# USRP2 LFTX maximum real baseband bandwidth?

Hi all,

I would like to send a real signal with a bandwith from DC to 25MHZ
thanks to the USRP2 and the LFTX daughterboard.

So, I generate a Real baseband signal and I use only the “I” channel.
But, it seems that the minimun interpolation of USRP2 is 4 (USRP2faq)
that’s why I use interp=4. So, I have a bandwidth of 12.5 MHz because
the sampling frequency is fs=25 MS/s ==> fs/2 = 12.5 MS/s. For me, it
seems good.

So, my first question is:
Is it possible to use an interpolation of only 2 to have bandwidth=25
Mhz?

Secondly, I’m not sure but if we use interp=2 we have before the DUC a
data rate = 50 MS/s * 32 (float32) = 1600 Mbits/s ==> > Gb/s.
However, if we use interp=2 and “short” format, we have a data rate = 50
MS/s * 16 (short) = 800 Mbits/s ==> < Gb/s.

Is it correct?

Thanks,

Philippe.

Hi,
I believe the short format isn’t implemented for the USRP2 yet, or has
this been done recently?

/Ulrika

Ulrika U. wrote:

Hi,
I believe the short format isn’t implemented for the USRP2 yet, or has this been done recently?

/Ulrika

Hi,

Indeed, the short format seems to be not implemented for the USRP2 yet.
However, usrp2_sink_16sc can be used (I tried to send a cosine with
short format but with no success) but do not work.

If we can used the short format and an interpolation = 2, do you think
that the data rate will be 50 MS/s * 16 (short) = 800 Mbits/s before the
Digital Up Converter?

and

Is it possible to have an interpolation = 2 ?

Philippe.

Quoting TANGUY Philippe [email protected]:

Indeed, the short format seems to be not implemented for the USRP2

My understanding is that we need 32bits per sample, 16bits I and 16bits
Q.

BR/
Per

Per Z. wrote:

Hi,

Is it possible to have an interpolation = 2 ?

My understanding is that we need 32bits per sample, 16bits I and 16bits Q.

BR/
Per

Thanks for the response. Indeed, you are right.

In fact, I was confused between the GNUradio format representation and
the appropriate “on the wire” representation, depending on the current
USRP2 configuration which is apparently 16-bit I & Q. (see in
documentation here:

Now, my problem is to know if it is possible to use an interpolation of
2 in order to have a data rate of 50 MS/s to transmit a real baseband
signal with a bandwidth of 25 MHz.

Regards,

Philippe.

On 02/25/2010 01:01 AM, TANGUY Philippe wrote:

Hi all,

I would like to send a real signal with a bandwith from DC to 25MHZ
thanks to the USRP2 and the LFTX daughterboard.

So, I generate a Real baseband signal and I use only the “I” channel.
But, it seems that the minimun interpolation of USRP2 is 4 (USRP2faq)
that’s why I use interp=4. So, I have a bandwidth of 12.5 MHz because
the sampling frequency is fs=25 MS/s ==> fs/2 = 12.5 MS/s. For me, it
seems good.

If you want a 25 MHz wide signal (0 to 25 MHz), then you should generate
it at baseband (+/-12.5 MHz), and have the digital upconverter in the
FPGA upconvert to a center frequency of 12.5 MHz That will give you 0
to 25 MHz coverage.

So, my first question is:
Is it possible to use an interpolation of only 2 to have bandwidth=25 Mhz?

Using an interpolation of 4 gives you a bandwidth of 25 MHz. Using an
interpolation of 2 would give you 50 MHz, but would require 8 bit I and
Q samples.

Secondly, I’m not sure but if we use interp=2 we have before the DUC a
data rate = 50 MS/s * 32 (float32) = 1600 Mbits/s ==> > Gb/s.
However, if we use interp=2 and “short” format, we have a data rate = 50
MS/s * 16 (short) = 800 Mbits/s ==> < Gb/s.

We don’t have an interp by 2 mode yet, but it could be done. However,
as I stated above, I don’t think you need it to do what you want to do.

Matt