Digital fast CORDIC for envelope tracking generation

Abstract

Disclosed is a coordinate rotation digital computer (CORDIC) having a maximum value circuit that selects a larger of the first component or the second component. A minimum value circuit selects a minimum operand that is a smaller one of the first component or the second component. Also included are N rotator stages, each corresponding to a unique one of N predetermined vectors, each of the N rotator stages having a first multiply circuit to multiply the maximum operand by a cosine coefficient of a predetermined vector to output a first rotation component, a second multiply circuit for multiplying the minimum operand by a sine coefficient of the predetermined vector to output a second rotation component, and an adder circuit for adding the first rotation component to the second rotation component to output one of N results, and a maximum value circuit for outputting a maximum one of the N results.

Description

FIELD OF THE DISCLOSURE

The present disclosure relates to a method and system that calculates the norms of a stream of vectors at a relatively high rate.

BACKGROUND

An envelope tracking system generates an envelope signal that is used as a reference input for a fast switched-mode power supply (Fast SMPS). In turn, the Fast SMPS uses the envelope signal to modulate a supply of a power amplifier for an increased efficiency. At present, an envelope signal generated by traditional methods is not fast or accurate enough for use with the long term evolution (LTE) standard wherein an envelope modulation bandwidth can be as high as 1.5 times a modulation bandwidth. In fact, a 20 MHz LTE bandwidth requires about 30 MHz envelope bandwidth, which further requires a digital sampling clock of 104 MHz for improved oversampling. As a result, there is a need for a method and system that generates fast digital envelope signals using in-phase (I) and quadrature (Q) signals that drive an RF modulator.

SUMMARY

The present disclosure provides a coordinate rotation digital computer (CORDIC) for computing the norm of a vector having a first component and a second component. The CORDIC includes a first maximum value circuit for outputting a maximum operand that is a larger one of either an absolute value of the first component or an absolute value of the second component. A minimum value circuit outputs a minimum operand that is a smaller one of either an absolute value of the first component or an absolute value of the second component. The CORDIC has N rotator stages, each corresponding to a unique one of N predetermined vectors. Each of the N rotator stages has a first multiply circuit for multiplying the maximum operand by a cosine coefficient of a corresponding predetermined vector to output a first rotation component and a second multiply circuit for multiplying the minimum operand by a sine coefficient of the corresponding predetermined vector to output a second rotation component. Moreover, each of the N rotator stages include an adder circuit for adding the first rotation component to the second rotation component to output one of N results. The CORDIC also has a second maximum value circuit for outputting a maximum one of the N results.

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

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 is a vector graph that is useful in illustrating the concepts of the present disclosure.

FIG. 2 is a block diagram of a first embodiment of a coordinate rotation digital computer (CORDIC) according to the present disclosure.

FIG. 3 is a block diagram of a second embodiment of the CORDIC according to the present disclosure.

FIG. 4 is a voltage versus time graph of an envelope tracking signal that can be generated using output from the present CORDIC.

FIG. 5 is a block diagram of a digital envelope tracking system that incorporates the present CORDIC.

DETAILED DESCRIPTION

The embodiments set forth below represent the necessary information to enable those skilled in the art to practice the embodiments and illustrate the best mode of practicing the embodiments. 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.

FIG. 1 is a vector graph that is useful in illustrating the concepts of the present disclosure. The vector graph includes a vector IQ having a first component I and a second component Q. A goal of the method and system of the present disclosure is to find the norm of a vector such as the vector IQ using calculations fast enough to use the norm for radio frequency (RF) modulation. Preferably, the calculations provided by the present disclosure processes vectors at a rate on the order of millions of vectors per second. In order to achieve this relatively high rate of vector processing, tradition floating point calculations used by functions based upon the Pythagorean theorem are too inefficient. Instead, the present disclosure provides a system and method that uses a novel coordinate rotation digital computer (CORDIC) for computing the norm of a vector.

In particular, the system and method of the present disclosure provides an array of unit vectors Ū(i) that are used to effectively rotate the vector IQ significantly onto a quadrant axis. In FIG. 1, quadrant axes are labeled X and Y. Each of the unit vectors Ū(i) has a first component a(i) and a second component b(i). An exemplary list of numerical values for the unit vectors Ū(i), the first components a(i), and the second components b(i) are listed in TABLE 1 below.

