This application claims priority to German Patent Application No. 10 2006 030 889.1-53, filed on Jul. 4, 2006, incorporated by reference herein as to its entirety.
In so-called direct digital synthesis (DDS), amplitude values of a period of an arbitrary 2π-periodic oscillation signal are stored as a lookup table (LUT) in a computer memory, for example a ROM (read only memory). As many amplitude values as possible are typically stored at as good as possible an amplitude resolution. A direct digital synthesizer (DDS) numerically calculates, in a clock cycle, a phase φ of the periodic signal using a phase accumulator, and determines associated amplitude values using the lookup table. Finally, an analog output signal is generated via a digital/analog converter (DAC) from the digital amplitude values. A tuning word forms a phase increment Δφ of the phase accumulator. That is, in a clock cycle n, phase φ[n] of the phase accumulator is increased by the phase increment Δφ, i.e. φ[n]=φ[n−1]+Δφ. A digital phase word of the accumulator has a specific number of bits, for example j-bits. Each time the phase accumulator overflows, for example in a transition from φ[n]=2j−1 to φ[n+1], a complete period of the periodic signal is generated. For this reason, phase increment Δφ of the phase accumulator and a clock frequency fclk of the direct digital synthesizer define an output frequency fout generated by the direct digital synthesizer.
One reason a direct digital synthesizer is often used is that its output frequency fout, its phase and amplitude may be altered in an accurate and rapid manner by means of digital signal processing.
In addition, one may fine-tune output frequency fout and phase φ[n] and quickly switch back and forth between different output frequencies. It is these properties that have rendered DDS technology popular, among other things, for radar applications, such as in motor-vehicle radar applications.
However, such a DDS system also has several disadvantages. Mapping of phase value φ[n] to a respective amplitude value is usually accomplished by means of a lookup table, as previously described. In a conventional configuration, such a lookup table may have 100,000 bits so as to suitably accomplish mapping of a phase value φ[n] to an amplitude value of the periodic signal. Such a large lookup table typically consumes a relatively large amount of power and additionally limits the clock frequency fclk of the DDS system. Also, for a high-resolution DDS system using a high-resolution lookup table, a digital/analog conversion of the amplitude value will be realized with a corresponding level of accuracy. For example, a DDS system having a 229,376 bit lookup table and a 14 bit DAC, clocked at a frequency fclk=400 MHz, results in a power consumption of approximately 500 mW.
A good level of linearity and a low power consumption of a DDS system may be achieved with digital/analog converters having a low digital resolution, known as low-bit DACs. However, these low-bit DACs generate, at the output of the DDS system, a portion of interference signals which is not to be neglected, so that their immediate use will cause other problems.
Various aspects are described herein. For example, various illustrative apparatuses and methods are described. An example of a described method is as follows: accumulating a phase increment that comprises an overflow to a phase information signal; mapping the phase information signal to an amplitude information signal; quantizing the amplitude information signal while feeding back a quantization noise using a first filter of an nth order, so that the feedback of the quantization noise comprises zeros at least two mutually different frequencies in a transfer function of the first filter; and performing digital/analog conversion of the quantized amplitude information signal to an oscillation signal. In addition, various illustrative apparatuses are described for performing the above method and other methods.
These and other aspects of the disclosure will be apparent upon consideration of the following detailed description of illustrative aspects.
A more complete understanding of the present disclosure may be acquired by referring to the following description in consideration of the accompanying drawings, in which like reference numbers indicate like features, and wherein:
The various aspects summarized previously may be embodied in various forms. The following description shows by way of illustration various examples in which the aspects may be practiced. It is understood that other examples may be utilized, and that structural and functional modifications may be made, without departing from the scope of the present disclosure.
Except where explicitly stated otherwise, all references herein to two or more elements being “coupled” or “connected” to each other is intended to broadly include both (a) the elements being directly connected to each other, or otherwise in direct communication with each other, without any intervening elements, as well as (b) the elements being indirectly connected to each other, or otherwise in indirect communication with each other, with one or more intervening elements. Furthermore, it should be appreciated that functional blocks or units shown in the drawings may be implemented as separate circuits in some embodiments, but may also be fully or partially implemented in a common circuit in other embodiments.
