3D frequency dithering for DC-to-DC converters used in multi-mode cellular transmitters

Description

RELATED APPLICATIONS

This application claims the benefit of provisional patent application Ser. No. 61/60,116, filed Mar. 13, 2009, the disclosure of which is hereby incorporated herein by reference in its entirety. This application also relates to co-pending U.S. patent application Ser. No. 11/756,909, now U.S. Pat. No. 7,928,712, entitled “Low Noise Fast Dithering Switching Power Supply”, filed on Jun. 1, 2007.

FIELD OF THE DISCLOSURE

The present disclosure relates to switching power supplies that may be used in radio frequency (RF) communications circuits.

BACKGROUND

A desirable component for a mobile phone is a switching power supply such as a direct current to direct current (DC-to-DC) converter that transfers energy from a source to a load. Typically, DC-to-DC converters offer much greater efficiencies than do linear voltage regulators while regulating power being transferred to a load such as the circuitry of a mobile phone. However, linear voltage regulators at present offer an advantage by generating much less spectral noise than DC-to-DC converters. Moreover, incorporating DC-to-DC converters into mobile phone circuitry is made more problematic by third-generation (3G) and later mobile phone standards, which are increasingly restrictive with regard to spurious radio frequency (RF) transmissions. As a result of a need to reduce spurious RF transmissions from mobile phones, any leakage of DC-to-DC converter switching noise into a mobile phone's transmitter circuitry is preferably reduced. Traditional attempts to reduce switching noise leakage have resulted in either prohibitively expensive filtering components or unacceptable results when mobile phone emission standards are applied.

Making the switching noise of a DC-to-DC converter less periodic by a frequency dithering of the DC-to-DC converter's switching signal is helpful. However, unlike a frequency dithering scheme for a crystal oscillator that provides acceptable electromagnetic interference (EMI) performance, a frequency dithering of a DC-to-DC converter's switching signal often results in an unacceptable increase in output voltage ripple. One reason for the increase in output voltage ripple is that a fundamental criterion for steady state operation of a DC-to-DC converter is violated by using frequency dithering.

Typical DC-to-DC converters have two distinct phases of operation. A first phase occurs when energy is stored in an inductor's magnetic field, and a second phase occurs when the stored energy is transferred to a load. An imbalance of energy transfer will often result in either an energy buildup in the inductor's magnetic field or a collapse of the inductor's magnetic field. As a result, undesirable output voltage variations will likely occur as an inductor current creating the inductor's magnetic field is integrated by one or more output filter capacitors. A resulting output voltage ripple will increase as a charge transfer to the filter capacitors becomes imbalanced. As illustrated in FIG. 1, the inductor current will show an imbalance as a pulse width modulated (PWM) switching signal is stepped from one period to another. In other words, the changes in switching signal frequency may introduce a variation of the inductor average current, which in turn results in an undesirable output voltage ripple. In FIG. 1, the switching signal labeled “PWM” may also be referred to as the “Lnode” signal. The “Lnode” signal is located on a node of an inductor that is either driven low or high by one or more power Field Effect Transistors (FETs) making up a part of a DC-to-DC converter's circuitry.

A challenge is how to change the frequency of a DC-to-DC converter without introducing a charge imbalance. FIG. 2 illustrates an adjusted inductor current resulting from an ideal PWM waveform. The ideal PWM waveform, shown in FIG. 2, and a corresponding flywheel voltage, will provide a constant average current independently of a frequency deviation. A plurality of transitional periods (T₁, T₂, and T₃) may be calculated. Moreover, an output voltage ripple can be reduced to the integral of an inductor ripple current. A first order linear design approach based on inductor current is preferably used to simplify an implementation of a DC-to-DC converter having reduced output voltage ripple.

A transitional period required when switching between a switching frequency Fn and another switching frequency Fn+1 may be defined as:

T Transitional = ( T F ⁡ ( n ) + T F ⁡ ( n + 1 ) ) * ( 1 - D ) 2 , Eq . ⁢ 1
where D is the Duty cycle, and T_F(n)+T_F(n+1) are the periods of the two frequencies Fn and Fn+1.

Equation 1 above may appear to be an over-simplification, but the generation of a PWM signal is as simple as comparing an error signal with a ramp using a voltage comparator. A fast dithering method of generating a PWM switching signal that allows the PWM switching signal to be changed every cycle without introducing a significant output ripple voltage is disclosed in co-pending patent application Ser. No. 11/756,909, now U.S. Pat. No. 7,928,712, filed Jun. 1, 2007.

Spurious RF transmissions due to leakage of DC-to-DC converter switching noise into a mobile phone's transmitter circuitry may be reduced somewhat by a fast dithering of a DC-to-DC converter's switching signal. The fast dithering of a switching signal is a two-dimensional process that spreads a plurality of switching frequency spurs over a wide range while improving output voltage ripple rejection. While fast dithering of a switching signal provides good a design starting point it has been discovered that every data burst (FIG. 3) from a Global System for Mobile communication (GSM) mobile phone will have very similar signatures. As a result, fast dithering may result in a poor quality communications link as explained below.

