TX influence on RX in TDD mode

I use two LimeSDR-USB (v1.4) as modems for data transfer. Modems operate in TDD mode. Tx and Rx modes change at approximately 380 Hz.
One antenna is used, connected through a switch. The switch provides approximately 30 dB of isolation between Tx and Rx.
When the power of its own Tx signal is low, everything works well. As the power of its own Tx signal increases, the Rx sensitivity decreases and the receiver malfunctions.
It is understood that this is due to intrinsic Tx power penetration into the Rx channel.

  1. Why does the intrusion of its own Tx signal into the Rx channel disrupt the operation of the modems if Tx and Rx operate at different time intervals (not simultaneously)? Is this a feature of the LMS7002?
  2. How to get rid of the influence of the own Tx signal on the Rx? It is clear that the problem can be solved by increasing the switch isolation between Tx and Rx to 100dB. Are there other ways?

Conducted additional testing.
Testbed # 1 is shown below.
Time diagram of work

LimeSDR USB # 1: RX_gain = 30, TX_gain = 40. TX slot - harmonic signal Fsin = 355kHz

Oscillograms at the RX input for LimeSDR USB # 2 (RX_gain = 30, TXgain - see Fig. 1-5)

In Figure 6, TX LimeSDR # 2 port is connected to a separate load (50 Ohm).

You can see that the TX gain affects the RX, although they work at different times!
In this case, the problem disappears when TX and RX are not combined into an RF-switch.
It turns out a problem in the RF-switch, which does not provide the required isolation, but the modem TDD mode !!!.

I would be grateful for any thoughts on this matter.

Switch LMS7002M transmitter output off during receive frame.

Please explain how to do this?
With which API commands or registers switch LMS7002M transmitter output off?

Your LNA+PA give you more gain, than switch isolation, so it is equal that your Lime TX connected to RX with 8 db gain. I think analog on/off tx time is more than reguired in tdd mode. You may use digital off by muting IQ. In addition, you should shift your bans up or down from zero IF poin to avoid leakage of residual TX signal at DC (center) point of the bandwidth. B
If you use external LNA, try to set RX gain near to 0.

Another approach is using of gated PA with activation threshold more than residual TX spower.