As will be discussed in more detail, in some illustrative embodiments, an apparatus is provided for generating an oscillation signal that may include a phase accumulator for accumulating a phase increment, including an overflow to a phase information signal, a phase quantizer for quantizing the phase information signal, a mapper for mapping the quantized phase information signal to an amplitude information signal, an amplitude quantizer for quantizing the amplitude information signal while feeding back the quantization noise by means of a first filter of an nth order having fixed filter coefficients, so that the feedback of the quantization noise includes zeros in at least two mutually different frequencies in the transfer function, and a digital/analog converter for converting the quantized amplitude information signal to an oscillation signal.
The zeros of the frequency response of the transfer function of the feedback of the quantization noise of the amplitude quantizer may be located in a range below the clock frequency fclk of the phase accumulation means, or of the digital synthesizer. The zeros may be at an output frequency fout of the digital synthesizer, or in a range symmetrical around output frequency fout.
In accordance with further illustrative embodiments, a digital synthesizer may include a phase accumulator with an overflow, a phase-to-amplitude mapper, and a digital/analog converter. The phase-to-amplitude mapper may be connected between the phase accumulator and the digital/analog converter. The digital synthesizer may further include a phase increment control having a fixed phase increment variation curve within a predetermined phase increment variation interval, the control's output being coupled to a phase increment default input of the phase accumulator, and a quantizer connected between the phase-to-amplitude mapper and the digital/analog converter and including a noise shaping feedback loop into which a filter of an nth order having fixed filter coefficients is connected, so that the noise shaping feedback loop includes zeros at least two mutually different frequencies in the transfer function.
The zeros of the frequency response of the noise shaping feedback loop of the quantizer may be in a range below the clock frequency fclk of the digital synthesizer. The zeros may be at an output frequency fout of the digital synthesizer, or in a range around output frequency fout. In other words, the zeros may be located, for example, at different frequencies between f=0 and f=fclk, and in particular near the output frequency fout, such as within the frequency range from ½fout to 3/2fout.
Phase accumulator 11 may include a j-bit phase register (not shown), which stores a digital phase increment Δφ present at an input 10a, and adds this phase increment Δφ to a phase register content φ[n] via a j-bit adder. The digital phase increment Δφ is input into the phase register by, for example, a phase increment control. At each clock n, this digital phase increment Δφ is added to the content φ[n] of the phase register. A clock signal having a clock frequency fclk is present at the input 10b of DDS 10. The phase increment Δφ represents a phase angle increment which is added to the preceding phase value every 1/fclk seconds so as to generate a linearly increasing digital phase value φ[n]. The phase value is generated by means of the 2j overflow property of the j-bit phase accumulator 11. The rate of overflows corresponds to output frequency fout in accordance with:
Formula (I) stems from the sampling theorem according to Shannon. Phase increment Δφ is an integral number, which is why a frequency resolution Δf of DDS 10 results, with Δφ=1, in which:
Lookup table 12 maps the digital phase information signal φ[n] to amplitude values r[n] of a sinusoidal oscillation. In an ideal case, i.e. with no phase and amplitude quantization, what results is an output sequence of lookup table 12:
wherein φ[n] represents the j-bit phase register value of the nth clock period.
The output of lookup table 12 is fed to digital/analog converter 13, which gives rise to an analog oscillation signal. It can be shown that the output spectrum of DDS 10 includes frequencies f=n·fclk±fout (n=0, 1, . . . ). The amplitudes of these frequency components are weighted with a function:
This effect may be responded to, for example, with an inverse sinc(f/fclk) filter in accordance with DAC 13.