In GSM, the quality of the link is measured in terms of bit error rate (BER) and a frame erasure rate (FER) for various traffic and control channels. Fast dithering affects a quality of service (QoS) for a mobile phone in at least two ways. For one, the quality of audio carried in audio channel will deteriorate from a tone produced by periodic switching noise generated by a mobile phone's DC-to-DC converter having a switching signal undergoing fast dithering. Secondly, a call in progress via a mobile phone may be dropped due to a repeated corruption of a slow control channel word or byte due to EMI from switching noise generated by a mobile phone's DC-to-DC converter having a switching signal undergoing fast dithering.

An examination of information carried over a GSM network is helpful in evaluating other negative impacts of a fast dithering of a DC-to-DC converter's switching signal. FIG. 3 depicts a GSM Frame with Burst structures. There eight bursts in a GSM frame of a mobile phone call. One burst is used for reception and another burst is used for transmission. The remaining bursts may be used to by a mobile phone controller to monitor an adjacent cell's signal power along with other information in case a handover of a call is needed. A normal burst typically includes a traffic channel that can carry audio and/or other types of data including control information.

In a case involving audio data, the audio data is segmented into 20 ms segments and is usually encoded using a Regular-Pulse-Excited-Linear-Predictive-Coder (RPELCPC). Each of the 20 ms segments generates 260 coded bits, which are classified in three groups known as group 1a, group 1b and group II. Based on subjective testing, the three groups exhibit significant variations in sensitivity to errors. Errors within the first 50 coded bits in group 1a cannot be tolerated, whereas the coded bits of group II can tolerate some errors. However, the coded bits of group II are relatively tolerant to errors.

A low signal to noise ratio or poor link quality due to EMI from switching noise generated by a mobile phone's DC-to-DC converter having a switching signal undergoing fast dithering will introduce errors into the coded bits of groups 1a, 1b and II. Individual errors in the coded bits may be corrected by a decoding process, but blocks of errors in coded bits of groups 1a, 1b and II are relatively harder to correct. If any of the group 1a coded bits remain incorrect after the decoding process a frame erasure is likely to occur. As a result of multiple frame erasures, a higher than acceptable BER and/or FER will result in unacceptable poor audio quality, slow data rates, dropped calls and failures to respond to paging. Moreover, other mobile phones in their respective downlinks will be negatively affected by a mobile phone having a DC-to-DC converter with a switching signal undergoing fast dithering. Thus, it may be tempting to try a slow dither the switching signal of a DC-to-DC converter. However, experience has shown that whereas a fast dither of a switching signal results in an increase in BER relative to FER, a slow dither of a switching signal results in an increase in FER over BER. Thus, there remains a need to provide switching signal dithering circuit and method for a switching power supply such as a DC-to-DC converter for a mobile phone that will yield a low voltage ripple while reducing EMI and RF spurious transmissions. Moreover, there is a need for a switching signal dithering circuit and method that will result in less stringent filtering requirements, reduced cost, and increased efficiency for switching power supplies.

SUMMARY OF THE DISCLOSURE

In order to meet the various 3rd Generation Partnership Project (3Gpp) requirements that relates to spurious emissions and noise, the switching frequency of the DC-to-DC converter which gets modulated onto the RF carrier must be spread over a certain bandwidth that is actually much wider than a communications measurement's bandwidth. Modeling and measurements have revealed that a superior ripple spreading could be achieved if another dimension beyond a two dimensional fast dither of a switching signal for a switching power supply is added. The added dimension of the present disclosure takes advantage of measurement averaging whenever possible. Such an advantage is especially desirable in GSM standards testing. This added third dimension provides a slower changing aspect to the fast dithered spectrum. An ultimate result of the third dimension is a greater ripple suppression in receive (RX) mobile phone bands and other areas of concern where averaging is used for communications measurements.

A circuit of the present disclosure includes a switching circuitry adapted to transfer energy from a source to a load using a switching signal having a series of switching cycles and a switching frequency. Also included is a control circuitry that provides another dimension to the two dimensional fast dither by being adapted to generate a pseudo-random value near a beginning of each of the series of switching cycles to determine a maximum switching frequency value based upon the pseudo-random value. A disclosed method includes adjusting the switching frequency of the switching signal incrementally from a minimum switching frequency value to the maximum frequency value and vice versa as a function of time during each of the series of switching cycles of the switching circuit.

Those skilled in the art will appreciate the scope of the present disclosure and realize additional aspects thereof after reading the following detailed description of the preferred embodiments in association with the accompanying drawing figures.

BRIEF DESCRIPTION OF THE DRAWING FIGURES

The accompanying drawing figures incorporated in and forming a part of this specification illustrate several aspects of the disclosure, and together with the description serve to explain the principles of the disclosure.

FIG. 1 depicts an inductor current imbalance as a pulse width modulated (PWM) switching frequency is stepped from one period to another.

