PLL With Dual Modulus Prescaler

This example shows how a PLL (phased locked loop) model works with a dual modulus prescaler frequency divider.

model = 'MSPLLDualModulusPrescaler';
open_system(model);
sim(model);

PLL Model Architecture

The PLL is relatively simple and consists of the following:

• A pulse generator block as an ideal clock reference source operating at 10 MHz.

• Phase and Frequency detector (PFD) block from Mixed-Signal Blockset™.

• Charge pump and loop compensator filter modeled with Simscape™ components with values specified in the subsystem mask.

• Ring oscillator based Voltage Controlled Oscillator block from Mixed-Signal Blockset™.

• A spectrum analyzer block from DSP System Toolbox™

• A dual modulus prescaler from Mixed-Signal Blockset™ that implements a divider ratio equal to 8*18+6=150, thus generating a PLL output frequency of 1.5 GHz.

Dual Modulus Prescaler

A prescaler is typically used before a programmable divider to get the frequency down to a manageable value for the programmable divider which is a more complex design than the fixed value prescaler. The consequence of this design is that the frequency resolution of the PLL is now ${\mathit{F}}_{\mathrm{ref}}•\mathit{N}$where N is the prescaler divider value.

A clever solution to this problem is to dynamically switch the prescaler division between N and N+1 by using two low frequency programmable dividers. These dividers are clocked with ${\mathit{F}}_{\mathrm{vco}}/\mathit{N}$or ${\mathit{F}}_{\mathrm{vco}}/\left(\mathit{N}+1\right)$which makes them far easier to implement. By dithering the divider ratio between two factors you can achieve the desired resolution.

open_system([model '/Dual Modulus Prescaler'],'force')

If the two counters are called P and S the overall frequency division is $\mathit{N}•\mathit{P}+\mathit{S}$.

Since ${\mathit{F}}_{\mathrm{out}}={\mathit{F}}_{\mathrm{ref}}•\left(\mathit{N}•\mathit{P}+\mathit{S}\right)$ and P has a resolution of 1 the PLL resolution is back to ${\mathit{F}}_{\mathrm{ref}}$. Aside from more hardware (two counters) and the complexity in programming them, there is one other important detail to consider and that is S is bounded in value and needs to be in the range of $2\le \mathit{S}\le \mathit{P}$. Therefore certain values of N may not be realizable.

Additional Information

In order to achieve optimal simulation duration and accuracy, change the default solver settings. In particular, this model uses 'ode45'.

myConfigObj = getActiveConfigSet(model);
openDialog(myConfigObj)

You can inspect the solver settings in the Model Configuration Parameters for more details.