Is this the correct way to measure the Back EMF Voltage?

[rstofer]

The thing about the tach is the issue of capturing the change or ramp in RPM in fine enough steps to be useful in determining inertia.  If the motor accelerates in near zero time, the tach will never see anything other than the final value.

Ramping of the terminal voltage in a coupled unloaded motor seems ideal.  Adding resistors across the terminal will add even more load to the driving motor.

I don't have any problem with inertia causing a huge curve in the various parameters.  I have attached a .zip of my model and the .mat files with the variable definitions.  I'm sure you know but 'load' DCmotor2.mat and 'open' DCmotor2.slx then run.

[rstofer]

About that 'b' value, I wonder if we should add just enough damping to account for the no load current perhaps including the gearbox.

[fishandchips]

I am still stuck with this problem so I bought an oscilloscope and used it to observe the input signal to the DC motor. It is a train of square pulses with amplitude 11.1V and width 300 ms. I also observed that with each square pulse to the motor, the shaft of the DC motor makes about 1.25 turns.

I used a signal generator block in simulink to generate one square pulse with the amplitude and width mentioned above. Then, I used it as a input signal to the DC motor model to turn a rotational joint. As soon as the signal generator generates the rising edge of the square pulse, the joint turns many times very quickly even I attached a very heavy load to it. I am unable to make the DC motor to turn slowly like the real one (i.e. shaft makes 1.25 turns in 300 ms).

[rstofer]

I don't understand the pulse.  What is the frequency and what is the duty cycle?  300 ms doesn't seem to be enough information (to me).  Unless it is a one shot pulse of 300 ms width.

If you pulse the motor and it turns 1.25 revs, does that mean it moves from stop to some rpm and back to stop in 300 ms?

If the model turns more than the real motor, there is something wrong with either the friction or inertia in the model.  I don't have a similar motor so I really can't do much in determining the constants.

It looks like 'b' in my Motor2 Matlab project is viscous friction.

I revised the project to display 'w' the total rotation and I get around 9 revolutions in 300 ms.  Reduced by 10x for the gearbox, this would be about 0.9 revs.  The thing is, inertia keeps it rotating.  The RPM very slowly drops off.  You can see this by changing the simulation run time from 0.5 sec to, say, 100 sec.  You will see that it takes nearly 1 minute to stop rotating.  So, we have a boatload of inertia and no friction.  If 'b' in my .mat file is truly friction, and I think it is but I forget, it's set to a very small number 5x10^-6.

Since I didn't know how to define a 300 ms one-shot pulse, the pulse generator sends a 300 ms pulse every 100 sec (I guess I could have chosen a bigger number) so don't simulate beyond 100 seconds without changing the generator.

Attached .zip file contains the .mat and .slx files.

[rstofer]

Set b=0.012 and simulation time to 2
Look at the 'w' graph, it maxes out at 12 which when reduced by the gearbox is 1.2 revs.
Originally b=5.0e^-6 which is a very small amount of friction.

I still don't have any confidence that after changing 'b' that the model represents anything like the real motor.  I'm still not convinced about inertia.  I suspect that the parameters still need a little tweaking.  I didn't go back and put in an 11.1V step function instead of the pulse.

ETA:  Nuts!  'w' is in radians, not revs.  w(t) is in rads/sec and so on.  So a little more tweaking of the constants will be required.  But, it seems like some adjustment to 'b' is the way to go.

[rstofer]

So, 1.25 revs is about 8.2 rads.  Multiply backwards through the gear box and w=82
b=1.67e-3 gets very close.