FIG. 2 depicts an adjusted inductor current resulting from an ideal PWM waveform.

FIG. 3 depicts a GSM Frame with Burst structures.

FIG. 4 depicts a switching power supply useable to transfer power from a source to a radio frequency (RF) power amplifier (PA), according to one embodiment of the disclosure.

FIGS. 5A and 5B illustrate the impacts of three-dimensional fast dithering.

FIG. 6 is a block diagram of a circuit that is useable to generate the five types switching frequency dithering supported by the present disclosure.

FIG. 7 is a block diagram of a sub-circuit that is useable to provide the dithering and frequency control function of the circuit shown in FIG. 6.

FIG. 8 is a logic truth table for the address decoder of the dithering and frequency control function of FIG. 7.

FIG. 9 is a block diagram of a Maximum Length Shift Register (MLSR) pseudo-random number generator having a variable length control.

FIG. 10 is a decoder truth table for the control for controlling the MLSR pseudo random number generator of FIG. 9.

FIG. 11 is a table of expected decimal values for various pseudo-random word selections for that may be generated by the MLSR of FIG. 9.

FIG. 12 is a graph of a fast frequency dithering according to the present disclosure.

FIG. 13 is a graph of a fast frequency dithering including a random span according to the present disclosure.

FIG. 14 is a graph of a slow frequency dithering according to the present disclosure.

FIG. 15 is a graph of a slow frequency dithering including a random span according to the present disclosure.

FIG. 16 is a graph of a three dimensional frequency that has a random maximum frequency according to the present disclosure.

FIG. 17A is a graph representing a first case for generating Fmax values in which a random number is added to a fixed Fmin value to create a random Fmax value.

FIG. 17B is a graph showing the resulting dithering frequencies for the Fmin and Fmax values of FIG. 17A.

FIG. 18A is a graph representing a second case in which a random number is added to and subtracted from an average frequency value Fmean to generate a random Fmax value and a random Fmin value, respectively.

FIG. 18B is a graph showing the resulting dithering frequencies for the Fmin and Fmax values of FIG. 18A.

FIG. 19A is a graph representing a third case in which two random numbers are generated and wherein one random number is useable to generate a random Fmax value and the other random number is useable to generate a random Fmin value.

FIG. 19B is a graph showing the resulting dithering frequencies for the Fmin and Fmax values of FIG. 19A.

FIG. 20 depicts an application example of a mobile terminal that includes an embodiment of a switching power supply according to the present disclosure.

DETAILED DESCRIPTION

The embodiments set forth below represent the necessary information to enable those skilled in the art to practice embodiments of the disclosure and illustrate the best mode of practicing the principles of the disclosure. Upon reading the following description in light of the accompanying drawing figures, those skilled in the art will understand the concepts of the disclosure and will recognize applications of these concepts not particularly addressed herein. It should be understood that these concepts and applications fall within the scope of the disclosure and the accompanying claims.

In order to meet the various 3Gpp requirements that relates to spurious emissions and noise, the switching frequency of the DC-to-DC converter which gets modulated onto the RF carrier must be spread over a certain bandwidth that is much wider than a communications measurement's bandwidth. Modeling and measurements have revealed that a superior ripple spreading could be achieved if another dimension was added to a fast dithering method. This added dimension would provide a slower changing aspect to the fast dithered spectrum. As a result, there is greater ripple suppression in the RX bands and other areas of concern where averaging is used for communications measurements.

Adding a third dimension to fast dithering will allow less stringent filtering requirements and thus simplify designs for switching power supplies for mobile phones, while at the same time reduce the cost, and increase efficiency of such designs. Moreover, such a design approach in accordance with the present disclosure is useable to design switching power supplies such as DC-to-DC converters that deliver DC power to single and multi-mode radio frequency (RF) power amplifiers (PAs). Moreover, such a design approach incorporating three-dimensional (3D) frequency dithering is also advantageous for applications other than powering RF PAs. For example, RF transceivers having a Voltage Controlled Oscillator (VCO) and Phase-Locked Loop (PLL) used to generate an RF carrier exhibit very high gain with extreme sensitivity to noise and power supply ripple. Thus, RF transceivers having VCOs and PLLs will benefit from the 3D frequency dithering provided by the present disclosure.

In accordance with the present disclosure, FIG. 4 shows a switching power supply 10 for transferring power to a radio frequency power amplifier (RF PA) 12. The switching power supply 10 receives a direct current (DC) supply signal V_DCSUPPLY and a set point signal V_RAMP, and provides a direct current signal V_DCPA to a power supply input to the RF PA 12. The switching power supply 10 regulates direct current signal V_DCPA based on the set point signal V_RAMP. The RF PA 12 receives and amplifies an RF input signal RF_IN to provide an RF output signal RF_OUT using the direct current signal V_DCPA.

