Myriad-RF Transmit I/Q Balance Calibration


#1

I am trying to better understand the requirements for properly calibration the Myriad-RF board and specifically the LMS6002D chip. The Transmist I/Q Balance Calibration is relatively vague on what needs to be done, but one of the calibration procedures it says is to apply a 1 MHz tone to the to the connector, X4, via a differential I/Q analog signal generator. The “recommended” generator is the Agilent N5182A MXG signal generator, but this is a very expensive piece of equipment. I am wondering what other hardware can be used to perform the same calibration as describe in the development kit pdf, or if there is another way to perform the Transmit I/Q balance calibration?


#2

First of all it would be best to understand the Math. When I first started RF development I simulated all the math in Matlab or do it by paper. If you have a understanding of the Math you can figure out how to calibrate the TX I/Q Imbalances.

There are really two TX signals you want to calibrate out:

  1. The TX LO leakage
  2. The TX Unwanted Sideband

Now imagine if you put a DC I and Q value into the dac’s of the LMS6002d. Than mixed it with a frequency, lets say 900 MHz. The output would be 900 MHz, since there is no lo leakage or unwanted it is impossible to calibrate those signals.

Now imagine if you want to use 3 MHz Sine wave into I and 3 MHz Cos into Q (Or vice versa depending on if you want upper or lower sideband). The Wanted TX signal would be 903 MHz. The unwanted would be 897 MHz. The Lo leakage would be 900 MHz. (Do the math so you understand how I came up with those numbers) The goal is to cancel out the unwanted and LO leakage. Register 41 and 42 Add a dc offset to calibrate out the LO leakage. If you hooked up your transmit into a spectrum analyzer you can manually calibrate out the LO leakage. This is only really helpful to visualize how the LO leakage cancellation works because in the real world you want the transmit going to some RF electronics and eventually an antenna. Therefore people loop the TX into the RX all the way back into the RX ADC’s. Than we take an FFT/DFT to find the frequency content. So what you do is change a value in the LO leakage register and than see the effect on the FFT (of course you are looking in the bin of the Lo leakage or Unwanted). Keep doing this procedure until it is tuned.

For I/Q unwanted signal calibration there is a Magnitude and phase imbalance. Again I recommend doing the Math on paper or in matlab. For Magintude Imbalance you just need to a scaling factor on I and Q till they are properly matched. Same thing with Phase imbalance you multiply the I Q by a phase to adjust. This is documented in LMS calibration guide I believe. Also check out the Lime microsystem Google group. A lot of questions are asked on that forum to calibrate the LMS6002d.

In conclusion, you don’t need a Signal generator to calibrate. You can create an NCO (numerically controlled oscillator) in software/FPGA to create a sine/cos wave into I Q and do the same thing.


#3

Hi Cory,

Thank you for the response. This gives me a good idea of what i need to do.