Noise and further interference-signal portions may arise, for example, by cutting off the j-bit phase accumulator output signal φ[n] in that only k most significant bits of the j-bit phase information signal φ[n] are utilized for addressing the amplitude values of lookup table 12. If only the m most significant bits are taken into account by a one-bit amplitude output signal of lookup table 12 for the digital/analog conversion, the same may apply. Also, distortions may arise due to a compression of the amplitude values of lookup table 12 and to the finite accuracy of the amplitude values stored within lookup table 12.
In an ideal case, i.e. without phase and amplitude information being cut off and/or quantized, the output sequence r[n] of DDS 10 is given by:
wherein n designates the nth clock period.
The memory space that would be used for encoding the complete word width (j bits) of phase accumulator 11 within lookup table 12 may be typically too large to be economical. Therefore, for example, only the k most significant bits of the phase accumulator output signal may be used for addressing lookup table 12 so as to determine the amplitude values of the periodic oscillation signal. If the output of phase accumulator 11 is limited, or quantized, to the k most significant bits before the lookup operation is performed, the output sequence of DDS 10 would change to:
Equation (6) may be rewritten to read:
wherein eφ[n] signifies a phase error signal that may arise as a result of the quantization of the j-bit phase information signal to k most significant bits. Phase error signal eφ[n] may be limited, in terms of its magnitude, in accordance with eφ[n]<2j-k, and also may be periodic. An example of a sequence of the phase error signal eφ[n] is shown in
As may be seen from
Quantizing the digital amplitude signal r[n] at the output of lookup table 12 from 1 to m most significant bits may also lead to an interference in the output spectrum of direct digital synthesizer 10. If one assumes that the phase quantization error eφ[n] does not exist, the input signal to digital/analog converter 13 is given:
wherein eA[n] represents the amplitude quantization error caused by a lookup-table amplitude word of a finite length. The sequence of the amplitude quantization error eA[n] also may be periodic.
To cancel the periodicity of the quantization errors of the phase information signal and of the amplitude information signal, and thus to spread the interference energy of the quantization noise into a broad-band noise, a principle of quantization error feedback may be applied. Quantization error feedback may result in noise shaping. Noise shaping is understood to mean methods of shaping the spectrum of undesired interference and/or noise signals in a controlled manner. In principle, by means of noise shaping, interference power may be transformed, from certain specific spectral ranges, to frequency ranges located outside a useful-signal band. In this context, the transformation may be conducted such that the spectral components shifted within the frequency range will have less influence on the useful signal, i.e. the interference power in the useful-signal band may be reduced. What this may mean, in concrete terms, for the application of noise-shaping methods within the DDS, is that interference-signal components located near the synthesized frequency fout may be shifted to more remote ranges.
In the direct digital synthesizer 100 shown in
Phase accumulator 11 may have, for example, a 17 bit wide phase increment Δφ, i.e. j=17, supplied to it via input 100a of direct digital synthesizer 100. However, a tuning word of a smaller width may alternatively be used. In the present example, phase accumulator 11 may work with a word width of 17 bits, which enables a high clock frequency fclk of the DDS system. The 17 bit phase accumulator output signal is quantized to an eight bit input signal (k=8) for lookup table 12 by means of phase quantizer 110 that includes a phase quantization error feedback structure as will be described in more detail below with reference to
Amplitude quantizer 120 as shown includes a phase information signal input 120a and an amplitude information signal output 120b. An adder, such as a one-bit adder 200, adds a one bit amplitude information signal to an output signal of a first FIR filter 210 (FIR=finite impulse response) switched into a feedback loop. Adder 200 is coupled to input 120a. The m most significant bits of the one-bit output of adder 200 are fed to output 120b of amplitude quantizer 120, whereas the (1−m) least significant bits of the one-bit amplitude information signal are fed, at the output of adder 200, to the input of FIR filter 210 of the feedback loop.
In some embodiments, FIR filter 210 includes, for returning the amplitude quantization error eA[n], two two-fold delay members 212, 214 and a four-fold delay member 216. The term “n-fold delay member” is intended to designate a device that outputs an incoming digital value in such a manner that same is delayed by n clocks. First two-fold delay member 212 has first and second outputs. The first output is coupled to an input of a multiplier 218, whereas the second output is coupled to an input of four-fold delay member 216. One output of multiplier 218 is connected to an input of second two-fold delay member 214.