TABLE 1
iterator i	(i)	a(i)	b(i)
0	(0) = 1∠4.5°	0.9969	0.0785
1	(1) = 1∠13.5°	0.9724	0.2334
2	(2) = 1∠22.5°	0.9239	0.3827
3	(3) = 1∠31.5°	0.8526	0.5225
4	(4) = 1∠40.5°	0.7604	0.6494

As will be explained in greater detail further in the disclosure, the vector IQ is multiplied by each of the unit vectors Ū(i) in order to calculate the norm for the vector IQ. In an exemplary case shown in FIG. 1, the vector IQ has a first component I and a second component Q. Approximations of the norm are given by the following equation (1) with the assumption that the first component I is larger than the second component Q:norm≈(a(i)(|I|)+(b(i)(|Q|)  (1)

The absolute values of the first component I and a second component Q are taken to ensure that the vector IQ is or becomes a first quadrant vector. Converting a vector to a first quadrant vector does not change the norm of the vector. Moreover, whenever the vector IQ has a second component Q that is larger than the first component I, the first component I and the second component Q are swapped in equation 1. In this way, the vector IQ is restricted to locations between angles ranging from 0° to 45° (i.e., 0 to π/4 radians).

A multiplication by one of the unit vectors Ū(i) yields a largest value. This largest value is the most accurate calculation of the norm for the vector IQ for the unit vectors U(i). The one of the unit vectors Ū(i) that yields the most accurate norm calculation will have the smallest angle difference with the vector IQ. An angle Θ represents the smallest angle between one of the unit vectors Ū(i) and the vector IQ.

In the exemplary case shown in FIG. 1, the vector IQ has a first component I having a value of 3599 and a second component Q having a value of 1822. Using equation (1), approximations of the norm of vector IQ for the exemplary case are shown in Table 2 below.

TABLE 2
iterator i	(i)	norm ≈ (a(i)(|I|) + (b(i)(|Q|)
0	(0) = 1∠4.5°	3731
1	(1) = 1∠13.5°	3925
2	(2) = 1∠22.5°	4022
3	(3) = 1∠31.5°	4021
4	(4) = 1∠40.5°	3920

In this exemplary case, the unit vector Ū(2) has the smallest angle with the vector IQ. As can be seen in Table 2, the largest calculated norm is 4022, which was calculated using the unit vector Ū(2). Referring to FIG. 1 the norm calculation of 4022 is reasonable because the unit vector Ū(2) is slightly closer to the vector IQ than the unit vector Ū(3). Using the Pythagorean Theorem, the actual norm for the vector IQ is close to 4034. Accuracy can be improved by increasing the number of unit vectors Ū(i) in the array.

FIG. 2 is a block diagram of a first embodiment of a coordinate rotation digital computer (CORDIC) 10 that is in accordance with the present disclosure. The CORDIC 10 is made up of circuitry that digitally calculates the norms of a stream of vectors such as the vector IQ (FIG. 1). In particular, the CORDIC 10 includes an in-phase (I) input 12 and a quadrature (Q) input 14 that both except signed values that are both n bits wide. The I input 12 accepts the first component I of the vector IQ, while the Q input 14 accepts the second component Q of the vector IQ.

A first absolute value (ABS) circuit 16 outputs the absolute value of the first component I. A second ABS circuit 18 outputs the absolute value of the second component Q. In the particular case of the first embodiment CORDIC 10, the outputted values are d bits wide, wherein d=n−1.

A first maximum value (MAX) circuit 20 receives both the first component I and the second component Q and in turn outputs the larger value of the first component I and the second component Q. In contrast, a first minimum value (MIN) circuit 22 receives both the first component I and the second component Q and in turn outputs the smaller value of the first component I and the second component Q.