A two dimensional fast dithering method for generating a pulse width modulation (PWM) signal that provides for a switching frequency adjustment at every cycle is disclosed in U.S. patent application Ser. No. 11/756,909, now U.S. Pat. No. 7,928,712. The present disclosure improves the fast dithering method, whereby a minimum frequency and a maximum frequency used for the lower and upper limits of the fast dithering are changed at a slower rate than an averaging rate of a communication test measurement. The slower rate can either be changed continuously or on a per burst basis. The slower rate actually takes advantage of the fact that, for GSM, the measurement setup averages between 50 and 200 bursts (FIG. 3) depending on a particular GSM test being performed. As long as the slow dithering rate is actually slower than the averaging rate of the GSM test measurement, a significant ripple suppression gain can be added to the fast dithering gain.

Implementing fast dithering based on randomly selecting a frequency within a range of uniformly distributed frequencies would also get a noise-like spreading gain. However, since an inductor charge imbalance is proportional to a step in frequency, an average step would have a mean close to the maximum number of frequencies/2, and thus will have a larger mean ripple. Furthermore, this approach would not benefit as much from the averaging process which can add an extra 10 dB to 15 dB or so of rejection.

Fast dithering is a two-dimensional process which spreads the switching frequency spurs over a wider range, which results in some ripple rejection. However, every burst (FIG. 3) in a GSM frame will look very similar to each other. If every burst had a different signature, a resulting average would provide more ripple rejection above and beyond the gain of using fast dithering alone.

The fast dithering signature of the bursts of a GSM frame can be changed by providing upper frequency limits and/or different lower frequency limits from time to time or for every burst. Randomly changing the upper and lower frequency limits of a switching signal is one aspect of three-dimensional fast dithering of a switching signal for a switching power supply.

FIGS. 5A and 5B illustrate the impacts of the three-dimensional fast dithering of a switching power supply's switching signal. FIG. 5A shows a graph of fifty GSM bursts using two-dimensional fast dithering with a two stage output filter. An overlay of a European Telecommunications Standards Institute (ETSI) spectral mask in a Low GSM band shows a marginal passing of the ETSI standards at best. FIG. 5B shows a graph of a linear average of fifty GSM bursts using three-dimensional fast dithering in which a maximum switching frequency limit and/or a minimum switching frequency limit is randomly changed. As a result of changing the maximum and/or minimum switching frequency limits while using fast frequency dithering, an averaging effect inherent in the ETSI measurement yields a conclusive passing of the ETSI test standard.

Five types of switching frequency dithering are supported by the present disclosure. Four of the switching frequency dithering types are two-dimensional and a fifth is three-dimensional. Two of the two dimensional dithering types are slow and two others are fast. A stepping sequence from a minimum switching frequency limit to a maximum switching frequency limit can either follow an incremental relationship, or be purely random or preferably be pseudo-random. An incremental stepping sequence is preferably linear.

FIG. 6 depicts a block diagram of a circuit 14 that is an embodiment of the present disclosure that is useable to generate the five types switching frequency dithering supported by the present disclosure. The circuit 14 includes a frequency generator 16 and a dithering and frequency control function 18 that are useable to set a desired frequency dithering configuration for a switching signal of the switching power supply 10 (FIG. 4). The frequency generator 16 includes a variable slope ramp generator 20 for generating a triangular waveform having an increasing voltage section and a decreasing voltage section. The variable slope ramp generator 20 comprises a discharging digital-to-analog converter (DAC) 22 and a charging DAC 24 that are useable to discharge and charge an on-chip capacitor 26. The discharging DAC 22 is always active and provides a negative sloped voltage ramp for generating the decreasing voltage section of the triangular waveform. The charging DAC 24 is enabled to provide a positive sloped voltage ramp for generating the increasing voltage section of the triangular waveform by supplying more current than the discharging DAC 22. Both, the discharging DAC 22 and the charging DAC 24 are each programmable to deliver variable amounts of current. In a preferred embodiment the charging DAC 24 is programmed to supply twice as much current as the discharging DAC 22. The triangular waveform is available at an output node 28 during operation of the variable slope ramp generator 20.

A window comparator 30 includes a first voltage comparator 32 having an input 34, a threshold voltage input 36 and an output 38. The window comparator 30 also includes a second voltage comparator 40 having an input 42, a threshold voltage input 44 and an output 46. The input 34 of the first voltage comparator 32 and the input 42 of the second voltage comparator 40 are coupled to the output node 28 of the variable slope ramp generator 20. The output 38 of the first voltage comparator 32 and the output 46 of the second voltage comparator 40 are coupled to a logic circuit 48.

During operation of the variable slope ramp generator 20, the first voltage comparator 32 compares the triangular waveform at the input 34 to an adjustable reference voltage (VREF_HIGH) provided to the threshold voltage input 36. A responsive output signal is generated by the first voltage comparator 32 and presented at the output 38. In similar fashion, the second voltage comparator 40 compares the triangular waveform at the input 42 to an adjustable reference voltage (VREF_LOW) provided at the threshold voltage input 44. A responsive output signal is generated by the second voltage comparator 40 and presented at the output 46. The logic circuit 48 generates an internal clock signal 50 in response to the output signals provided by the first voltage comparator 32 and the second voltage comparator 40.

