Using servo-walls to compress an isotropic sample until 100kPa

Hi. I have been trying to use six servo-controlled mesh walls to prepare an isotropic sample with a target pressure of 100kPa but encounter an issue that I could not prepare a "stable" sample.

I have 8000 spheres in the granular assembly. The particle size ranges between 1.1e-4m and 4.4e-4m, and the hard_particles is used with Young's modulus of 7e10Pa. Thus a tiny timestep of 5e-10s is adopted, less than one percent of both the Rayleigh time and the Hertz time. During the running, I extract the positions of six walls, based on which the area of each wall in contact with spheres is calculated. This allows me to apply a constant pressure multiplied by the area as the target_val (target force) for the servo-walls. When I use the parameters of servo-walls with kp=0.1 ki=0.0 kd=0.0 and vel_max=10, the target pressure of 100kPa can be nearly obtained after a number of steps, but the void ratio of the sample is still decreasing mildly (like a 5e-7 drop in 10000 steps).

To judge whether the sample is stable or not, I replace the six servo-walls with the velocity-controlled walls and put the velocities as zero, i.e., do not move these velocity-controlled walls. Then I let it run (relaxation steps) but notice the pressure of the sample is dropping (it can drop around 80kPa in a number of steps), signaling that the sample is not stable or ready to use. Since later I will apply constant volume shearing, if I see the reduction of pressure, I do not know whether the reduction is attributed to the shearing or the non-stable sample. I have played the values of kp near 0.1 and vel_max near 10 but could not achieve what I want. I have also used the mode auto but until now I have not got what I want.

Can anyone provide some insights or suggestions into this issue? Your help is really appreciated. Thanks.