You can see it right here in this picture that was already posted.
View attachment 54834
At the front, the radar would pass through the radome and hit the plane's radar. If the J-10's radar's face is set straight, like it was in the original, the signals from the threat radar would bounce directly back to the threat. But the J-10B & C's AESA is set at an inclined rest angle, the signals would bounce up and not towards the other radar.
Radar signals would also be bouncing at the firewall separating the radar and the back end. In here, the firewall itself is angled. That's more bounce away in an angle.
Another area of bounce would be the rim of the radome, the seam between the radome and the metal. On the previous J-10, like the firewall, it would be straight. On the angled radome, the rim around the radar would also be angled to bounce upwards.
You don't want this. Firewall, radar, the radome rim, all straight.
View attachment 54836
You want this. The radar, the wall behind it, the rim around it, all angled.
View attachment 54838 View attachment 54839
Another area of radar bounce would be the rim that makes up the intake. The original J-10's boxy and straight intake would bounce radar waves back. Here, the intake is angled to bounce upward. Note the angle of the diverter intake's rim and the rim of the radome has the same angle. That's intentional, so the radar that is bouncing off on both is towards the same upward direction. You want less directions to scatter radar around.
The DSI intake further bounces radar waves away from the engine intake.
There are things that still need to improvement however. Most notably is the circular rim in front of the cockpit over the hud. Refueling probe doesn't help either. And there is always that tall vertical tail. But from a head on perspective, RCS would be reduced from the angled radar, firewall, the rim around it, as well on the angled rim of the diverter intake and from the DSI itself.
View attachment 54835