A phase splitter is a circuit that inputs a reference oscillation and splits it into two components, oscillating at the same frequency but with a fixed phase separation. In many cases of interest in RF design, the required phase separation is 90 degrees, in which case the circuit is known as a quadrature phase splitter. These circuits are required in a variety of applications, most prominently in transmitters and receivers for modern cellular communication systems.
For good accuracy, a quadrature phase splitter must produce output signals with amplitudes as closely equal as possible, and with phase separation as close to 90 degrees as possible. The first generation of phase splitter circuits worked by simply passing the reference signal through passive R-C or L-C networks. The phase and amplitude mismatch of these circuits was directly proportional to the mismatch in component values, making it difficult to achieve high accuracy.
A second generation of phase splitter circuits attempted to fix the amplitude imbalance by applying amplitude-correcting feedback around the passive networks. In these circuits, the R components in the R-C networks are realized with voltage-controllable resistors (for example, MOS transistors operating in the ohmic region). The amplitudes of the outputs are compared and an error signal generated, which is fed back to the voltage-controlled resistors to control the amplitude shift in the passive network.
The application of amplitude feedback results in good amplitude equality, but phase mismatch in the second-generation circuits is still sensitive to component matching. The obvious next step is to apply phase feedback in addition to the amplitude feedback, to achieve good phase quadrature even in the face of poor component matching. My phase splitter circuit represents this third generation of circuits, which employ both amplitude and phase feedback.
The phase feedback is particularly easy to apply in the case of a quadrature phase splitter. In this case you can simply plug the output signals into a multiplier, such as a Gilbert cell, and low-pass the product; the result is a DC signal directly proportional to the deviation of the signals from phase quadrature. When the two signals are exactly 90 degrees out of phase, you get zero output signal. This means you can feed back the output of the phase comparator to the passive phase shifting network, varying the phase shift of the network until the phase comparator output goes to zero.
This last point was first recognized by Joe Havens, who designed a phase splitter with feedback from a Gilbert cell type phase comparator. I then noticed that the amplitude equality of Joe's circuit was not perfect, and added in some amplitude feedback of the type used in the second generation circuits. However, the way in which I was combining the amplitude and phase feedback signals was not ideal; this point was recognized by Bruce McNeill, who proposed a much more elegant and efficient scheme for combining the two feedback signals. I then implemented Bruce's proposal to arrive at the present circuit, which we patented.
For the patent application, I wrote a (medium-complexity) explanation of the circuit's operation, which maybe eventually I'll convert to HTML and put on this page. Until then, however, please check out this report, which should suffice to convey the basic principles.
U. S. Patent Number 6313680 was issued for this invention, under the title "Phase Splitter," on November 6, 2001.
Dual Loop Feedback Phase Splitter, by Homer Reid
|Last Modified: 11/16/16|