Output signal quality

[cochenob]

I've been noticing some strange behavior for an application I'm developing, and have possibly tracked the issue down to the Tx path of the USRP.  Instead of attaching my code...you can replicate this behavior using the "TX Finite Sync.vi" example with the following parameters...

Carrier - 70MHz

I/Q rate - 2MS/s

Gain - -10

#samples - 1000

tone frequency - see below

Attached are 4 pictures I took of the USRP going directly into a LeCroy scope.  The scope was fixed at a 20Gsps rate, which is more than sufficient to capture the FM modulation around the 70MHz carrier.  Each acquisition was captured using a single trigger to get a "snapshot" of USRP output.  The scope has a deep memory as well so that it can acquire the entire USRP output burst.  In other words...I took care to make sure I'm not seeing any artifacts of the scope trying to retrigger off a signal of changing frequency and I'm not undersampling. 

I show two situations...one using a 1Hz tone....which is basically me just trying to fudge the radio into outputting a single CW tone at 70MHz.  Here is a long capture which shows that, as expected, the amplitude is constant. 

However, if you zoom in, you see that the output is nowhere near sinusoidal...though it does appear to have the correct frequency.

The second situation is a 250kHz tone.  The 'long' capture shows a varying envelope (I am confident this isn't the digital scope aliasing, since the sample rate is enormously high).  At first I thought that this "extra" amplitude modulation was just due to the uneven/uncalibrated magnitude response of all the RF components...but...I would expect then that the envelope 'peaks' would be spaced at the modulation frequency.  It isn't.  It's about 4 times less.  So, not sure what's going on here.   

In any case...if you zoom in, you will notice that the output is again rather odd.

So...doing FM (or FSK) modulation results in 'extra' amplitude modulation since the components aren't calibrated (expected), but manifests itself in a manner that doesn't jive with the rate of modulation (unexpected).  Additionally, the carrier has an odd shape that isn't sinusoidal (really unexpected).

My thoughts was that this may have been some relationship between the I/Q rate and the modulation frequency.  However, increasing the IQ rate relative to the mod frequency does not seem to make a difference.  I'm not sure if there's a good rule of thumb for determing this relationship.  For example...if you were doing FSK instead of analog FM....how many samples per FSK symbol does the modulator need to faithfully represent your signal at the RF carrier?  One?  Two?  16? 

Regardless...what is the reasoning for this poor RF carrier waveform shape?  Is it to be expected?  If so, why?

Thanks!

---

Brandon

[ErikL]

That waveform seems inconsistent with what we would expect. Can you 1. Confirm you are using a NI USRP-2920 and 2. Share your VI generating the waveform?

[cochenob]

All that info and I forgot to add what radio I was using!  Yes...it's the 2920.  I could post my code...but the pictures I sent were using one of the shipped examples - "niUSRP EX Tx Finite Sync.vi".  I'm not really using that for any of the "sync" features...rather, it was the only example that did a finite 'burst' of data (which was similar to the application I'm trying to develop).  I used this example as a "sanity check" to convince myself it was the hardware and not my code.

---

BC

[ErikL]

Decrease full scale waveforms by multiplying the IQ signal by 0.7.

[cochenob]

Good idea.  This changes things....but not for the better.  The 'shape' of the modulation just gets weirder.  I'm going to try all of this on my 'Ettus radio' (with the low frequency daughter boards) and see if I notice the same thing.

[Anthony_F]

---

@cochenob wrote:

I show two situations...one using a 1Hz tone....which is basically me just trying to fudge the radio into outputting a single CW tone at 70MHz.  

---

You should be able to achieve CW output by having a constant component for I and a zero component for Q. As Erik suggested, use a max value around 0.7

All of the images were rather small and could not be enlarged. Is it possible to attach some high resolution pictures? Also what LeCroy scope are you using? (I am assuming the 20GSa/s rate is being done via time-interleaved-sampling.)

I would also be curious to know if you set the LeCroy scope for a 50 ohm input impedance or if it was unchanged.

Our applications engineers should be able to verify this with their USRP and might have additional questions.

[cochenob]

Attached is a screen dump where I modified the example VI code to output a single tone by setting I=.707 and Q=0. 

Clearly, I'm still figuring out how to use the save/capture features of the scope, but this should at least give you a better sense of what's going on.  Note that the modulation frequency (70MHz) appears to be correct.  And yes...I'm going into 50ohms.  Also...the pk-pk amplitude of ~200mV jives well with the -10dBm output that I have configured (give or take a bit due to the uncalibrated response).

I'm using the Lecroy 640zi Waverunner.  http://teledynelecroy.com/oscilloscope/oscilloscopemodel.aspx?modelid=4778

4GHz analog BW.  20GSps (40GSps interleaved).  I did configure the scope for a lower sampling rate, which yeilded the same results. 

---

BC

[ErikL]

I think I figure it out.  Can you share a screen shot of the time domain waveform on your LabVIEW front panel with your IQ rate and number of samples (waveform size)?

Erik

[cochenob]

See below.  These were the settings of the scope shots I just sent this AM.  You can ignore the "tone frequency" input...as I replaced the FM modulation with DC values for I and Q (as seen in the graph).

The IQ rate and number of pts is slightly different than yesterday evening's post....but I've tried a few combinations, all with the same result...though judging from your request, perhaps I wasn't very exhaustive?

---

BC

[ErikL]

I believe there is a problem with the relationship between number of samples, IQ rate, and tone frequency. It's hard to make recomendations with out seeing the VI itself.

Can you confirm that the shipping example with no chanes works as expected?