tomp wrote:
Essentially the amplifier used in the servo system looks at the input signal and compares it to the positional feedback from the sensor and if there is a difference between the input and what the sensor should feed back it does a correction. It can do so as an additional drive level if the driver is lagging behind or a decreased or even opposite polarity signal if the driver is overshooting the signal. The key is that the "correction" circuit must have two things:
First, there has to be enough gain/bandwidth product so that the correction can be applied in the right amount at the right time. If not, the correction circuit, if it lags, can actually introduce more distortion than it fixes. That is why up to now these servos have been used mainly on subs. The problem has not been so much the speed of the amplifiers as the delays caused by slow feedback sensors.
Second, the design of the circuit has to be smart enough to realize when the motor structure has run out of steam and any magnitude applied correction signal can no longer make the necessary correction. Otherwise the correction circuit can potentially send a large enough signal out that it can destroy the driver because the feedback sensor says there is not enough correction. That is one of the many hurdles that Velodyne had to deal with in the early days.
BTW, you have to remember with any feedback circuit, whether in amplifiers or speakers, an error has to occur before a correction can be made. The more competent the feedback circuit, the smaller the error and therefore lower the distortion can be. That is why it is always best to make the system as linear as it can be before applying feedback. If you start with an essentially non-linear system the feedback circuit has a very difficult time to try to make it linear.
Then of course there is the delay within the feedback network and the response time of the sensor at the woofer, which typically may be an accelerometer, an additional voice coil winding, the combined woofers back EMF read off two parallel 0.1 ohm resistors, or a strain gauge, that adds to the errors. I am not sure any of those schemes can replace a really well designed driver and enclosure method.