RF Design Magazine


Low power, giga speed programmable divider in 0.18 m CMOS
Jul 1, 2006 12:00 PM  By Girish N. Jadhav
 
Resources
Spotlight on Automotive Wireless Connectivity

For the PDF version of this article, click here.

To meet the stringent low-power consumption and high-frequency operation at low switching noise of on-chip programmable dividers used in the design of PLL circuits, this article will focus on the design and simulation of a new programmable divider architecture for use in the 2.412 GHz to 2.484 GHz frequency bands. The design is based on 0.18 micron CMOS technology.

By comparison, the conventional programmable divider consists of a dual-modulus prescaler (DMP), a program (P) counter and a swallow (S) counter and is depicted in the block diagram shown in Figure 1. A DMP is used to get higher resolution. A DMP allows the prescaling factor to be changed between N and N+1. Prescaler divides 2.4 GHz down to the 75 MHz, which is used for the following loadable counters to minimize the silicon area and power. Here, a program counter acts as coarse tuner and a swallow counter as fine tuner. The counter/divider M is given as:

M = (N+1) S + N (P - S) = NP + S

For proper functioning of the counters, S should be smaller than P and S must also be less than N. The numbers of N, P and S should be chosen carefully according to the maximum limitation of the allowable input frequency of the counters.

The conventional design has the following issues:

  • Requires two loadable down-counters.

  • Reduces speed, due to delay introduced in the counter path.

  • Increases design complexity while designing the loadable flip-flops.

  • Reduces robustness of the circuit.

  • Offers high power consumption at ~24 mW.

  • Increases hardware and silicon area.

Proposed programmable divider

The proposed programmable divider consists of a dual-modulus prescaler (DMP), an up-counter, an equality detector, an analog MUX and PCFR logic. The conceptual diagram is shown in Figure 2.

In Figure 2, Fin = sinusoidal input frequency, Fout = output frequency, CR = coarse register, FR = fine register and PCFR logic = prescaler, coarse, fine selector and reset generation block. Now the formula for the proposed programmable divider is:

M = {( N * CR ) + ( FR * 2 )}

The DMP divides sinusoidal input signal by either N or N+1. That is, DMP will perform divide-by-32; when MC = 0, and divide-by-33; when MC = 1. Thus, the input frequency is divided down to 75.375 MHz, 77.625 MHz from 2.412 GHz, 2.484 GHz respectively for MC = 0. The up-counter block operates with this downed frequency. The up-counter is itself a programmable frequency divider whose function is to increment the counter output on the rising edge of the downed frequency (CLK) and reset the counter output asynchronously when reset = 1.

The up-counter is an asynchronous divider constructed by combining 8-cascaded divide-by-2 circuits. It starts counting from 0 to 2 L, where L is the length of the equality detector. (In this design L = 8).

Initially, the MUX select signal is low, which means the equality detector compares FR value with the output of the up-counter, if both values match then a pulse is generated. This pulse is used to toggle (switch) the MUX select signal from low to high.

After toggle, the MUX select is high, which means the equality detector compares CR value with the output of the up-counter, if both values match then again a pulse is generated and this pulse is used to reset the up-counter.

Both coarse and fine detected pulses are generated and merged on the same output of the equality detector.

We can observe that both coarse and fine generated pulses are merged on the same output line of the equality detector. It's important to separate coarse and fine pulses, which is implemented in PCFR logic.

Coarse pulse is used to reset the up-counter, set mode control (MC) and toggle the MUX select signal. Fine pulse is used to reset the MC and toggle the MUX select signal.

Implementation of PCFR logic

The implementation of PCFR and pulse separation logic is as shown in Figure 3.

Simulated vs. calculated results are shown in Table 1 for Fin = 2.412 GHz, prescaler N = 32, Tout = 1/Fout = 1/(Fin/M) and M = {( N * CR ) + ( FR * 2 )}

Based on the above, the current consumption was 2.84 mA at 1.3 V, with power consumption at 3.692 mW at 1.3 V.

References

  1. Xi Li, “Evaluation of RF CMOS IC Technology for Wireless LAN Applications,” A proposal presented to the graduate school of the University of Florida in partial fulfillment of the requirements for the degree of Doctor of Philosophy, February 2002.

ABOUT THE AUTHOR

Girish N. Jadhav is a senior member of technical staff 1 at Ikanos Communications in Fremont, Calif. where he is involved in silicon characterization of analog transmitters and receivers. Prior to Ikanos, Jadhav worked on analog PLL design as a design engineer at Acer Laboratories. He holds a BE in electronics and telecommunication from Karnataka University in India.

Table 1. Simulated vs. calculated results.
Cases Programming CR & FR Calculated Simulated
Course Register C R CR7 … CR1, CR0 Fine Register F RFR7 … FR1, FR0 M Fout (MHz) Tout (nsec) Fout (MHz) Tout (nsec)
I 0000 1110 → CR=14 0000 0100 → FR=04 456 5.28947 189.0547 5.28946 189.055
II 0001 1110 → CR=30 0000 1100 → FR=12 984 2.45121 407.9601 2.45121 407.96
III 0010 0100 → CR=36 0000 1000 → FR=08 1168 2.06506 484.2454 2.06506 484.246



June Defense
 
Back to Top