Phase monopulse sensor outputs wrong angle for weaker signals.

I am currently have a problem with my monopulse sensor. It is classical design with one TX antenna and two RX antennas. The main difference is that I use passive FET mixers instead of classical Schottky diodes. Distance between RX antennas allow to measure angles in range -45...+45°, but actually a little less, because TX main beam is designed to be much narrower (-30...+30°), so angle would not warp around.

I have perfect angle readings when using corner reflector indoors for distances from 1 to 9 meters and almost full angle range of 90° (-45...+45). I obtain angle by mapping received IF signal phase shift from -180...+180° to -45...+45° range. Distance and angles are all confirmed by few adhesive tape markers on the ground.

The problem: when testing sensor outdoors big angle error appears from approx 10 to 20meters. Instead of corner reflector small metallic cart is pushed away and moving in outgoing direction.
For simplicity here is an example when cart moving trajectory is aligned with sensor antenna axis (0 degree)
Cart moving from 0 to 10 meters: angle readings are good 0°, error is +-1°. Starting from 10 meters: angle "jumps" few points to the left from 0° to -7°, and then I get several readings around -7° (-6...-8°) like if cart position and angle was changed. At the same time I obtain correct distance, so in my opinion signal amplitude must be enough to obtain correct angle too.

The same problem appears if cart is pushed towards different angle:
Pushed at -20°, good readings to 10m, and then angle change to -27° after 10m.
Pushed at +10°, good readings to approx 10m, then angle goes from +10° to +3°.

I can't understand where this problem is coming from. I understand if signal is very weak (near noise), then monopulse phase will have many error readings with +- distribution. But in my case I observe almost constant phase error after certain signal amplitude is reached, and this ampitude is above noise. All angles are shifted to the left.

Currently my setup is:
1 TX antenna, 2 RX antennas.
Each antenna have FET LNA and FET mixer.
IF output of each antenna connected to an OpAmp with few MHz of gain bandwidth.

Could this unwanted phase shift of -7° come from LNA, mixer or an opAmp, or RC low-pass and high-pass filters?

1)LNA using FET. Antenna1-FET1,antenna2-FET2.
Is it possible that received signal reach certain amplitude, and as a result FET1 have S21 phase that differs from FET2 S21 phase by few degrees? How does phase shift through LNA depends on signal amplitude?

2) FET passive mixers. Each antenna have one FET for mixing. FetMix1, FetMix2
Assume incoming RF signal is strong enough to bias one FET more than another (each mixer is fed with LO and amplified RF). Then probably it changes S-parameters of FETs, even if they passive FET mixers. Maybe phase shift come from here?

3)OpAmps to amplify IF output of mixers. I use high frequency OpAmps, big margins, so even 100kHz modulation waveform should not be a problem. Classical inverting amplifier with low-pass and high-pass RC filters. Tests was performed with single stage OpAmp amplifier on each antenna output. Gain is below 30dB, no any obvious sine distortions in digitized IF signal. Difference in amplitude seems to be negligible after ADC: I observe sine signal from each antenna, amplitude difference is few percent, maybe 10% max.
Is it possible that OpAmp phase shift difference become significant for weak signals?

4) RC passive filters and other passive parts.
I only expect some DC shift in signal levels may be possible, and should not affect phase of signal.

Overall problem seems to be unwanted phase jump in phase monopulse when signal amplitude crosses certain value. Difference between antenna phases jumps by approx 28° (or 7° when mapped to physical position). It seems that this phase shift is amplitude dependent, but I am open to all ideas and suggestions.

I would be very grateful if someone lead me to the right direction of solving this problem.

What a nice work! I think, you should check your receive channels without antennas. Apply a test signal using a power divider into both channels. Adjust the input power similar to the received power in real experiments. I hope you have suitable attenuators. Then, you can realize how you are sensitive to the signal power. If you do not observe the problem there, antennas and also propagation issues are left to be investigated. What is the operating frequency? Did you calculate your far zone distance?

I am short on equipment. Frequency is 10GHz, currently using FSK modulation with step of 2MHz (max theoretical distance is 74.9m without direction sensing or 37.5m with direction sensing). Please comment on far zone distance? Do you mean transition from near field to far field? I thinked about ground reflections, but can't imagine what happens there. Cart position is correct, moving away having direct line trajectory, then around 15 meters trajectory bents to the left and I have few meters of wrong position data until it reaches too weak signal with expected random errors in distance and angle.

I put most effort to design aperture fed patch antenna array, vco and IF amplifier. Other parts mostly prototyping and trial and error. I already moved from block design to single board design.

I may try to apply low frequency signal instead of mixer outputs. Applying RF is a problem, I am thinking about what could be done.

Also I learned signal from commercial k-band monopulse sensor some time ago, it was based on mmic vco + mmic mixer. I hoping to see similar signal output. Signal from commercial sensor: when signal was weak angle started to "swing" to both directions + and -, and after few meters further distance (FSK phase) was totally unusable. I did not observed any bent of trajectory toward one direction. It was sometime more to the left, or to the right, but always in some different way depending on position and object type. Now I have many doubts and some disappointment, but still hope to push through.

Yes, I meant far field distance. I do not know how to calculate far field distance for a patch array. First, I had thought that such strange behaviors may happen in near fields. However, because the problem arises beyond a distance, actually, it is not looking near field related.
The problem may also be AM-PM distortion as you have pointed in your first post. As you mentioned, amplifiers may show different phase response depending on the input level. Have a look at https://na.support.keysight.com/pxi/help/latest/tutorials/am-pm.htm
If your channels have different AM-PM conversion/distortion characteristics, the problem you told may emerge.
I keep following the thread.

- - - Updated - - -

Yes, I meant far field distance. I do not know how to calculate far field distance for a patch array. First, I had thought that such strange behaviors may happen in near fields. However, because the problem arises beyond a distance, actually, it is not looking near field related.
The problem may also be AM-PM distortion as you have pointed in your first post. As you mentioned, amplifiers may show different phase response depending on the input level. Have a look at https://na.support.keysight.com/pxi/help/latest/tutorials/am-pm.htm
If your channels have different AM-PM conversion/distortion characteristics, the problem you told may emerge.
I keep following the thread.