Purpose/Design

The purpose of the charge pump is to source and sink current to and from the loop filter.   As shown in the diagram below, the charge pump has inputs U, D, and Tx.  Inputs U and D come from the Phase Frequency Detector (PFD).  When the VCO frequency is too low, the PFD sets the U signal high, which means that current needs to be sourced into the loop filter so that the voltage going in to the VCO will rise.  If the VCO frequency is too high, the PFD sets the D signal high, which tells the charge pump to sink current from the loop filter, causing the voltage in to the VCO to fall.  The Tx signal is high when the transceiver is in transmit mode, so when this input is high, the currents should be shut off.  Below is a general diagram of the charge pump circuit.

When U is high and Tx is low, the output of the NAND gate is low, which turns on the top switch.  This causes current to be sourced to the output.  When D is high and Tx is low, the output of the AND gate is high, which turns the bottom switch on.  This causes current to be drawn from the output.  If Tx is high, the output of the NAND gate is high and the output of the AND gate is low, so there is no current at the output.  A schematic of the charge pump circuit is shown below.

As stated in the diagram, all of the FETs are regular and most of them have dimension W/L = 1.5mm/0.5mm.  It can be seen in the diagram that the FETs used for the current sources are quite a bit longer and wider that all of the other FETs.  In simulation, it was found that the amount of current during current source was considerably lower that the amount of current during current sink.  Using longer channel FETs helped compensate for this.
 
 

Simulation

The charge pump circuit was simulated in Hewlett Packard Advanced Design System.  For the U,Tx, and D inputs, a digital bit sequence was used.  A schematic of the circuit used for simulation can be found in the link below.

The output voltage, which represents the voltage across the loop filter capacitor was first set to 1.5 volts, and the output current was plotted against the U, D, and Tx as shown below.  From the simulation below, when U is high and Tx is low, the charge pump is sourcing 8mA. When D is high and Tx is low, the charge pump sinks 8mA. If Tx is high, no current sourcing or sinking is present.  The case where both U and D are high is a case that should never occur, so it can be ignored here.
 

To make sure the charge pump can still source and sink current at higher and lower voltages, the circuit was also simulated with output voltages of 2.3 and 0.7 volts.  The output current waveforms for these simulations are shown below.

From these two simulations, it can be seen that the charge pump still sources and sinks about 8mA of current between 0.7V and 2.3V.
 

Layout

Layout was done by team member Aaron Orsborn in MAGIC for Linux.  A diagram of the Layout is shown below.  For readability purposes, this layout does not show probe pads.  There will be two probe pads to test both bias voltages at the current sources.  There will also be two probe pads to test that the digital logic is working correctly.