Different behavior is shown by the multibody backlash block, which is designed to more accurately model the backlash of gears. See picture below. Note: it is speeds, not angles. What I am referring to is initial constantness of speed when it changes direction. Also ignore staircase-like behavior: obviously, it is more accurate model then simulink’s one so it takes more physical processes into account.

You can compare it to the output of simulink backlash block (for convenience, see below from official example, note: this is angles). Here one can see initial angle constantness.

I think, simulink’s block does not simulate gears backlash correctly (and multibody block does). Gears disengage when torque changes sign => when speed changes direction (because at this moment motor will start pushing in the opposite direction then before, but there is no contact to the opposite gear => backlash occures. Exactly at this moment torque = 0 (changes its sign) and speed tries to change direction: before this moment gear was accelerating and after it starts to slow down). So this constantness should occure on the speed level. But disangagment should be calculated on the angles level. Therefore it can’t be simulated by simulink’s block (and indeed simulink’s multibody block shows different behaviour).
P.S. Sorry for my english, it is not my native language.Different behavior is shown by the multibody backlash block, which is designed to more accurately model the backlash of gears. See picture below. Note: it is speeds, not angles. What I am referring to is initial constantness of speed when it changes direction. Also ignore staircase-like behavior: obviously, it is more accurate model then simulink’s one so it takes more physical processes into account.

You can compare it to the output of simulink backlash block (for convenience, see below from official example, note: this is angles). Here one can see initial angle constantness.

I think, simulink’s block does not simulate gears backlash correctly (and multibody block does). Gears disengage when torque changes sign => when speed changes direction (because at this moment motor will start pushing in the opposite direction then before, but there is no contact to the opposite gear => backlash occures. Exactly at this moment torque = 0 (changes its sign) and speed tries to change direction: before this moment gear was accelerating and after it starts to slow down). So this constantness should occure on the speed level. But disangagment should be calculated on the angles level. Therefore it can’t be simulated by simulink’s block (and indeed simulink’s multibody block shows different behaviour).
P.S. Sorry for my english, it is not my native language. Different behavior is shown by the multibody backlash block, which is designed to more accurately model the backlash of gears. See picture below. Note: it is speeds, not angles. What I am referring to is initial constantness of speed when it changes direction. Also ignore staircase-like behavior: obviously, it is more accurate model then simulink’s one so it takes more physical processes into account.

You can compare it to the output of simulink backlash block (for convenience, see below from official example, note: this is angles). Here one can see initial angle constantness.

I think, simulink’s block does not simulate gears backlash correctly (and multibody block does). Gears disengage when torque changes sign => when speed changes direction (because at this moment motor will start pushing in the opposite direction then before, but there is no contact to the opposite gear => backlash occures. Exactly at this moment torque = 0 (changes its sign) and speed tries to change direction: before this moment gear was accelerating and after it starts to slow down). So this constantness should occure on the speed level. But disangagment should be calculated on the angles level. Therefore it can’t be simulated by simulink’s block (and indeed simulink’s multibody block shows different behaviour).
P.S. Sorry for my english, it is not my native language. backlash, deadband, dead zone MATLAB Answers — New Questions

​