1、专 业 学 号 班 级 指导教师 2010年 4月 . INTRODUCTIONTECHNIQUES for reducing power consumption have bemultimedia devices. Since digital signal processing is pervasive in such applications , it is useful to consider how algorithmic approaches may be exploited in construction low-power solution.A significant numbe
2、r of DSP function involve frequency-selective digital filtering in which the goal is to reject one or more frequency bands while keeping the remaining portions of the input spectrum largely unaltered. Examples include lowpass filtering for signal upsampling and downsampling , bandpass filtering for
3、subband coding, and lowpass filtering for frequency-division multiplexing and demultiplexing. The exploration of low-power solutions in these areas is therefore of significant interest.To first order, the average power consumption, P, of a digital system may be expressed as (1)Where Ci is the averag
4、e capacitance switched per operation of type i (corresponding to addition, multiplication, storage, or bus accesses), Ni is the number of operation of type i performed per sample, Vdd is the operating supply voltage, and fs is the sample frequency.Real-time digital filtering is an example of a class
5、 of applications in which there is no advantage in exceeding a bounded computation rate. For such applications, an architecture-driven voltage scaling approach has previously been developed in which parallel and pipelined architectures can be used to compensate for increased delays at reduced voltag
6、es . This strategy can result in supply voltages in the 1 to 1.5 V ra-nge by using conventional CMOS technology. Power supply voltages can be further scaled using reduced threshold devices. Circuits operating at power supply voltages as low as 70 mV (at 300 K) and 27 mV (at 77 K) have been demonstra
7、ted .Once the power supply voltage is scaled to the lowest possible level, the goal is to minimize the switched capacitance at all levels of the design abstraction. At the logic level, for example, modules can be shut down at a very low level basedon signal values. Arithmetic structures (e.g., rippl
8、e carry versus carry select) can also be optimized to reduce transi-tion activity. Architectural techniques include optimizing the sequencing of operations to mini mize transition activity, avoiding time-multiplexed architectures which destroy sig-nal correlations, using balanced paths to minimize g
9、litching transitions, etc. At the algorithmic level, the computational complexity or the data representation can be optimized for low power .Another approach to reduce the switched capacitance is to lower N,. Efforts have been made to minimize N, by intelli-gent choice of algorithm, given a particul
10、ar signal processing task. In the case of conventional filter design, the filter order is fixed based on worst case signal statistics,which is inefficient if the worst case seldom occurs. More flexibility may be incorporated by using adaptive filtering algorithms, which are characterized by their ab
11、ility to dynamically adjust the processing to thedata by employing feedback mechanisms. In this paper, we illustrate how adaptive filtering concepts may be exploited to develop low-power implementations for digital filtering.Adaptive filtering algorithms have generally been used to dynamically chang
12、e the values of the filter coefficients, while maintaining a fixed filter order. In contrast, our approach nvolves the dynamic adjustment of the filter order. This approach leads to filtering solutions in which the stopband energy in the filter output may be kept below a specified hreshold while usi
13、ng as small a filter order as possible. Since power consumption is proportional to filter order, our approach achieves power reduction with respect to a fixed-order filter whose output is similarly guaranteed to have stopband energy below the specified threshold. Power reduction is achieved by dynam
14、ically minimizing the order of the digital filter.The idea of dynamically reducing cost (in our case, power consumption) While maintaining a desired level of output quality (in our case, stopband energy in the filter output) emanates from the concept of approximate processing in computer science. Wh
15、ile approximate processing concepts may be used to describe a variety of existing techniques in digital signal processing processing (DSP), communications, and other areas, there has recently been progress in formally using these concepts to develop new DSP technique . Since our adaptive filtering t
16、echnique falls into this category, we refer to our approach as adaptive approximate filtering, or simply approximate filtering. DIGITAL FILTERING TRADE-OFFSA frequency-selective digital filter may have either a finite impulse response (FIR) or an infinite impulse response (IIR). It is well known tha