A first rotator stage 24, a second rotator stage 26 and an Nth rotator stage 28 simultaneously receive the larger value output from the first MAX circuit 20 and the smaller value output from the MIN circuit 22. One approximation of a norm for the vector IQ is calculated and output from the first rotator stage 24 based upon the larger value output from the first MAX circuit 20 and the smaller value output from the MIN circuit 22. Simultaneously, another approximation of the norm for the vector IQ is calculated and output from the second rotator stage 26 based upon the same larger value output from the first MAX circuit 20 and the same smaller value output from the MIN circuit 22. A last approximation of the norm for the vector IQ is simultaneously calculated and the output from the Nth rotator stage 28 is based upon the same larger value output from the first MAX circuit 20 and the same smaller value output from the MIN circuit 22. In this way, N approximations of the norm for the vector IQ are calculated by the CORDIC 10. A second MAX circuit 30 having an output 32 simultaneously receives the N approximations of the norm for the vector IQ and outputs the largest one of the N approximations of the norm for the vector IQ. This largest one of the N approximations of the norm for the vector IQ is the closest approximation for the norm for the vector IQ. A larger number of N rotator stages will yield a more accurate approximation for the norm for the vector IQ at the cost of additional circuitry and additional power needed to operate the additional circuitry.

The first rotator stage 24, the second rotator stage 26, and the Nth rotator stage 28 each have a multiply circuit 34 for multiplying the larger value, which is a maximum operand, by corresponding cosine coefficients a(i), wherein i is a natural number that ranges from 0 to N−1. For example, the cosine coefficient a(0) is stored in a memory location 36 for use by the first rotator stage 24. The cosine coefficient c is stored in a memory location 38 for use by the second rotator stage 26, and a cosine coefficient a(N−1) is stored in a memory location 40 for use by the Nth rotator stage 28. The memory location 36, the memory location 38, and the memory location 40 are each c bits wide. An equation for calculating the cosine coefficients is given by the following equation (2).

$\begin{array}{cc}a\left(i\right)=2^c·\mathrm{COS}\left(\frac{\pi }{4N}i+\frac{\pi }{8N}\right)·\mathrm{CorrectionFactor}& \left(2\right)\end{array}$ The CorrectionFactor is a dimensionless correction factor that is usable for reducing the mean error over a plurality of calculations and is given by the following equation (3).

$\begin{array}{cc}\mathrm{CorrectionFactor}=\frac{1}{\frac{8N}{\pi }·\mathrm{SIN}\frac{\pi }{8N}}& \left(3\right)\end{array}$

Similarly, the first rotator stage 24, the second rotator stage 26, and the Nth rotator stage 28 each have a multiply circuit 42 for multiplying the smaller value, which is a minimum operand, by corresponding cosine coefficients b(i), wherein i is a natural number that ranges from 0 to N−1. A sine coefficient b(0) is stored in a memory location 44 for use by the first rotator stage 24. A sine coefficient b(1) is stored in a memory location 46 for use by the second rotator stage 26, and a sine coefficient b(N−1) is stored in a memory location 48 for use by the Nth rotator stage 28. The memory location 44, the memory location 46, and the memory location 48 are each c bits wide. An equation for calculating the sine coefficients is given by the following equation (4).

$\begin{array}{cc}b\left(i\right)=2^c·\mathrm{SIN}\left(\frac{\pi }{4N}i+\frac{\pi }{8N}\right)·\mathrm{CorrectionFactor}& \left(4\right)\end{array}$ The CorrectionFactor is usable for reducing the mean error over a plurality of calculations and is given by the equation (3) above.

The first rotator stage 24, the second rotator stage 26, and the Nth rotator stage 28 each have an adder circuit 50 that receive m bit wide outputs from corresponding multiply circuits 34 and 42. Each adder circuit 50 outputs an o bit wide norm approximation to the second MAX circuit 30. Moreover, each adder circuit includes a results register 52 that is m bits wide plus an overflow bit 54 that controls a results multiplexer 56 that outputs either the addition results or a predetermined maximum value (MAXVAL) held in a memory location 58. In this particular embodiment the output of the multiplexer 56 will be addition results if the overflow bit is set to 0. In contrast, the output of the adder circuit will be the constant MAXVAL if the overflow bit is set to 1. A round and right shift (RS) circuit 60 rounds the addition results and then right shifts the results until only an o bit wide result remains.

