application. In radio astronomy, they typically point the array at a
well-characterized “calibrator source” and make phase measurements, and
then tweak based on known baseline geometries.

For commercial MIMO, using a distant transmitter of known properties and
arrival angles might be useful, or a local calibrator source.

On 2015-03-03 11:41, Pengyu Z. wrote:

In Radio Astronomy, they do this all the time–there’s always a
phase-calibration process that is undertaken prior to observations, because
despite careful layout of common LOs, carefully-measured feedlines from the same
batch of cable, etc, differentials creep in.
On 03.03.2015 08:18, [email protected] wrote:
If you use the phase-resynch feature of the SBX card, and you have common
refclock and 1PPS across all your USRPs in the array, then you should be OK. In
general, synthesized downconverters use fractional-N synthesizers where their
starting phase after a re-tune is random with respect to any other synthesizer
using the same reference clock. That’s just the way fractional- N synthesis works.
It happens that SBX uses an ADF4351 synthesizer, which has a (rare!) hardware
feature to allow phase-resynch after tuning. In order to make this work, all SBXes
in your “array” must be tuned at exactly the same time, so the timed_commands
feature must be used to do the tuning. See:
USRP Hardware Driver and USRP Manual: Device Synchronization [1] On 2015-03-03 11:05, Pengyu Z.
wrote: Hi Guys, I am planning to build a mimo system which supports beam forming.
Is USRP a good platform for my target? I do not want to use WARP because I need to
operate at frequencies lower than 2.4GHz. A potential problem of USRP
is phase alignment. As the following doc suggests, phase alignment can
be done by MANUALLY tuning the phase. This process can be labor
intensive when the number of antennas increases. I am also worried about
the problem that the manual phase calibration might need to be done each
time you use the mimo system. Is there any approach that can make this
process much easier?
http://www.ettus.com/content/files/kb/mimo_and_sync_with_usrp_updated.pdf
[2] Would be great if you can give me some suggestions about what
devices should I use for building a mimo system for beam forming.
Thanks. Pengyu _______________________________________________
Discuss-gnuradio mailing list [email protected]
mailto:[email protected]
Discuss-gnuradio Info Page [3]
_______________________________________________ Discuss-gnuradio mailing
list [email protected]
Discuss-gnuradio Info Page [3]

Discuss-gnuradio mailing list
[email protected]
Discuss-gnuradio Info Page [3]

Links:

7

Marcus - I am thinking about sending a training signal using one TX
antenna
and figure out the phases of other antennas by looking at the maximum
correlation. Does this sound like a reasonable solution for a practical
mimo system?

8

On 03.03.2015 08:41, Pengyu Z. wrote:

Martin - I am thinking about doing beam forming around 900MHz. The
number of antennas that I want to use is around 8x8. I am thinking about
using USRP N210+SBX+OctoClock-G suggested by ettus. But not sure how
much phase mis-alignment will I observe.

In that case, SBX will work fine. How much phase alignment you’ll
observe depends on your actual setup… there’s a lot of factors (which
is why, as Marcus mentioned, overall systems usually get calibrated
before usage).

If you have specific questions or concerns, I suggest you contact Neel
(CC’d).

Cheers,
Martin

9

When I read this, pCell comes in mind:
http://www.rearden.com/artemis/An-Introduction-to-pCell-White-Paper-150224.p
df

Quite exciting stuff!

Ralph.

to have?

A lot of people have done MIMO with USRPs, up to 100x100 antennas, so
there is some precedent. For beam forming, you’ll need very good phase
alignment, depending on what you’re planning to set up. This is always
difficult, because all sorts of things affect phase. You should, in any
case,

In general, synthesized downconverters use fractional-N synthesizers
where their starting phase after a re-tune is random with respect to
any other synthesizer using the same reference clock. That’s just the
way
fractional- N synthesis works. It happens that SBX uses an ADF4351
synthesizer, which has a (rare!) hardware feature to allow
phase-resynch after tuning. In order to make this work, all SBXes in
your