The adjustable reference voltages VREF_HIGH and VREF_LOW are useable to provide a selectable propagation delay compensation for the first voltage comparator 32 and the second voltage comparator 40, respectively. If no propagation delay compensation is selected, the VREF_HIGH and VREF_LOW thresholds are not adjusted with a change in frequency for a switching signal. However, when propagation delay compensation is selected, the VREF_HIGH and VREF_LOW thresholds are adjusted as the frequency of the switching frequency is changed.

It is preferred that a propagation delay compensation be provided to maintain linearity between generated switching frequencies and a set of generated pseudo-random codes. In this way, the frequency generation for the set of generated random codes can be accurately predicted. The propagation delay compensation also makes predicting a spur spreading gain more accurate. At least one reason the propagation delay compensation is preferred is that as frequency increases a finite comparator propagation becomes a more important portion of the frequency's total period. As a result, the frequency is reduced by an equal amount as the frequency increases.

A propagation delay compensation circuitry 52 that is useable to adjust the VREF_HIGH threshold comprises a current DAC 54 that is coupled to a multi-tap resistor 56. An adjustable current controlled by the current DAC 54 flows through the multi-tap resistor 56 to generate the adjustable VREF_HIGH threshold. An analog multiplexer (MUX) 58 receives inputs from taps 60 of the multi-tap resistor 56. The MUX 58 has an output 62 that is coupled to the voltage threshold input 36 of the first voltage comparator 32. A control signal PROPDELAYHIGH[2:] is useable to select Individual ones of taps 60 for output to the voltage threshold input 36 of the first voltage comparator 32.

Another propagation delay compensation circuitry 64 that is useable to adjust the VREF_LOW threshold comprises a current DAC 66 that is coupled to a multi-tap resistor 68. An adjustable current controlled by the current DAC 66 flows through the multi-tap resistor 68 to generate the adjustable VREF_LOW threshold. An analog multiplexer (MUX) 70 receives inputs from taps 72 of the multi-tap resistor 68. The MUX 70 has an output 74 that is coupled to the voltage threshold input 44 of the second voltage comparator 40. A control signal PROPDELAYLOW[2:] is useable to select Individual ones of taps 72 for output to the voltage threshold input 44 of the second voltage comparator 40. The independent control PROPDELAYLOW[2:] is provided for propagation compensation circuitry 64 because the second voltage comparator 40 operates closer to ground potential 76 and will behave differently than the first voltage comparator 32. An adjustable voltage reference 74 having a voltage reference input (VREF_—2.4) and a control input (FREQCAL[3:0]) is useable to supply the current DACs 22, 24, 54 and 66 with an accurate voltage reference (VDAC).

FIG. 7 shows details of the dithering and frequency control function 18 depicted in FIG. 6. The frequency control function includes a 3-bit counter 76 that generates a SLOWCLOCK signal that is input for a digital multiplexer (DMUX) 78. A DMUX 80 receives an enable signal (MAINCP ENABLE) and an event signal (SDIWRITE EVENT). The SDIWRITE EVENT signal is generated upon a Serial Device Interface (SDI) write event that is useable to preset values of registers and counters such as the 3-bit counter 76. Either the enable signal or the event signal is selected as output for input into the DMUX 78 via a control signal (CTRL_N). The DMUX 78 is controlled by the CTRL_N signal, which is useable to select either the output from DMUX 80 or the SLOWCLOCK signal for output. A 6-bit counter 82 includes a clock input that is fed by a either the SLOWCLOCK signal or a SWITCHING CLOCK signal that is selectable through a DMUX 84. A 6-bit output from the 6-bit counter 82 is input for a DMUX 86. A Maximum Length Shift Register (MLSR) 88 that is useable to generate pseudo-random numbers accepts input from a DMUX 90. The CTRL_N signal is received by the DMUX 90 to select between the SLOWCLOCK signal, the SWITCHING CLOCK signal, the MAINCP ENABLE signal and the SDIWRITE EVENT signal. A 6-bit adder 92 receives random numbers generated by the MLSR 88. The 6-bit adder 92 adds the random numbers to Fmin values inputted into the 6-bit adder via an Fmin[5:0] signal. Individual summations of the random numbers and Fmin values are passed to the frequency generator 16 (FIG. 6) as a 6-bit word via the DMUX 86.

A decoder logic circuit 94 receives and decodes control bits sent via a DITHERINGCTRL[2:0] signal. The decoded control bits are useable to control or enable the 3-bit counter 76, the 6-bit counter 82, and the MLSR 88. FIG. 8 is a truth table that relates the various dithering modes of slow, fast, linear, and random to the dithering control settings carried by the DITHERINGCTRL[2:0] signal.

