Defective pll/vco chip?

[submmAstronomer]

Hi,

We recently bought an NI USRP 2920 (Ettus 210/WBX).  Currently, I am simply trying to output a few sine waves from the device using either the Labview or Matlab/Simulink driver.  

Observing the output on a spectrum analyzer, both the sine waves and the LO are unstable.  At 2 GHz, the LO moves around by ~200 Hz over the course of a few seconds.  It tends to drift downward in frequency and then periodically jump back to a higher frequency.  At 100 MHz, the motion of the LO is considerably less, drifting around in a range of about 20 Hz, but the behavior is still the same.  Note that both the spectrum analyzer and the n210/wbx are frequency locked to the same 10 MHz external rubidium source (the E LED is lit on the n210).

What I am trying to understand is if this behavior is expected and due to an imperfect calculation of frequency or if I have a defective pll/vco chip.  

Any assistance would be appreciated.

Thanks,
Loren 

[Sarah_Y]

Hi Loren,

The USRP 2920 has a TCXO crystal oscillator on it (see NI-USRP Help documentation » NI USRP-292x Specifications for more details).  According to the vendor's specs, the 2920 should have a frequency accuracy of 2.5 PPM.  However, this is not measured.  Based on that accuracy, at 2 GHz, you should be seeing a variation of +/- 5 kHz and +/- 250 Hz at 100 MHz.  That is all assuming that you are using the crystal on the USRP.  Since you are using a much more stable source, I would not expect to see that sort of behavior.

Let me know if that didn't answer your question or if there's anything else I can do, I'd be happy to help.

[Sarah_Y]

Hi Loren,

Having said this, I am also going to try a USRP that I have to see if I can reproduce similar behavior.  Would you mind posting your VI?  Also any screen shots of this behavior would be helpful.

[ErikL]

Hello Loren -

As Sarah said that type of behavior is expected with the internal reference and we expect much better performance with an external reference.  We replicated your setup here with an OCXO and a center frequency of 2 GHz.

We saw the following:

1. With the internal clock the frequency drift was consistent with your measurements.

2. When I attach the external reference (OCXO) and using the property node in LabVIEW to specify the RefIn Reference Frequency Property (using an niUSRP property node), we saw less thank 10Hz of drift over 10 minutes or so.

To enable the external reference, navigate to the NI USRP palette (Functions >> Instrument I/O >> Instr Drivers >> NI-USRP >> niUSRP Property Node.vi)  Drop it in-line connected to the purple reference wire before the while loop containing niUSRP Write Tx Data.vi.  Choose the property Reference Frequency Source (right click and set to Change to Write if necessary).  Then right click and select "Create >> Constant".  Choose RefIn as the constant.

For an example open the attached example.

[submmAstronomer]

Hi Erik, Sarah,

Thanks for the reply.  Indeed, you were right.  The front panel LED only indicated that the external clock was present but the pll/vco wasn't using this external clock.  Using the vi with the property node did the trick.

Thanks,

Loren