USRP2 + WBX How to use a 32 MHz signal band?

What would be the best way to work with a 32 MHz band resulting from
USRP2?
I guess, since the minimum decimation rate is 4 (resulting in a 25 MHz
signal band), some kind of interpolation will be required or maybe
resampling?!
Any ideas on how to proceed?
Thanks in advance!

Vlad.

View this message in context:
http://old.nabble.com/USRP2-%2B-WBX-How-to-use-a-32-MHz-signal-band--tp30335430p30335430.html
Sent from the GnuRadio mailing list archive at Nabble.com.

What would be the best way to work with a 32 MHz band resulting from USRP2?
I guess, since the minimum decimation rate is 4 (resulting in a 25 MHz
signal band), some kind of interpolation will be required or maybe
resampling?!
Any ideas on how to proceed?
Thanks in advance!

Vlad.

The maximum bandwidth over the GiGE interface is 25Msps. If you want to
do processing at higher bandwidths, you’ll
have to dive into the UHD code and move to smaller samples (8-bits x
2 instead of 16-bits x 2), or do your processing
on the FPGA. Either approach will require modifying significant
swaths of code on the FPGA.

Also, doing anything significant to a datastream arriving at 25Msps on
the host will require a CPU that’s up to the task.
My IRA radio astronomy code, which does spectral and radiometry
measurements can just-barely keep up at 12.5Msps
on a 4-core Phenom II running flat-out at 3.2GHz.


Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium

On 11/29/2010 08:57 PM, Abdalaleem Andy James Potter wrote:

You could use 2 USRPs and chop it down the middle?

Reconstruct on the computer?

Yup, although see my previous comments about host-side resources
required for such wide bandwidths.


Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium

On Mon, Nov 29, 2010 at 5:59 PM, Vladutzzz [email protected]
wrote:

What would be the best way to work with a 32 MHz band resulting from USRP2?
I guess, since the minimum decimation rate is 4 (resulting in a 25 MHz
signal band), some kind of interpolation will be required or maybe
resampling?!
Any ideas on how to proceed?
Thanks in advance!

Vlad.

I’m not entirely sure what you meant by interpolating or resampling,
but in case you were talking about what I think you were talking
about, that’s not going to work. Sure, you can come in at 25 Msps and
resample to 32 Msps, but you won’t be getting any more information out
of the signal by doing that. You’ll still only have the information in
the original 25 MHz bandwidth, just now sampled at a higher rate
(which doesn’t make any difference or, frankly, any sense).

Tom

On 11/30/2010 07:04 PM, Vladutzzz wrote:

How would the USRPs be connected to the same computer? (MIMO cable?)
How will the two 16 MHz bands be attached together to form the 32 MHz one
(some kind of 2:1 multiplexing - I’m just guessing here)?
I am aware of the great load exerted upon the system resources, but I’m
trying to make this work anyway.
Thanks.

Vlad.
Assuming that you have a big-arsed machine to do this with, here goes:

Assumptions

o phase-coherence between the two isn’t an issue (if you’re just
doing power measurements, it won’t be)
o you have a very-studly computer
o you have two good 1GiGe interfaces

Start out with two single-usrp sources, address them as appropriate for
your two USRP2s.

16MHz isn’t an available bandwidth out of the USRP2, so use 16.66667MHz,
and band-limit it to exactly 16MHz with an FIR bandpass filter.
Your two USRP2s will each be tuned to a frequency that is 16MHz away
from each other.

Once you have your two band-limited complex signals, detect them, using
a complex-to-mag-squared on each of them.
Then put the two signals into an adder, and low-pass filter with a
single-pole IIR or FIR low-pass filter. Decimate to taste
after filtering.


Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium

Tom,
You are right, it doesn’t make any sense, I was thinking about the
reverse method, taking for example just 10 MHz out of a total band of
12.5 by filtering or resampling, but that is totally different since you
have extra band and you just eliminate the extra part, but here you
can’t just make up the difference between 25 and 32 MHz.

Vlad.

