I am involved with a research project for a medical device using a AD 8333 EVAL board. I am having a problem with measure phase accurately. As an example in one set up we have a two channel internally synchronized waveform generator. One channel goes to the Local Oscillator and the other goes to the IN1 channel on the EVAL board. Then the I and Q outputs go directly into an Oscilloscope. The waveform generator settings are: LO at 20 Mhz, 1 Vpp, square wave (w/o a DC offset); and IN1 at 5MHz and roughly 100 mVpp (w/o a DC offset), sin wave, both phase settings are 0.0. The phase difference is set to 0 degrees. The board is powered by +5 V and -5V. In this particular example we are looking for I and Q outputs (two straight lines on the oscilloscope) and calculate both magnitudes and phase. When we keep the magnitude of the IN1 input voltages constant and vary the phase of IN1 input, the results match our expectations of
phase= atan(Q/I) + a constant
When we keep the phase constant and change the magnitude of IN1 the magnitude of the output conforms fairly well to
But, the inexplicable results we are seeing is that if the magnitude of the input is changed from say 100 mVpp to 101 mVpp the ratio of Q and I as well as there signs will change and therefore the phase. When the magnitude is changed back to 100 mVpp another configuration of signs on I and Q will be present. Looks like the signs of I and Q randomly change as amplitude of IN1 changes. The signs of (I,Q) can be (+,+), or (+,-), or (-,+), or (-,-) randomly for the same (or different) amplitudes of IN1 as we change the amplitude of IN1. When we further changed the amplitude of IN1 from 50 mV to 200 mV, the phases were randomly changed a wide range. We verified that the function generator has produced phase locked signals for LO and IN1.
Looks like AD8333 will not give us the correct phase as the amplitude of rf signal changes, while as the phase information is important to us in beamforming. Is it normal for AD8333? Is anyone can help us on this?