LFRX and the NI driver

[cochenob]

Hi!

I'm using an LFRX board with my N210 radio.  A few questions as to the operation...

1. I assumed that since the LF boards don't have a demodulator...that the "Center Frequency" input would be ignored.  However...I notice that if I change it to something other than 0 Hz....my input signal shifts around in frequency.  What's going on here?  Is this the CORDIC doing something even though there's not analog downconversion stage? 

2.  Similarly....I assumed that the "Sample Rate" input (R), which essentially defines the BW...would define the BW between 0-R Hz.  However, if I input a single tone, the spectrum I calculate goes from -R/2 -> +R/2...and my tone shows up at +f and -f.  This was at odds with my expectations.

I had initially assumed that using the LF boards essentially just turned the USRP into a digitizer for capturing real valued signals. 

However, the above observations seem to suggest that really the driver assumes that you've done a downconversion somewhere else and are just using the LF board to capture the baseband quadrature signals.  In this case, the interpretation of "center frequency" and "sample rate" (i.e. - BW) are essentially the same as if I was using the WBX.  The Center Frequency = 0 Hz because I've done the demodulation somewhere else, and the Sample Rate/BW are defined with the assumption that I'm capturing those baseband signals in quadrature.

If I want to just use each A/D channel to capture separate, un-related, real-valued signals....then the sample rate would have the interpretation of the 2*Nyquist frequency for each unrelated real-valued signal....rather than 2*Nyquist frequency of two baseband signals in quadrature, where the sample rate essentially defines the BW of a bandpass signal that was previously downconverted from some carrier.  Right?

Anyone give me a sanity check here? 

---

Brandon

[Evan_S]

Hi Brandon,

I'm currently looking into your questions on the exact interaction of the LFRX daughterboard and the N210 USRP. For a point of reference, are you using any of the NI USRP examples?

For your first question you mention that 

---

@cochenob wrote:

I notice that if I change it to something other than 0 Hz....my input signal shifts around in frequency. 

---

I wanted to confirm exactly how the input signal frquency is shifting, possibly if you could attached screenshots, so that I coan confirm whether that is truly with your current configuration.

[cochenob]

Hi Evan-

I think I eventually answered my own question.  I think I just have to remember that the outputs and configuration of the LFRX isn't really all that different from the WBX....only there isn't any analog demodulation on the front end.  The shifting I was seeing was likely the CORDIC...which was essentially the same behavior if I set different LO frequencies and Center Frequencies with the WBX.

One follow-up question though.....I was hoping to use the LFRX and N210 to digitize some low frequency signals DC coupled signals.  The signals aren't in quadrature...rather, I was just trying to use it as a 'generic' digitizer.  (Yes...there are better hardware options for this...but there were some compelling arguements for using it in our system that contained other USRPs).  Anyway...I noted that the LFRX doesn't really go all the way down to DC.  Things start to roll off in the 10-100 Hz range.  Where does this bandlimiting occur?  Is it on the RF board?  Or are the ADC's AC coupled?  Or am I seeing an artifact of the DC offset removal in the FPGA?

I did notice this occasionally at some frequencies with the WBX when my input tone and USRP center frequency were very very close.  Instead of getting DC (or near DC) I and Q signals, I'd get nothing (the DC component was completely filtered out).  It was as if the ADCs were either AC coupled...or the FPGA was removing the DC component.  Never quite understood what was happening here.

Thanks for the insight!

---

Brandon

[Sarah_Y]

Hey Brandon,

The LFRX board indeed, has no analog downconversion.  Instead, the DDC on the FPGA (based on a CORDIC) does a conversion of the real input from the LFRX into a complex-baseband signal at whatever frequency you request.  The "center frequency" is still meaningful, and the UHD code knows that there is no analog downconverer, so it instructs the DDC to effect the downconversion entirely digitally.

The output of the USRP is always a complex-baseband signal, even if it originated as a "real" signal.   So, if you request a center frequency of 25Mhz on the LFRX, with a sample rate of 5Msps, the DDC will downconvert the incoming 25Mhz signal down to baseband (it will appear as 0Hz, just like any other card), and the complex-baseband signal will then represent -2.5Mhz to +2.5Mhz  (22.5Mhz to 27.5Mhz).

[cochenob]

Hey Sarah!

Thanks.  This is basically the conclusion I came to.  Good to hear someone who actually knows what they're talking about confirm it though.  Thanks!  My impression though is that if I wanted to use the LFRX to digitize baseband signals that *were not* in quadrature...I'd have to set the center frequency to 0 Hz (so no frequency translation via CORDIC), and then split out the "I" and "Q" components in LabView (since they wouldn't really represent a complex pair to begin with). Agreed?

Any insight into DC (or near DC) signals going INTO the LFRX or N210 as per my second post?  i.e. - not signals that get downconverted to DC by the CORDIC?

---

BC

[Sarah_Y]

Hey Brandon,

When I am using the Basic or LF RX board, I do the same thing you mentioned.  I split the output into the IQ components in LabVIEW to look at them.  Unlike UHD, NI-USRP does not have a way to specify to return just RX_A or RX_B, it will always return both.  Antenna RX_A corresponds to I and RX_B corresponds to Q in LabVIEW.  There are similar ways to use the Basic and LF TX boards as well (start a separate thread if you would like to get into details about that some time).

I know that UHD does do some signal processing to reduce the DC offset.  That could be causing what you are seeing on the WBX.  I believe that this correction is applied to the signal regardless of the daughterboard connected, but I am not an expert on this low level driver detail.  I don't think that this is caused because of the hardware, my guess it that it is an artifact of the DC offset removal.  I will need to double check with our software team to verify this.  You may also want to try emailing our USRP users mailing list.  There are a lot of knowledgeable and very active users that might be able to get a response to you faster than I can.

[cochenob]

Thanks Sarah-

I asked about the DC removal over on the listserv a few months ago.  They also suggested that it was the DC offset removal in the FPGA.  It was suggested both there and here that I tune the LO to an IF...capture I and Q at an IF to get away from DC...then let the CORDIC bring things down to DC.  However...this didn't work as well as expected.  As I noted towards the end of this post...it didn't seem like the CORDIC got all the way down to DC, as there was still some very low frequency content on I and Q (i.e. I and Q would drift very slowly).

I also noticed this weird bit of behavior.....

If I left the LO and Center Frequency the same....and picked a Center Frequency that did NOT get coereced by the driver (usually a nice integter like 100 MHz, or something like that)....then I would observe the above behavior where there'd be a "hole" around DC.  HOWEVER....if I picked any other frequency that DID end up being coerced (like 101 MHz, or 100.00000001 MHz), then it looked like I COULD mix down directly to DC.  Couldn't really figure that one out!

---

BC

[4sb]

Hello,

I have a question in regard to your previous experience with N210 and LFRX/TX boards along with LabVIEW programming. Is it possible to use N210 to acquire continuously 20MHz RTBW signal (for like an hour)?

Thanks!