Table 3 below lists word size configurations for an exemplary CORDIC having five rotator stages and another exemplary CORDIC having six rotator stages. Table 3 also lists the maximum error for a least significant byte (LSB) and the mean error LSB. The probability for an error greater than or equal to 0.5 LSB is 11.3% for a CORDIC with five rotator stages and 7.4% for a CORDIC having 7.4%.

TABLE 3
N Ro-		Coef-	Oper-	Multi-		Max.	MEAN
tator	Input	ficient	and	ply	Output	Error	Error
Stages	Bits n	Bits c	Bits d	Bits m	Bits o	(LSB)	(LSB)
5	13	9	12	11	8	1.09	0.046
6	13	10	12	13	8	0.90	0.010

Table 4 lists digital values for the coefficients of the first components a(i), and the second components b(i) of the unit vectors Ū(i) used by the first rotator stage 24, the second rotator stage 26, and the Nth rotator stage 28.

	TABLE 4
	i	0	1	2	3	4	5
N = 5	a(i)	511	498	474	437	389
Rotator	b(i)	40	121	195	268	334
Stages
N = 6	a(i)	1023	1005	970	919	853	771
Rotator	b(i)	67	200	330	453	568	675
Stages

Referring back to Table 3, the five rotator stage example is configured for coefficients that are 9 bits in size. As a result, the maximum value for the coefficients of the five rotator stage example is 512. The six rotator stage example is configured for coefficients that are 10 bits in size. Therefore, the maximum value for the coefficients of the six rotator stage example is 1,024. Dividing the coefficients of the first components a(i), and the second components b(i) by the maximum values 512 and 1024 for the five rotator stage example and the six rotator stage example yields decimal values listed in Table 5 below. Notice that the coefficients of the first components a(i), and the second components b(i) of the five rotator stage example in Table 5 are close to the first components a(i), and the second components b(i) of Table 1 multiplied by a CorrectionFactor of 1.001 that is calculated using equation (3).

	TABLE 5
	i	0	1	2	3	4	5
N = 5	a(i)	0.9980	0.9727	0.9258	0.8535	0.7598
Rotator	b(i)	0.0781	0.2363	0.3809	0.5234	0.6523
Stages
N = 6	a(i)	0.9990	0.9814	0.9473	0.8975	0.8330	0.7529
Rotator	b(i)	0.0654	0.1953	0.3223	0.4424	0.5547	0.6592
Stages

FIG. 3 is a block diagram of a second embodiment of the CORDIC 10 that includes mode circuitry for providing outputs of norms that are divided by the square root of two. The mode circuitry is usable to process vectors associated with a square constellation such as a square quadrature amplitude modulation (QAM) constellation. In particular, the mode circuitry includes an input adder circuit 62 adds the first component I of the vector IQ (FIG. 1) to the second component Q of the vector IQ. A first divider circuit 64 divides the result of the input adder circuit 62 by two. A first multiplexer 66 receives output from the I input 12 and the first divider circuit 64. A mode bit memory 68 provides a mode bit to the first multiplexer 66, which in turn uses the mode bit to select between outputting the first component I of the vector IQ or outputting the divided sum of the first component I of the vector IQ and the second component Q of the vector IQ.

The mode circuitry further includes an input subtractor circuit 70 that subtracts the second component Q of the vector IQ from the first component I of the vector IQ. A second divider circuit 72 divides the result of the input subtractor circuit 70 by two. A second multiplexer 74 receives output from the Q input 14 and the second divider circuit 72. A mode bit of 0 will result in a regular norm, whereas a mode bit of 1 will result in a norm divided by the square root of two.

An additional round and RS circuit 76 can be implemented if the width d of the output of the first MAX circuit 20 is less than the width n−1 of the I input 12. Likewise, another additional round and RS circuit 78 can be implemented if the width d of the output of the first MIN circuit 22 is less than the width n−1 of the Q input 14. In operation, the round and RS circuit 76 and the round and RS circuit 78 round their respective n bit wide contents before right shifting their n bit wide contents to drop lower bits until only d bits remain.

