On Sat, Jul 21, 2012 at 7:03 AM, Thomas Tsou thomastsou@gmail.com wrote:
I tried calibrating LO leakage cancellation again and found out that it wasn't calibrated enough. The settings are very sensitive and increasing the I and Q shifts by one made a large difference on the E4406.
This is my experience as well. Near the optimum point even change of calibration registers by 1 makes big difference. When you're far off, a change of a calibration register could be almost unnoticeable.
I've started adding notes about LMS calibration here: http://code.google.com/p/umtrx/wiki/LMS6002DCalibration Please write down there all useful notes about all types of calibration we're doing.
The RSA seems to compensate, though, because the same changes don't affect it at all.
How could it compensate? For RSA it should look like a genuine sine signal.
I also tried some different frequencies. At the not useful frequency of 240 MHz, we are not far from target values. At 945 MHz, it is better than before, but still too high. The 1900 MHz band is not so good.
Well, we're more interested in GSM900, but this clearly shows that there is an issue with the LMS PLL noise.