— On Tue, 11/30/10, Tom R. [email protected] wrote:

From: Tom R. [email protected]
Subject: Re: [Discuss-gnuradio] USRP2 + WBX How to use a 32 MHz signal
band?
To: “Vladutzzz” [email protected]
Cc: [email protected]
Date: Tuesday, November 30, 2010, 4:48 AM

On Mon, Nov 29, 2010 at 5:59 PM, Vladutzzz [email protected]
wrote:

What would be the best way to work with a 32 MHz band resulting from USRP2?
I guess, since the minimum decimation rate is 4 (resulting in a 25 MHz
signal band), some kind of interpolation will be required or maybe
resampling?!
Any ideas on how to proceed?
Thanks in advance!

Vlad.

I’m not entirely sure what you meant by interpolating or resampling,
but in case you were talking about what I think you were talking
about, that’s not going to work. Sure, you can come in at 25 Msps and
resample to 32 Msps, but you won’t be getting any more information out
of the signal by doing that. You’ll still only have the information in
the original 25 MHz bandwidth, just now sampled at a higher rate
(which doesn’t make any difference or, frankly, any sense).

Tom

Thank you Marcus, for the very prompt and exhaustive reply!
Mostly it’s what I expected, but I am a wee bit worried about the IP
addressing process:
I need to change the IP of one of the USRP2 modules, how do I do that,
considering I use an UDP FPGA image and not UHD?
Also, what IPs should I use in the USPR2 source blocks, now that there’s
2
of them?
Thanks!

Vlad.

Marcus D. Leech wrote:

as
Thanks.

Then put the two signals into an adder, and low-pass filter with a


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


View this message in context:
http://old.nabble.com/USRP2-%2B-WBX-How-to-use-a-32-MHz-signal-band--tp30335430p30347806.html
Sent from the GnuRadio mailing list archive at Nabble.com.

I am interested in this 2 USRPs approach since I don’t have the
experience
and the knowledge to start messing about inside the FPGA firmware code
and
have an extra USRP2.
How would this go?
I tried looking up some info about this topic. I’ve read some bits and
pieces on a few forums. Most of the info is about having one USRP2
module as
a transmitter and the other as a receiver.
I want them both to be receivers(actually half a receiver) and to
complete
each other by receiving half of the 32 MHz band (so each would receive
16
MHz).
How would the USRPs be connected to the same computer? (MIMO cable?)
How will the two 16 MHz bands be attached together to form the 32 MHz
one
(some kind of 2:1 multiplexing - I’m just guessing here)?
I am aware of the great load exerted upon the system resources, but I’m
trying to make this work anyway.
Thanks.

Vlad.

Marcus D. Leech wrote:

[email protected]
Discuss-gnuradio Info Page


View this message in context:
http://old.nabble.com/USRP2-%2B-WBX-How-to-use-a-32-MHz-signal-band--tp30335430p30338714.html
Sent from the GnuRadio mailing list archive at Nabble.com.

On Tue, 2010-11-30 at 08:27 -0800, Vladutzzz wrote:

MHz).
How would the USRPs be connected to the same computer?
How will the two 16 MHz bands be attached together to form the 32 MHz one
(some kind of 2:1 multiplexing - I’m just guessing here)?
I am aware of the great load exerted upon the system resources, but I’m
trying to make this work anyway.
Thanks.

Vlad.

My suggestion would be to connect the two USRP2s as a MIMO pair, but
tune them with 16MHz offset relative to each other. Then you would need
to design a way (a signal processing method) to merge the two streams
together into a single wide-band stream.

BR/
Per

On 12/01/2010 05:49 AM, Vladutzzz wrote:

My advice would be to screw up your courage, bite the bullet, and
convert to UHD.

Theres a UHD utility, usrp2_addr_burner that can be used to change the
IP address.

The factory address is: 192.168.10.2, you’ll need to change it to an
address on a different
subnet, like 192.168.20.2 or some such. The idea is that you have
two GiGE ethernet ports,
and you configure your system with the appropriate subnet on each
ethernet port, and program your
USRP2 addresses accordingly.