FIG. 4 is a voltage versus time graph of an envelope tracking signal (ETS) that can be generated using output from the CORDIC 10 (FIGS. 2 and 3). The ETS is shown in dashed lines and matches the modulation envelope of a modulated radio frequency carrier (RFC). The output from the CORDIC 10 does not produce the ETS directly. Instead, the CORDIC 10 is incorporated into other circuitry that forms the ETS.

FIG. 5 is a block diagram of a digital envelope tracking system 80 that incorporates the CORDIC 10. The digital envelope tracking system 80 includes a transmitter section 82 that drives a power amplifier module (PAM) 84 having power amplifier stages 86 with a bias control 88. A front end module (FEM) 90 receives the output from the PAM 84 and passes the output through selectable filters 92 to a transmit antenna 94 via RF switches 96. A fast switch mode power supply (SMPS) converter 98 supplies power to the PAM 84. The fast SMPS converter 98 is controlled through a mobile industry processor interface (MIPI) RF front-end (RFFE) standard interface 100. A general purpose analog-to-digital converter (ADC) 102 is usable to monitor supply voltages provided to the PAM 84 by the fast SMPS converter 98.

The TX section 82 includes an ETS generator 104 that drives the fast SMPS converter 98 to produce the ETS (FIG. 4). The ETS generator 104 receives a digital gain control (GainControl_dB) signal along with a stream of norm outputs from the CORDIC 10.

The TX section 82 also includes a digital modulator 106 that separates a transmit signal TX into a digital in-phase (I) signal and a digital quadrature (Q) signal. A timing block 108 provides timing advances and delays for the digital I signal and the digital Q signal in response to base station requests. The timing block 108 also provides interpolation for achieving higher clock frequencies.

A digital gain control 110 provides gain to the digital I signal and the digital Q signal in cooperation with the GainControl_dB signal. The cooperation ensures that the amplitude of the ETS (FIG. 4) and the amplitude of the RFC (FIG. 4) substantially match. A fixed delay 111 on the order of nanoseconds ensures that the stream of norm values is synchronized with the propagation of the digital I signal and the digital Q signal that are output from the digital gain control 110. A first digital-to-analog converter (DAC) 112 converts the digital I signal into an analog I signal that is filtered by a first filter 114. Similarly, a second DAC 116 converts the digital Q signal into an analog Q signal that is filtered by a second filter 118.

A first mixer 120 mixes the analog I signal with an RF signal generated by an RF oscillator 122. A second mixer 124 mixes the analog Q signal with the RF signal. Mixed outputs from the first mixer 120 and the second mixer 124 combine to produce the modulated RFC shown in FIG. 4. A variable attenuator 126 is usable in cooperation with the GainControl_dB signal to adjust the gain of the RFC.

The ETS generator 104 includes a multiplier circuit 128 that multiplies GainControl_dB with the stream of norm values output from the CORDIC 10. A look-up-table (LUT) 130 provides pre-distortion to the stream of norms to match distortion produced by the power amplifier stages 86. A programmable delay 132 is usable to finely tune synchronization between the stream of norm values and the RFC (FIG. 4). A group delay compensator 134 is included to compensate for a dynamic bandwidth response of the fast SMPS converter 98. Lastly, the ETS generator has a third DAC 136 for converting the stream of norm values into a differential output that drives the fast SMPS converter 98.

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

Claims

1. A coordinate rotation digital computer (CORDIC) for computing a norm of a vector having a first component and a second component, comprising: a first maximum value circuit for outputting a maximum operand that is a larger one of either an absolute value of the first component or an absolute value of the second component;a minimum value circuit for outputting a minimum operand that is a smaller one of either an absolute value of the first component or an absolute value of the second component;N rotator stages, each corresponding to a unique one of N predetermined vectors, each having a first multiply circuit for multiplying the maximum operand by a cosine coefficient of a corresponding predetermined vector to output a first rotation component, a second multiply circuit for multiplying the minimum operand by a sine coefficient of the corresponding predetermined vector to output a second rotation component and an adder circuit for adding the first rotation component to the second rotation component to output one of N results; anda second maximum value circuit for outputting a maximum one of the N results.