FIG. 9 shows the details of the MLSR 88 of FIG. 7. The MLSR 88 is a digital circuit that periodically generates a set of permutations of binary numbers for a given selection of taps 96A-F for capturing binary levels from a feedback path 98. The binary numbers generated follow a uniform distribution with the exception of an all bits zero condition that prevents a sequence needed for binary number generation. The length of the MLSR 88 is adjustable to match the RandomSpan[1:0] selection of 3, 4, 5, or 6 bits. Individual coded selections of taps 96A-F are [2, 1] for a 3-bit length, [3, 2] for a 4-bit length, [4, 2] for a 5-bit length, and [5, 4] for a 6-bit length.

A D flip-flop 100 accepts input from the feedback path 98 and a clock signal 102. An output of the D flip-flop 100 is a most significant bit, and in this case is bit 5. The bit 5 output of D flip-flop 100 is fed into an exclusive or (XOR) gate 104, which also receives input from the tap 96A. A digital multiplexer (DMUX) 106 receives output from the XOR gate 104 as well as the tap 96B. An MLSRCTRL4 signal controls which of the signals received by DMUX 106 is outputted to a D flip-flop 108. An output of the D flip-flop 108 is of a bit 4 magnitude. The bit 4 output of the D flip-flop 108 is fed into a DMUX 110 that also receives input from the tap 96C. An MLSRCTRL3 signal controls which of the signals received by DMUX 110 is outputted to a D flip-flop 112. An output of the D flip-flop 112 is of a bit 3 magnitude. The bit 3 output of the D flip-flop 112 is fed into an XOR gate 114 that also receives input from tap 96E. An output of the XOR gate 114 is fed into a DMUX 116 that also receives input from the tap 96D and the output from the D flip-flop 112. An MLSRCTRL[2:1] signal controls which of the signals received by DMUX 116 is outputted to a D flip-flop 118. An output of the D flip-flop 118 is of a bit 2 magnitude. The bit 2 output of D flip-flop 118 is fed into an XOR gate 120 that also receives input from the tap 96F. An output from the XOR gate 120 is fed into a DMUX 122 that also receives the bit 2 output of D flip-flop 118. An MLSRCTRL0 signal controls which of the signals received by DMUX 122 is outputted to a D flip-flop 122. An output of the D flip-flop 122 is of the bit 1 magnitude. The bit 1 output of D flip-flop 122 is fed into a D flip-flop 126 that has an output of the bit 0 magnitude. The bit 0 output of D flip-flop 126 is outputted to the feedback path 98. A DMUX 128 receives the bit outputs of the D flip-flops 100, 108, 112, 118, 124 and 126. The DMUX 128 is controlled by the RANDOMSPAN[1:0] signal to select which of the bit outputs received by the DMUX 128 will be outputted as random switching frequency control bits. The clock signal 102 going to the D flip-flops 100, 108, 112, 118, 124 and 126 can be disabled to reduce power consumption.

FIG. 10 is a decoder truth table for a conversion of the RANDOMSPAN signal to the MLSRCTRL signal. With a step size of 125 kHz, a 3-bit span covers 875 kHz, a 4-bit span covers 1.875 MHz, a 5-bit span covers 3.857 MHz, and a 6-bit span covers 7.875 MHz. The dithering and frequency control function 18 (FIGS. 6 and 7) is configured to not allow the Fmin value and RandomSpan settings to exceed a binary maximum of 111111b. The reason for this is that the MLSR 88 (FIG. 7) is not allowed to go to all zeros. Therefore, minimum allowable switching frequency will be higher by 125 kHz compared to a disallowed 0b RandomSpan setting.

In order to ensure that the MLSR 88 will always start a sequence at a given point, a seed value is loaded into the MLSR 88. The seed value is preferably 1b and is loaded at power up when a power on reset (POR) for the dithering and frequency function 18 (FIG. 6 and FIG. 7) is triggered.

The two equations that follow can be used to predict the binary values, based on the RandomSpan[1:0] selection, and their decimal equivalent. The actual frequency can then be calculated by summing up the selected Fmin value and generated random number value and multiply the sum by 125 kHz and add 2 MHz. The first equation builds an array of binary values sized by a “Number of Trials” value. A set of “Seed_nBit” values are defined in binary as 001b for the 3 Bit, 0001b for the 4 Bit, etc.

FIG. 11 is a table that lists the expected decimal values generated by the MLSR 88. The pseudo-randomness of the expected values is provided by the MLSR 88. A uniform pseudo-randomness is generated for all selectable sequence lengths. Implementations of the various dithering modes are described below.