Principal Investigator
Shirleys Bay Radio Astronomy Consortium

The problem is that I am in the middle of a project, time is of the
essence
and I don’t have time to start stumbling around with UHD, right now I
have
to use Simulink, hence UDP. GnuRadio Companion still doesn’t work with
UHD,
right (I am very new to python)?
So no chance of UDP working with 2 USRPs at once, huh?!
Let me ask you something else then, if I manage to synchronize 2 USRPs
working on separate computers to start sensing a band of 16MHz each, and
their central frequency being 16MHz apart, saving those two bands in a
file
, what block (system of blocks) can I use to compose a 32MHz band (and I
mean the whole band not just their power measurements) from the two
16MHz
halfbands originating from the “From File” blocks?
Naturally the samplerate of the new signal should be the double of that
of
the halfbands.
If I just use a 2:1 multiplex the halfbands appear superimposed and the
resulting signal band still occupies 16MHz.
Any ideas?
Your help is really appreciated,

Vlad.

Marcus D. Leech wrote:

Thanks!
address on a different
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org


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


View this message in context:
http://old.nabble.com/USRP2-%2B-WBX-How-to-use-a-32-MHz-signal-band--tp30335430p30353714.html
Sent from the GnuRadio mailing list archive at Nabble.com.

On Wed, 2010-12-01 at 13:53 -0800, Vladutzzz wrote:

The problem is that I am in the middle of a project, time is of the essence
and I don’t have time to start stumbling around with UHD, right now I have
to use Simulink, hence UDP. GnuRadio Companion still doesn’t work with UHD,
right (I am very new to python)?

GRC has always worked with UHD. You might try downloading the latest UHD
and giving it a shot.

So no chance of UDP working with 2 USRPs at once, huh?!

No.

Let me ask you something else then, if I manage to synchronize 2 USRPs
working on separate computers to start sensing a band of 16MHz each, and
their central frequency being 16MHz apart, saving those two bands in a file
, what block (system of blocks) can I use to compose a 32MHz band (and I
mean the whole band not just their power measurements) from the two 16MHz
halfbands originating from the “From File” blocks?
Naturally the samplerate of the new signal should be the double of that of
the halfbands.
If I just use a 2:1 multiplex the halfbands appear superimposed and the
resulting signal band still occupies 16MHz.

Assuming your signals are mixed to baseband you can interpolate both to
32Msps and mix one sideband up and the other down so they sit where they
are supposed to, and then add the signals together. There’s probably a
smarter (faster) way to do it.

–n

Per,
At the moment we don’t have a MIMO cable, so we’re looking for
alternative
solutions until one makes its way to our lab :slight_smile:
Thank you for your suggestion though, it is much appreciated.
All the best,

Vlad.

Per Z.-2 wrote:

on a
trying to make this work anyway.

BR/
Per


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


View this message in context:
http://old.nabble.com/USRP2-%2B-WBX-How-to-use-a-32-MHz-signal-band--tp30335430p30353755.html
Sent from the GnuRadio mailing list archive at Nabble.com.

You should be able to setup an interpolating FIR filter to interpolate
by two and to shift your signal up or down by fs. Run each channel
though a filter (one up, one down) and add the resulting streams. The
resulting stream will have 2fs of bandwidth.

-Josh

On Wed, Dec 1, 2010 at 5:02 PM, Nick F. [email protected] wrote:

If I just use a 2:1 multiplex the halfbands appear superimposed and the
resulting signal band still occupies 16MHz.

Assuming your signals are mixed to baseband you can interpolate both to
32Msps and mix one sideband up and the other down so they sit where they
are supposed to, and then add the signals together. There’s probably a
smarter (faster) way to do it.

–n

There’s the pfb_synthesis_filterbank that could do this, too, but it’s
probably not faster than your way if you are just doing two
“channels.” This code also needs a bit of a SIMD burst for the
internal filters, and there’s another tweak that I will be making so
that you can more easily upsample and smash the channels together for
perfect reconstruction (that is, if you’re using a real Nyquist
filter).

But yeah, for right now, I’d do it the way Nick suggested.

Tom

NIck,
What do you mean by “mix one sideband up and the other down”?
How would that go?
Thanks

Vlad.

Nick F.-4 wrote:

GRC has always worked with UHD. You might try downloading the latest UHD

, what block (system of blocks) can I use to compose a 32MHz band (and I
are supposed to, and then add the signals together. There’s probably a

2


View this message in context:
http://old.nabble.com/USRP2-%2B-WBX-How-to-use-a-32-MHz-signal-band--tp30335430p30362146.html
Sent from the GnuRadio mailing list archive at Nabble.com.