You might try researching under Gilbert cell. Barrie Gilbert is the inventor of the multiplier. He came up with it as variable gain control for use in Tek scopes.
There are two incarnations out there. One is a six transistor design that is used for RF mixing (reference SA612). These are not suited to analytic functions because they lack the distortion reduction circuitry found in Gilbert's original design.
The following site has the internal schematic for Gilbert's original multiplier:
http://www.intersil.com/data/fn/fn2477.pdf
The circuit is on page 5 and the critical elements are the two transistors in the upper left corner. They pre-distort the voltage to the multiplier cell, thus linearizing it.
In addition to this circuit, there is the one quadrant multiplier, which is based on log and antilog amps. National has a good treatment on this:
http://www.national.com/an/AN/AN-30.pdf#page=1
There are other techniques as well, such as converting value x into a PWM and using it to actuate an analog switch to chop between value y and 0 volts. The output is then filtered to remove the high frequency noise from the PWM carrier.
In all these techniques, though, none is as cost effective and stable as pulling you signal into the digital domain. I've repeatedly started to use an analog method, and I keep going back to digital.
- Mike