17、t IIR filters use fewer taps than FIR filters in order to provide the same amount of attenuation in the stopband region. However, IIR filters introduce nonlinear frequency dispersion in the output signals which is unacceptable in some application. For such cases, it is desirable to use symmetric FIR
18、 filters because of there linear phase characteristic.An important family of symmetric FIR filters corresponds to the symmetric windowing of the impulse responses of corresponding ideal filters. For example, a lowpass filter of this type has an impulse response given by (2)Where is a symmetric N-poi
19、nt window. This filter has cutoff frequency and may be implemented using a tapped delay line with N taps. For the purposes of this paper, we refer to such a filter as having order N. In Fig. 1, we display the frequency response magnitudes for three different values of N when is a rectangular window
20、and =. It should be observed that the mean attenuation beyond the cutoff frequency increase with filter order. Furthermore, with respect to a tapped delay-line implementation (see Fig. 2), the taps of the shorter Type I filter are subsets of the taps of the longer Type I filters. This ensures that i
21、f the filter order is to be decreased without changing the cutoff frequency, we can simply power down portions of the tapped delay line for the higher order filter. The price paid for such powering down is that the stopband attenuation of the filter decreases.Butterworth IIR filter are commonly used
22、 for performing frequency-selective filtering in applications where frequency dispersion is tolerable. The frequency response magnitudes of such filters do not suffer from the ripples which can be seen in the frequency response magnitudes for FIR filters. These IIR filters are commonly implemented a
23、s cascade interconnections of second-order sections, each of which consist of five multiplies and four delays, as shown in Fig.3. Also in Fig.3 is an illustration of a cascade structure for an eighth-order IIR filter as the cascade of four second-order section For the purposes of this paper, we cons
24、ider the order of a Butterworth IIR filter to be equal to twice the number of second-orderFrequency, normalizedFig. 1 Frequency response magnitudes for FIR filters of orders N=20,80,and 140Fig. 2 Tapped delay line of an FIR filter structure, and the powering down concept To preserve phase linearity,
25、 powering down must be applied at both ends of the structure.Fig. 3 Cascade implementation of an IIR filter structure. The detail of one of the second-order section is shown.sections in its cascade implementation., An interesting property of IIR Butterworth filters is that if the second-order sectio
26、ns are appropriately ordered, one may sequentially power down the later second-order sections and effectively decrease the net stopban attenuation of the filter. ADAPTIVE APPROXIMATE FILTERING In this section we present the details of our approximate processing approach to low-power frequency-select
27、ive filter-ing. As discussed earlier, frequency-selective filters are used in applications where the goal is to extract certain frequency components from a signal while rejecting others. Suppose a signal, xn, consists of a passband component, xpn, and a stopband component,. That is, (3)If it were po
28、ssible to cost-effectively measure the strength of the stopband component, , from observation of, we could determine how much stopband attenuation is needed at any particular time. When the energy in increases, it is desirable to increase the stopband attenuation of the filter. This can be accomplis
29、hed by using a higher-order filter. Conversely, the filter order may be lowered when the energy in decreases. We have developed a practical technique, based upon adaptive filtering principles, for dynamically estimating the energy fluctuations in the stopband component, , and using them to adjust th
30、e order of a frequency-selective FIR or IIR filter. As described in the previous section, the decrease in filter order enables the powering down of various segments of the filter structure. Powering down of the higher order taps has the effect of reducing the switched capacitance at the cost of decr
31、easing the attenuation in the stopband. Assuming that the FIR delay line is implemented using SRAM, even the data shifting operation of the higher order taps can be eliminated through appropriate addressing schemes.Our overall technique is depicted in Fig. 4. The quantity dn, which represents the en
32、ergy differential between the input and the output, is obtained as (4)where (5)and (6)The filter order for sample period n, Order n, is updated at each sample period. One approach for the update process is to choose Order n to be the smallest positive integer which guarantees that the stopband energy, Qn, of the output signal will be maintained below a specified threshold y. Assuming that the stopband portion of the
copyright@ 2008-2022 冰豆网网站版权所有
经营许可证编号:鄂ICP备2022015515号-1