The output signal of FIR filter 210 is configured such that an output signal of the first output of the first two-fold delay member 212 is multiplied, by multiplier 218, for example by a factor of 2.5, and in that the output signal of multiplier 218, which corresponds to the result of the multiplication, is in turn delayed by the second two-fold delay member 214. In addition, an output signal of the second output of the first two-fold delay member 212 is delayed by the four-fold delay member 216. To form the output signal of FIR filter 210, the output signal of multiplier 218, the output signal of the second two-fold delay member 214, and the output signal of the four-fold delay member 216 are added. This may mean that the transfer function of FIR filter 210 in the feedback path for feeding back the quantization error eA[n] of the amplitude information signal includes a z transform of the pulse response of filter 210, in accordance with:
F
1(z)=2.5z−2+2.5z−4+z−6. (9)
The transfer function of amplitude quantizer 120 with regard to the amplitude quantization noise eA[n] represented in the z domain may thus result in, in accordance with the present illustrative embodiment:
H
A(z)=1−F1(z)=1−2.5z−2−2.5z−4−z−6. (10)
In some embodiments, the coefficients of filter F1(z) 210, or the coefficients of transfer function HA(z), may be selected such that the zeros of the frequency response of transfer function HA(z) of amplitude quantization means 120 are in a range below clock frequency fclk of the phase accumulation means and/or of the digital synthesizer, i.e. in a range 0<f<fclk. In this context, the zeros may occur at different frequencies. In particular, the filter coefficients of filter F1(z) 210 may be selected, in some embodiments, such that the frequency zeros of HA(z) are at an output frequency fout of the digital synthesizer, and/or in a symmetrical range around output frequency fout.
A frequency response of the transfer function HA(z) of amplitude quantization means 120 may be determined, with regard to the amplitude quantization noise eA[n], by means of the context:
wherein ω stands for an angular frequency 2πf, and Tclk stands for a clock duration of 1/fclk.
For forming the quantization noise eA[n] of the amplitude signal, filter 210 need not necessarily exhibit any zeros at the frequency of f=0 in the feedback loop. A zero at the frequency of f=fclk/4 signifies a filter coefficient of zero, so that the filter structure of the FIR filter 210 of amplitude quantization means 120 may be highly simplified. The transfer function of amplitude quantizer 120 with regard to the amplitude quantization noise HA(z)=1−F1(z) of the sixth order may be composed of three sections of the second order, respectively, in the form of, for example, HA(z)=(1+a·z−1+z−1)·(1+b·z−1+z−)·(1+c·z−1+z−2). This may allow the design to be simplified, since in this case only one filter coefficient, respectively, is adjustable, and the others automatically provide a realizable and stable filter. If, for example, a=0, a zero of the transfer function HA(z) would result at f1=fclk/4. This may be explained by the fact that a filter having a transfer function (1+a·z−1+z−2) exhibits a zero at f=arccos(−a/2)·fclk/(2π), and thus the following results: a=0 f1=fclk/4.
When in the equation HA(z)=(1+a·z−1+z−2)·(1+b·z−1+z−2)·(1+c·z−1+z−2), i.e. a=0, and the coefficients b and c are set to have the same magnitude with different signs, specifically are set to the value of b=−c=−√2/2=−0.707, the transfer function set forth in equation (10) will result from multiplying. The filter coefficient of 2.5 may be realized in hardware by means of, for example, shifting and adding operations. The zeros of HA(z) resulting from b and c occur at the frequencies f2=arccos(−b/2)·fclk/(2π)=arccos(0.707/2)·fclk/(2π) and f3=arccos(−c/2)·fclk/(2π)=arccos(−0.707/2)·fclk/(2π). With a clock frequency of, for example, fclk=240 MHz, the feedback loop of amplitude quantizer 120 would have zeros at the frequencies f2≈46 MHz, f1≈60 MHz=fout=fclk/4 and f3≈74 MHz. The zeros at f2 and f3 thus may be located symmetrically around the zero at f1=fout=fclk/4.
