Hi Hieraldo:
Your scope capture above already answers the test-pin question, so I’ll start there.
Correction to my earlier post: D_TST does not carry demodulated data. Per the 1999 TMS3705A manual’s test-pin section, in normal modes it carries the internal digital reset (active low); the current datasheet only calls it “for testing the internal logic.” Decoded data leaves on SCIO. The ESD clamp I mentioned belongs to the charge phase, where TI expects it, and the diagnosis resistor at SENSE is off during read. A_TST is the pin you want: it sits at the limiter input after the band-pass, so it shows the receive signal just before slicing.
The clipping is expected. The problem is what happens after the drive stops: the coil rings down at its own carrier for about 10 to 15 ms, starting near +1 V.
That ring-down is what kills the read. After the RF stage (gain 5) and the band-pass plus limiter, cumulative gain is well over 1000, so even +1 V at the input is far above the limiter threshold of about 5 mVpp. The ring-down sits at roughly 134.7 kHz, which the demodulator counts near 119, inside its accepted window of 112 to 125. So instead of calibrating from the tag preamble, it calibrates from your ring-down. It never sees a clean start bit, and the whole frame drops. A shorter ring-down is very likely the dominant reason removing the full bridge gets you a read: less stored energy, the preamble finally gets through, but the range drops.
Two fixes, both in your trace.
One: damp the tank actively. A resistor switched across the coil for the first 15 ms after drive-off bleeds the ring-down fast. That is much longer than the 100 microseconds. I estimated earlier. Your capture shows the real decay is 10 to 15 ms, so the damping window has to cover that.
and two: during the HDX read window, disconnect the discrete FDX front end from the coil. FDX is load modulation, so that front end presents a low impedance across the coil right when you are trying to receive. Open a series switch or MOSFET in that path, but only during the HDX read window so FDX still works otherwise.
To confirm without a full rework: on the same board, switch a bleed resistor across the coil that is on only for the first ~15 ms after TX-off, then capture A_TST. If the start-bit signature cleans up, that confirms the ring-down calibration hypothesis; if not, the limiter or threshold path is the next thing to check. Your SENSE capture already shows the envelope stays high for about 10 to 15 ms, so size the bleed window to cover that.