Linear Fast Dithering

Referring to FIG. 12, a switching frequency of a switching power supply such as switching power supply 10 (FIG. 4) is changed every switching cycle between the Fmin value and Fmax value. A possible implementation of a method to provide fast dithering may be based on a pre-settable 6-bit up/down counter such as the 6-bit counter 82 (FIG. 7). In such a case, the value of the Fmin[5:0] signal is used as a preset frequency value for a lower frequency limit, and the value Fmax[5:0] signal is used as a preset frequency value for an upper frequency limit. The 6-bit counter 82 will count up or down incrementally between the Fmin and Fmax values. As a result of a repetitive cycle incrementing and decrementing between the Fmin and Fmax values, a plurality of discrete frequency tones will be generated. The spectral spacing between each of the plurality of tones is equal to the dithering frequency. The plurality of discrete frequency tones will extend all to way to zero frequency (i.e., DC), so the amplitude of each tone's fundamental frequency along with any significant harmonics is an important design consideration. The resulting dithering frequency is calculated as follows:

F Dithering = 1 1 F ⁢ ⁢ min + 1 F ⁢ ⁢ max + 2 ⁢ ∑ n = 1 N - 1 ⁢ 1 F ⁢ ⁢ min + n ⁢ ⁢ F steps ,
where N is the number of steps between Fmin and Fmax ((Fmax−Fmin)/F_steps), and F_steps is a step size such as 125 kHz.
Random Fast Dithering

Referring to FIG. 13, the frequency of the switching power supply 10 (FIG. 4) is again changed every switching cycle, but in a random fashion. In this case the 6-bit up/down counter is replaced by the maximum length shift register (MLSR) 88 (FIGS. 6 and 7) and the 6-bit full adder 92. At near the beginning of every switching cycle of the switching frequency, a new random number will be generated by the MLSR 88, which is controlled by the bits of the RandomSpan[1:0] signal. The new random number is then added to the Fmin value, and the sum is passed to the frequency generator 16 (FIG. 6) as a 6-bit word. Linear Slow Dithering

Referring to FIGS. 6 and 14, the frequency of the switching power supply 10 (FIG. 4) is changed when a trigger event occurs. One trigger event occurs on a rising edge of the MainCpEnable bit. Another trigger event occurs on the completion of a Serial Device Interface (SDI) write cycle. Either of these trigger events can be selected with the bit SlowTrigSelect[1:0]. A possible implementation can a based on a pre-settable 6-bit up/down counter, where Fmin is used as preset value, and Fmax is used as the upper limit. The counter will count up or down incrementally between Fmin and Fmax. Furthermore, in case the time lapse between successive events is too short, a trigger event counter allows that time to be extended. The number of trigger events to be counted depends on the TriggerCount[2:0] bits which allow 1, 2, 4, 8, 16, 32, 64, or 128 events to be counted. The dithering rate, if it exists, can also be calculated as in the Linear Fast Dithering by taking into account the rate of occurrence of the selected trigger event and the number of it that are being counted as per the TriggerCount[2:0] bits. In practice, the dithering rate will most likely be at very low frequency that is significantly below the audible range of humans and will only exist if the dithering rate is changed repetitively within one burst (FIG. 3).

Random Slow Dithering

Referring to FIG. 15, the frequency of the switching power supply 10 (FIG. 4) is again changed every so often when the so-called trigger events occur, but in a random fashion. In this case, the operation of the 6-bit counter 82 (FIG. 7) is replaced by an operation of the MLSR 88 (FIG. 7) and the 6-bit full adder 92 (FIG. 7). Every switching cycle, a new random number will be generated by the MLSR 88 as controlled by bits sent via the RandomSpan[1:0] signal. The generated random number is then added to the Fmin value. The sum is passed to the frequency generator 16 as a 6-bit word. There are two events being supported at this time and they are; a rising edge of the MainCpEnable bit, and the completion of an SDI write cycle. Either one can be selected with a bit use a SlowTrigSelect[1:0] signal. Furthermore, in case the time lapse between successive events is too short, a trigger event counter allows that time to be extended. The number of trigger events to be counted depends on the TriggerCount[2:0] bits which allows 1, 2, 4, 8, 16, 32, 64, or 128 events to be counted (only valid for the SDI write events). In practice, the dithering rate will most likely be at very low frequency that is significantly below the human audible range and will only exist if it is changed repetitively within each burst.

3D Dithering

Referring to FIG. 16, the frequency of the switching power supply 10 (FIG. 4) is again changed every switching similar to fast frequency dithering, but in this case the maximum frequency value Fmax is changed randomly while the minimum frequency value Fmin is fixed. However, the Fmin value may be changed via the SDI. In operation, the 6-bit counter 82 (FIG. 7) begins incrementing from the Fmin value and continues for every switching cycle until the first Fmin+random value is reached, after which the 6-bit counter 82 decrements towards the Fmin value. Upon reaching the Fmin value, the MLSR 88 is incremented once, and the new random number is again added to the Fmin value. The 6-bit counter 82 counts up incrementally until it reaches the Fmax value, after which the 6-bit counter 82 decrements towards the Fmin value. The only bits used to control three-dimensional frequency dithering bits sent by the Fmin[5:0] and RandomSpan[1:0] signals. In this case of three-dimensional frequency dithering, the pseudo-randomness of a switching frequency signal that gets modulated into a radio frequency signal cannot be easily captured or measured via a radio frequency spectrum analyzer. As a result, three-dimensional frequency dithering of a switching signal of a switching power supply is the preferred mode of operation of a mobile terminal during normal use. The other frequency dithering modes discussed above are primarily useful for testing the operation of switching power supply 10 (FIG. 4) and circuit 14 (FIG. 6).

