When a system has a linear response you can run it in traditional feedback control and use the Amplitude Controller to boost the command as needed to obtain the desired waveform levels. For example, if you want to achieve a waveform that is +/-100N you would select the load cell as the feedback transducer and run the desired waveform. The amplitude controller will automatically adjust the command waveform peaks/valleys until the +/-100N waveform peaks are met. Where this control approach comes up short is when the system response is non-linear or discontinuous.
Take for example a system that uses compression platens to apply a 1 to 100N load to a disk shaped UHMWPE sample(Fig I). Any time compression is applied to the sample the rsponse is somewhat linear. However, if the motor is retracted (ie: commanded in the tension direction) the upper platen looses contact with the specimen and the load drops to zero (Fig. II). Once the command changes back from tension to compression the load signal starts to increase at a constant rate. Figure III shows the system response which has a kink in the slope at the zero cros-over point. This is what a non-linear discontinuous response looks like.
If you run this system using the load cell as the feedback strange things can happen. For example, because of the spring rate of the test specimen and linear motor design, the motor works much harder to achieve 100N than it does to achieve 1N. It's kind of like lifting barbells.... it's easier letting them down once you have lifted them. Since the motor has such an easy time reducing the load (in fact the specimen is pushing it back towards zero load) there is a good chance it will overshoot the 1N point and the upper platen will lift off the specimen. When this happens the controller tells the motor to reverse direction (after all it is trying to maintain a 1N compressive load) and the platen (with nothing to resist it) begins to accelerate toward the test specimen. When it makes contact with the specimen the platen's speed translates into a specimen impact which spikes the load in compression. This impact triggers a high frequency oscillation that can only be stopped by de-energizing the motor. If you have ever seen this happen it can be alarming.
So how do we get around this non-linear discontinuity? You need to set your feedback as something other than the system load cell while at the same time setting up the Amplitude Controller to increase the system command until the desired load peaks are met. You can either set your system feedback as Displacement or Null. The Null feeback means that you basically set the feedback to a zero reading. Another way to do this (assuming your controller has this feature) would be to set the PIDs to zero and set up a Feed Forward input from the command signal into the control loop. If you use displacement control you set the starting displacement waveform with a peak/valley displacement level that you know will generate a load profile that falls between the 1 to 100N compressive load. You do the same for the Null and the Feed Forward approaches but instead of using displacement units you use volts or load units. Once running, the Amplitude Controller will adjust the waveform peak/valley until the desired load peak and valley are achieved. This process is easier than it sounds and once you have done it you will have a better understanding for how it behaves.
At MDT we use all of the tools available to provide the most accurate test possible and the Amplitude Controller is one of the more powerful ones in the toolbox.