Feedforward and Feedback Working Together



next up previous contents
Next: Feedforward Details Up: Overview Previous: Feedforward

Feedforward and Feedback Working Together

At this point the poor reader may be somewhat confused. Not only do both feedback and fast feedforward control the same phase shifters, but the beam energy that feedback measures is being constantly adjusted by feedforward. How do they keep it all straight and avoid confusing each other?

First consider the sharing of the sector 17 and 18 phase shifters. This is handled with special feedforward hardware. In simple terms, it has two inputs: the amount of energy feedback wants the two sectors to contribute and the amount of energy fast feedforward wants to contribute. It effectively adds these two requests and calculates the necessary phase settings to obtain that total energy gain.

The two systems share the sectors (rather than using two separate pairs of sectors) for two reasons.

  1. Only two phase shifters are needed instead of four.
  2. The two systems can share energy headroom. That is the sharing avoids the case where feedback complains that it needs to increase the energy but can't because its two sectors are contributing their maximum possible amount (both phases set to the BNS phase) while at the same time the feedforward sectors have several hundred MeV of spare energy because at the moment the beam intensities are not at their maximum values.

To understand how feedback and fast feedforward avoid confusing each other it is best to consider a few examples.

  1. Assume the klystrons' energy gains remain constant but the positron intensity suddenly drops (because the flux concentrator tripped off). Fast feedforward will detect the intensity change, calculate the expected change in beam loading, and adjust the phase shifters before the low intensity pulse is accelerated. Assuming that all the fast feedforward gains are set correctly, the low intensity pulse will have exactly the right energy. Feedback will see no energy change and hence will do nothing.
  2. Now take the previous case but assume that the feedforward gains are not set exactly right so it makes a 10% error in its calculation of the beam loading. In this case feedback will see a small energy (only 10% of what it would have been if fast feedforward did not exist) and will adjust its outputs to correct it. This takes several beam pulses.
  3. Assume the beam intensities remain constant but Klystron 5-1 trips off. The beam intensity will not change hence fast feedforward will not change anything. Feedback will read its BPMs, detect the energy change, and adjust its outputs to correct it. This will take several beam pulses.



next up previous contents
Next: Feedforward Details Up: Overview Previous: Feedforward



SLAC Controls Software
Fri Nov 4 11:34:56 PST 1994