I am a little confused, since the Rake Receiver has one finger for each multipath path, and the channel estimator models the impulse response of the channel and can be used to find the multipaths, why do I still need a Path Searcher?
Freescale AN2253 "Channel Estimation for a WCDMA Rake Receiver" by Ahsan Aziz 2004 has a pretty good explanation -- see Figure 1.
The signal from from an antenna is split into several fingers.
Each finger takes that input signal and produces one output signal by:
Then the receiver combines the output from all outputs from all the fingers from all the antennas, hopefully producing a combined signal with a better signal/noise ratio than it would get without all this rake complexity.
The part of the radio that independently tunes the delay for each path in part (a) is the Path Searcher. The part of the radio that independently tunes the complex attenuation for each path in part (c) is the Channel Estimator.
People familiar with AM and FM radio may think that part (c) is the only one needed -- in those kinds of radios, time delays have approximately the same effect as phase changes. But with spread spectrum technology, time delays act very different from phase changes. A phase change of +20 degrees can be compensated by either an additional phase change of -20 degrees or a phase change of +340 degrees or a phase change of +700 degrees -- all three methods of compensation are identical. A spread-spectrum signal with a relative time delay of -21.5 chips must be compensated by a time delay of close to +21.5 chips; anything below +20.5 chips and anything above +22.5 chips is useless.
The part of the radio that "can be used to find the multipaths" is the Path Searcher. I suppose you might have one integrated component that simultaneously tunes the time delays in (a) and the complex attenuations in (c); that component includes the Path Searcher and the channel estimator. (This happens often with highly-integrated electronics -- you have a bunch of conceptually distinct blocks on the overview block diagram, but often there is some single physical component that is shared between two blocks).
The RAKE-receiver is a simple parallel correlation receiver. The advantage is, that (if you know the path delays) you can delay the signals and then correlated with the spreading sequence. Afterwards just combine the results.
Take a look at the picture below (source: wirelesscommunication.nl). It shows the principles.
You need the Path searcher to find the beacon signals from other basestations (eg 3G towers) in order to help with smooth handover, communication must take place to multiple basestations Jason