FIGS. 17A and 17B illustrate that the randomness of numbers generated by the MLSR 88 (FIG. 9) does translate into absolute and dithering frequency randomness. In a first case, a random number generated by the MLSR is added to an Fmin value to create Fmax. In this first case the Fmin value is fixed while the Fmax value is randomly changed. As a result, the switching frequency linearly changes back and forth between the Fmin value and the Fmax value. Thereby, generating a set of pseudo-random dithering frequencies. The Fmin value was chosen to be 2.0 MHz, and a 5-bit setting for the MLSR 88 that was used to generate the random numbers.

FIGS. 18A and 18B illustrate a second case in which a random number generated by MLSR 88 is added to an average Fmean value to generate a random Fmax value and subtracted from the Fmean value to generate a random Fmin value. In this second case, both the Fmin value and the Fmax value are randomly changed. Then the switching frequency linearly changes back and forth between the Fmin value and the Fmax value, thereby generating a set of dithering frequencies. The Fmean value was chosen to be 6.0 MHz, and the 5-bit setting for the MLSR 88 was used for generating the random numbers.

FIGS. 19A and 19B illustrate a third case in which two random number generators, identical to the MLSR 88 (FIG. 9), are used. Both random number generators have the same properties and taps, but the seed values for the random number generators are different. For example, a seed value of 10000b may be used for one random number generator, while a seed value of 00001 b may be used for the other random number generator. In this way, one set of random numbers is added to an average Fmean value to generate a random Fmax value. Another set of random numbers is subtracted from the Fmean value to generate a random Fmin value. In this case, both the Fmin value and the Fmax value are randomly changed. As a result, the switching frequency of the switching power supply 10 (FIG. 4) linearly changes back and forth between the Fmin value and the Fmax value. Thereby, generating a random set of dithering frequencies. The Fmean value was chosen to be 6.0 MHz, and the 5-bit setting was used for the random number generators.

As illustrated by FIGS. 17B, 18B and 19B, the variation of randomness between the three cases does not justify additional hardware and complications that may be required to support cases 2 and 3.

The implementation that supports case 1 will produce similar spurious spreading gain as the other cases and is therefore the preferred implementation. However, any implementation may be used without deviating from the teachings of the present disclosure.

FIG. 20 depicts an application example of the switching power supply 10 (FIG. 4) feeding power to a power amplifier circuitry 130 that comprises a part of the basic architecture of a mobile terminal such as a wireless smart phone 132. The wireless smart phone 132 may include a receiver front end 134, a radio frequency transmitter section 136, an antenna 138, a duplexer or switch 140, a baseband processor 142, a control system 144, a frequency synthesizer 146, and a user interface 148. The receiver front end 134 receives information bearing radio frequency signals from one or more remote transmitters provided by a base station (not shown). A low noise amplifier 150 amplifies the signal. A filter circuit 152 minimizes broadband interference in the received signal, while downconversion and digitization circuitry 154 downconverts the filtered, received signal to an intermediate or baseband frequency signal, which is then digitized into one or more digital streams. The receiver front end 134 typically uses one or more mixing frequencies generated by the frequency synthesizer 146. The baseband processor 142 processes the digitized received signal to extract the information or data bits conveyed in the received signal. This processing typically comprises demodulation, decoding, and error correction operations. As such, the baseband processor 142 is generally implemented in one or more digital signal processors (DSPs).

On the transmit side, the baseband processor 142 receives digitized data, which may represent voice, data, or control information, from the control system 144, which it encodes for transmission. The encoded data is output to the radio frequency transmitter section 136, where it is used by a modulator 156 to modulate a carrier signal that is at a desired transmit frequency. The power amplifier circuitry 130 amplifies the modulated carrier signal to a level appropriate for transmission, and delivers the amplified and modulated carrier signal to the antenna 138 through the duplexer or switch 140.

A user may interact with the wireless smart phone 132 via the user interface 148, which may include interface circuitry 160 associated with a microphone 162, a speaker 164, a physical or virtual keypad 166, and a touch screen display 168. The interface circuitry 160 typically includes analog-to-digital converters, digital-to-analog converters, amplifiers, and the like. Additionally, the interface circuitry 160 may include a voice encoder/decoder, in which case the interface circuitry 160 may communicate directly with the baseband processor 142.

The microphone 162 will typically convert audio input, such as the user's voice, into an electrical signal, which is then digitized and passed directly or indirectly to the baseband processor 142. Audio information encoded in the received signal is recovered by the baseband processor 142 and converted by the interface circuitry 160 into an analog signal suitable for driving the speaker 164. The keypad 166 and the touch screen display 168 enable the user to interact with the wireless smart phone 132, input numbers to be dialed, address book information, or the like, as well as monitor call progress information.

The control system 144 includes a memory 170 for storing data and software applications 172, and a processor 174 for running the operating system and executing the software applications 172.

Those skilled in the art will recognize improvements and modifications to the present disclosure. All such improvements and modifications are considered within the scope of the concepts disclosed herein and the claims that follow.