2. The CORD IC of claim 1 wherein the N predetermined vectors are unit vectors.

3. The CORD IC of claim 1 wherein each adder circuit of the N rotator stages comprises: a results register having an overflow bit that is set when a result of an addition performed by the adder circuit is larger than the results register; anda results multiplexer that outputs the result of an addition performed by the adder circuit if the overflow bit is not set or outputs a predetermined maximum value if the overflow bit is set.

4. The CORDIC of claim 1 wherein each adder circuit of the N rotator stages include a round and right shift (RS) circuit adapted to round addition results and then right shift the addition results until a predetermined bit wide result remains for output.

5. The CORDIC of claim 1 wherein each cosine coefficient a(i) is given by the following equation,

6. The CORDIC of claim 5 wherein each sine coefficient b(i) is given by the following equation,

7. The CORDIC of claim 6 wherein the dimensionless correction factor is given by the following equation,

8. The CORDIC of claim 1 further including mode circuitry adapted to output norms divided by the square root of two.

9. The CORDIC of claim 8 wherein the mode circuitry comprises: an input adder circuit to add the first component to the second component;a first divider circuit to divides results of the input adder circuit by two; anda first multiplexer for selecting either the first component or the results of the input adder circuit divided by two.

10. The CORD IC of claim 9 wherein the mode circuitry further comprises: an input subtractor circuit to subtract the second component from the first component;a second divider circuit to divides results of the input adder circuit by two; anda second multiplexer for selecting either the second component or the results of the input subtractor circuit divided by two.

11. An envelope tracking system comprising: a power amplifier module (PAM);a fast switch mode power supply (SMPS) converter adapted to supply modulated power to the PAM in response to an envelope tracking signal (ETS);an envelope tracking signal (ETS) generator adapted to drive the SMPS converter in response to a stream of vector norms; anda CORDIC for computing a stream of vector norms, each having a first component and a second component, comprising: a first maximum value circuit for outputting a maximum operand that is a larger one of either an absolute value of the first component or an absolute value of the second component;a minimum value circuit for outputting a minimum operand that is a smaller one of either an absolute value of the first component or an absolute value of the second component;N rotator stages, each corresponding to a unique one of N predetermined vectors, each having a first multiply circuit for multiplying the maximum operand by a cosine coefficient of a corresponding predetermined vector to output a first rotation component, a second multiply circuit for multiplying the minimum operand by a sine coefficient of the corresponding predetermined vector to output a second rotation component and an adder circuit for adding the first rotation component to the second rotation component to output one of N results; anda second maximum value circuit for outputting a maximum one of the N results.

12. The envelope tracking system of claim 11 wherein the N predetermined vectors are unit vectors.

13. The envelope tracking system of claim 11 wherein each adder circuit of the N rotator stages comprises: a results register having an overflow bit that is set when a result of an addition performed by the adder circuit is larger than the results register; anda results multiplexer that outputs the result of an addition performed by the adder circuit if the overflow bit is not set or outputs a predetermined maximum value if the overflow bit is set.

14. The envelope tracking system of claim 11 wherein each adder circuit of the N rotator stages include a round and right shift (RS) circuit adapted to round addition results and then right shift the addition results until a predetermined bit wide result remains for output.

15. The envelope tracking system of claim 11 wherein each cosine coefficient a(i) is given by the following equation,

16. The envelope tracking system of claim 15 wherein each sine coefficient b(i) is given by the following equation,

17. The envelope tracking system of claim 16 wherein the dimensionless correction factor is given by the following equation,

18. The envelope tracking system of claim 11 further including mode circuitry adapted to output norms divided by the square root of two.

19. The envelope tracking system of claim 18 wherein the mode circuitry comprises: an input adder circuit to add the first component to the second component;a first divider circuit to divides results of the input adder circuit by two; anda first multiplexer for selecting either the first component or the results of the input adder circuit divided by two.

20. The envelope tracking system of claim 19 wherein the mode circuitry further comprises: an input subtractor circuit to subtract the second component from the first component; anda second divider circuit to divides results of the input adder circuit by two; and a second multiplexer for selecting either the second component or the results of the input subtractor circuit divided by two.