IQ Imbalance and DC offset in sample streaming project

[irad]

Hi, 

I am running the NI USRP Simple Streaming project (without any modifications) on a 2944R and noticing a large DC offset and I/Q imbalance. The screen captures below show

outputs without anything connected to the RF 0 receive channel. 

DC offsetIQ imbalance

These vary over time. For example, at the start of one run the DC offset was at -60 dBm. After about ten minutes it sank down to -80 dBm and after another ten minutes it crept back to -70 dBm and higher. 

How do I turn on impairments correction in the Simple Streaming project? There is no 

control on the front panel. If it is already turned on by default, it isn't producing great results. 

Regards, 

Ivan 

[psisterhen]

The corrections are enabled by default in the sample project, and I believe you are seeing good results.  I would say LO Feedthrough at -70 dBm is good for this device.  However, you have some additional options.

There is a self-cal utility for Rx DC Offset (LO Feedthrough).  I recommend the following steps

1. Backup your existing correction data in case you don't like the results of the self-cal (run <LV DIR>\vi.lib\LabVIEW Targets\FPGA\USRP\niusrprio_tools.llb\Backup Persistent Device Data.vi"
2. Warm up the device (run an acquisition for 10-20 minutes)
3. Disconnect any connections to the RF terminals and run "<your sample project folder>\Host\SubVIs\niusrprios Self Correct Rx IQ Offset.vi"
4. If the results aren't any better, you can restore the previous correction data using <LV DIR>\vi.lib\LabVIEW Targets\FPGA\USRP\niusrprio_tools.llb\Restore Persistent Device Data.vi"

If this is still unacceptable, you can

1. Use LO Offset to move the LO Frequency outside of your band of interest
2. Implement a high-pass filter in your Rx DSP chain to remove DC offset on-the-fly

[sunaenae]

Hi Ivan,

Thanks for reaching out! Unfortunately, the USRP is an uncalibrated/unwarranted device, so this behavior is expected and is somewhat common. The I and Q signals may go through separate signal paths that could cause different amounts of imbalance in the final reading. It also may be due to temperature fluctuations when starting acquisition/the device.

There are a couple steps we can take to potentially compensate for this, though:

• Manually calibrate the device (i.e. terminate the RX channel to find the average gain offset for I and Q, and then manually set the software to compensate for that)
• If you've programmed your device using UHD before, you can follow these directions to self-calibrate the device

Also, it would be good to know how you are planning on using this device/what measurements you're trying to take. The IQ imbalance may not be ultimately critical in terms of some tests.

Thanks,
Sunayna

[zengmimei]

hi, did you solve this problem?  I am also have this problem,and I have no idea how can I solve it

[irad]

As psisterhen suggested using the LO offset to shift the DC bias out of the filter passband can cut the DC bias quite substantially. The I/Q imbalance is also diminished. However this only works if you reduce the bandwidth. It doesn't work if you need to use the full sample rate of 200 MS/s. 

Calibration would probably help but I haven't had time to explore this option yet. 

[zengmimei]

Thank you !