The symmetry of the frequency zeros of HA(z) around output frequency fout, in particular fout=fclk/4, may allow the filter function F1(z) for be simplified for amplitude quantization, since every other filter coefficient would disappear. In this manner, circuit expenditure may be reduced, for example, by savings in terms of adders.
Filter 210 shown in
It is be noted that other FIR filter structures also may be feasible in the feedback path for feeding back the quantization error eA[n] of the amplitude information signal. One may generally represent, for example, a z transform of the impulse response of a sixth-order FIR filter in accordance with:
F
1(z)=a0+a1z−1+a2+z−2a3z−3+a4z−4+a5z−5+a6z−6 (12)
wherein a0, a1, a2, a3, a4, a5, a6 stand for the specified filter coefficients. The choice of the filter coefficients may depend, for example, on the clock frequency fclk of the DDS, and/or on the desired spectral properties of the output signal.
FIR filter 310 for returning the phase quantization error eφ[n] has three delay members 312, 314, 316 connected in succession. First delay member 312 has first and second outputs. The first output is coupled to an input of a multiplier 318, whereas the second output is coupled to an input of second delay member 314. Second delay member 314 also has first and second outputs, one output being connected to an input of third delay member 316.
The output signal of FIR filter 310 is formed in that an output signal of the first output of first delay member 312 is multiplied, by multiplier 318, for example by a factor of −1, and in that the output signal of multiplier 318 is added to the sum of the outputs of the two subsequent delay members 314 and 316. This may mean that the transfer function of FIR filter 310 in the feedback path for feeding back the quantization error eφ[n] of the phase information signal has a z trans-form of the pulse response of the filter in accordance with:
F
2(z)=−z−1+z−2+z−3. (13)
The transfer function of quantizer 110 with regard to the phase quantization noise eφ[n], represented in the z domain, thus results in:
H
φ(z)=1−F2(z)=1+z−1−z−2−z−3. (14)
Phase quantizer 110 may be implemented, for example, using only adders and/or subtractors, and registers, all of which are readily available using a variety of technologies. As was already described previously, Hφ(z), too, may be composed of a second-order portion (zero within the frequency band) and a first-order portion (zero at the edge of the frequency band at f=0). Phase quantizer 110 may exhibit a zero in the transfer function Hφ(z) of phase quantization means 110 with regard to phase quantization noise eφ[n], since the phase information is in a range of [0,2π] in a binary representation. The zero of Hφ(z) at f=0 may be desirable, since the output signal of phase accumulator 11 may be only positive and thus have a steady component which, however, may not be changed by filter 310. This is why, in accordance with the invention, a third-order FIR filter may be employed.
It should be noted that other FIR filter structures may be feasible in the feedback path for feeding back the quantization error eφ[n] of the phase information signal. Generally, a z transform of the pulse response of a third-order FIR filter may be represented in accordance with:
F
1(z)=b0+b1z−1+b2z−2+b3z−3 (15)
wherein b0, b1, b2, b3 stand for the fixed filter coefficients. The filter coefficients may be chosen to depend, for example, on the clock frequency fclk of the DDS, and/or on the desired spectral properties of the output signal.
A quantization error feedback may cancel a periodicity of the quantization errors eφ[n] and/or eA[n], and may thus convert interference energy concentrated in spectral terms to broad-band noise. Quantization noise of phase and amplitude may be formed by the respective quantization error feedback structures 110 and 120. In accordance with some embodiments, quantization noise shaping may be performed by means of feedback filters 210 and 310, respectively, with fixed filter coefficients, in such a manner that a relatively large part of the output spectrum of direct digital synthesizer 100 may exhibit a reduced quantization noise power, and such that the entire quantization noise power may be concentrated near DC (direct current) and a frequency of f=fclk/2. In accordance with some embodiments, a frequency range of approx. 0.3·fclk/2 may remain as a usable frequency range, for example for FMCW radar applications without necessarily needing to specifically adapt the filter coefficients of the quantization error feedback filters to the output frequency fout of direct digital synthesizer 100.
