I have tested the AD9361 with 2 FMCOMMS3, both configured with 50 MSPS and 12,5 Mhz RF, setting in the device tree 1 Channel mode. The emitting AD9361 has a Tx LO Freq = 2,4 GHz and Rx LO Freq = 2,45 GHz, while the receiving signal has Tx LO Freq = 2,45 GHz and Rx LO Freq = 2,4 GHz. I set different frequencies in the Tx and Rx channels of each AD9361 in order to avoid the VCO pulling effect.
I send the sequence [011111111111, 100000000000], each value repeated 4 DAC_Valid pulses, through the DAC_Data_I0 input of the AXI_Ad9361 IP core and set the rest of the DAC_Data channels to 0.
I receive the emitted signal through both I0 and Q0 and I encounter several phase shifts (180º) between them. Sometimes they occur in I0, sometimes in Q0, sometimes one phase shift is followed by the next one only ~50 samples later, sometimes it occurs hundreds or thousands of samples away.
Next, I tried feedbacking the Tx and Rx channels of only one FMCOMMS3 and I found that those phase shifts do not occur within the same transmission as before, but each time I initialise the board I find a different phase between I0 and Q0. Note that I still find the emitted signal at both I0 and Q0 channels of Rx.
Why do I receive DAC I0 through both ADC I0 and Q0? Why those phase shifts? Do I need to synchronise both AD9361 in order to solve these problems? If so, why do I still find some of those problems when feedbacking one AD9361? In the case I needed to synchronise both AD9361, how could I perform a wireless transition when they are far away from each other?
I mess up the screenshots of the IIO Oscilloscope. Both Tx and Rx are working at 12.5 MHz RF.
In both boards.
TX and RX have random phase differences which change every time the VCOs are retuned. The AGC can also cause phase shifts over time.
A phase rotation will spread energy across the I and Q channels.
The AD9361 provides no synchronization between disjoint devices, this is left up to the implementer and is specific to your waveform.
Would using external oscillators for both Tx and Rx channels solve this problem?
In the Chipscope screenshot I am using 71dB Rx manual gain control.
In the case of using 2 FMCOMMS3, would using the same Reference Clock alleviate the phase shifts that, contrarily to what happens with only one board, occur over and over again within the same initialisation?
When feedbacking only one board, is there any general method to make the relationship between the phases of Tx and Rx deterministic along multiple startups?