At a suitable clock frequency fclk, the usable spectral range adjusted by the fixed filter coefficients and, thus, by the zeros of the transfer function of the noise feedback, may be sufficiently large and may even be adapted to utilize direct digital synthesizer 100, as a reference and/or modulation source, for example for a PLL (phase locked loop) of a motor-vehicle radar system. In accordance with some embodiments, the structure of direct digital synthesizer 100 may be optimized for generating so-called frequency sweeps as are often used, for example, in FMCW radar systems.
Lookup table 12, depicted in
Next to an output signal 500 of a direct digital synthesizer having an output frequency of, e.g., fout=60 MHz at a clock frequency of, e.g., fclk=240 MHz, the zeros of the transfer function HA(z) of amplitude feedback structure 120 depicted in
The performance of quantization error feedback structure 120 in this example becomes clear in terms of a low noise level in the region between f1≈46 MHz and f3≈74 MHz. In this region, a spectral noise power density is at approx. −115 dBc/Hz. It is only the interference-signal components within this frequency band of f1≈46 MHz to f3≈74 MHz that may negatively affect a system performance of direct digital synthesizer 100, since all other frequencies outside this frequency band may be filtered out by means of a bandpass filter, not shown in
As was already described above, such an apparatus may be employed, for example, in motor-vehicle radar systems, such as in FMCW radar systems. A schematic block diagram of an illustrative circuit for frequency generation in such a radar system is depicted in
As shown in
Output frequency fRF may be higher, by a factor of N, than the frequency fout generated by direct digital synthesizer 100. The reference frequency fout generated by the DDS 100 for the PLL 620 may range from, for example, 100 MHz to 1 GHz. Phase frequency detectors may operate within a frequency range of, for example, several hundred KHz up to several hundred MHz.
In FMCW radar systems, frequency sweeps are typically performed. This frequency sweep may be created, for example, in that the carrier frequency fRF of the radar system is increased, within a specific time interval T, from a starting value to a final value. This increase may be linear, for example. A modulation of radar output frequency fRF may be achieved with the system depicted in
As was already shown in
It is also noted that in various embodiments, word widths may be used that differ from the above-described word widths for the phase information signal, the phase quantization noise, the amplitude information signal, and the amplitude quantization noise. A choice of these word widths may depend, among other things, on the performance of the hardware used.
In various embodiments, the phase quantizer and/or the amplitude quantizer may each be implemented as higher-order multiple error feedback structures for quantizing the phase information signal and the amplitude information signal, respectively, and may be optimized for a hardware and/or software implementation.
Due to the quantization of the phase information signal, a relatively small lookup table may be used for mapping the phase information signal to the amplitude information signal. Moreover, due to the quantization of the amplitude information signal, low-bit DACs may be employed, and it may be desirable, due to the multiple quantization error feedback structures used, to spectrally shape the quantization noise such that it has a particularly low spectral power density in a range around output frequency fout of the DDS.
Thus, various embodiments as described herein may enable a more performance-efficient DDS system using smaller interference-signal components that may be particularly suited to the low spectral power density of the quantization noise in a range around output frequency fout of the DDS system, for applications such as LFMCW radar systems (LFMCW=linear frequency modulated continuous wave).
In addition, the features described herein are not limited to radar applications, but may also be employed in other applications, such as in function generators that generate various signal shapes. In addition, various embodiments may be suitable for generating fast and accurate frequency and phase changes, for example for modulation purposes.
Any of the various functions described herein may also be implemented in software. The implementation may be effected by storing computer-readable instructions and/or data on a computer-readable storage medium, such as a magnetic and/or optical disk, magnetic tape, and/or memory. These computer-executable instructions may be read and executed by a programmable computer.
Number | Date | Country | Kind |
---|---|---|---|
102006030889.1-53 | Jul 2006 | DE | national |