PAPR and the Cubic Metric for Low PA Power consumption Dr.Terence E. - PDF document

PAPR and the Cubic Metric for Low PA Power consumption Dr.Terence E. Dodgson Nokia Siemens Networks, Dr. Yanyan Wu Xi‟an Jiaotong Liverpool University  and the requirement for spectrum efficiency (accessible Abstract - With the every-increasing complexity of signal bandwidth being a relatively rare resource). Due to waveforms generated by transmitters of modern digital environmental considerations, a third driver, the requirement communications systems, there comes a corresponding demand for for power efficiency, needs to be taken into account, aiming transmitters to continue to operate efficiently, producing ultimately towards the development of Power Efficient minimum distortion, within specified limits. These days, Technology (or Green Technology). equipment requirements also include the need to operate in an energy efficient way, minimizing power consumption. In order to II. G REEN T ECHNOLOGIES - G ENERAL operate in the most efficient manner possible, instantaneous control of a Power Amplifier (PA) by measuring appropriate PA Perhaps one of the first, most obvious, areas where high power input signal characteristics, for example the Peak to Average efficiency has always been a requirement is the transmitter PA. Power Ratio (PAPR) or Cubic Metric (CM), should result in setting the PA Back-off (sometimes referred to as de-rating) as The increase in data rate, which allows for the delivery of accurately as possible. PAPR and a CM approaches to enhanced user services needs to be complimented by the determining PA Back-off are compared for OFDM signals, “efficient” delivery of that service, that is to say, resources together with a suitable PAPR reduction technique, in terms of should not be needlessly wasted and power should not be their ability to minimize power consumption and ensure efficient PA operation. unnecessarily consumed. Optimisation of such system resources is a necessary task in order to drive down the cost per bit to the Index Terms — CM, Green Radio, PAPR, Optimum Selective end user and associated Operational Energy (OpEn) costs. Mapping Multi-media (e.g. internet) service users require relatively high data rates combined with low data access delays, yet internet I. I NTRODUCTION sessions tend to be bursty in nature, with defined periods of When looking at services which require high user bit rates, activity and inactivity. To ensure efficient resource usage, users wireless communication systems have tended to lag behind should be assigned only the minimum resources which would wired systems due to problems caused by the less benign radio allow them to remain connected during these inactive times (i.e. when no data is being sent on the Uplink). In times of no user transmission environment. To the end user service is all important, however, when considering the environment, low activity ideally (ignoring transient off/on switching aspects) PA‟s should be turned off, otherwise their „b ack-off ‟ should be power consumption becomes as important. Current perception is that users would like to have the same service offered over a kept as small as possible, to ensure efficient operation, whilst also ensuring minimum signal degradation which might wireless link as they would be able to obtain over a wired one, despite any additional performance requirements. This has led otherwise occur (for example through signal clipping). A low to the design of cellular wireless systems which are capable of back-off value without signal distortion not only ensures working with relatively high data rates, in particular over the increased efficiency but also ensures power is channeled into the desired signal and not “wasted” on spurious signals. wireless link. Recent advances in the third generation (3G) system, currently referred to as Wideband Code Division Multiple Access (WCDMA) – formerly the Universal Mobile III. PAPR AND THE CM Telecommunications System (UMTS) – have resulted in Input signals to a PA tend to be very dynamic and this is concepts such as High Speed Packet Access (HSPA) where becoming increasingly true with modern communications theoretical data rates of 14.4Mbps and 5.76Mbps in downlink systems which involve the deployment of an increased number and uplink respectively are achievable. The next generation of channels and associated channel configurations. The way mobile phone system (LTE) aims to increase these data rates these channels are configured, and their subsequent processing further, in a spectrum efficient manner, using “bandwidth on by the PA, is likely to have an impact on any distortion that demand” techniques (the allocation of appropriate resource might be produced by that PA. With second generation, 2G, blocks depending on, amongst other items, the amount of data a systems PAPR-based back-off schemes were deemed sufficient user requires). The main, driving, forces behind the design of for efficient PA operation. However, more recent literature e.g. WCDMA and LTE have been the thirst for increased data rates [2], [3], [4], [5] suggests that PAPR „back - off‟ schemes are not sufficiently accurate when working with the, increasingly, TO BE ADDED

       complex signals that occur with enhanced 3G systems (and by 2 2 3 2 2 3 2 2 3 P ( a b ) ( a b ) ... ... ( a b )   CUBED 0 0 1 1 N 1 N 1 implication future, fourth generation, 4G, systems). A measure (3)     t N 1  2 2 3 is required which reflects better a signal‟s constituent parts, e.g. P ( a b )  CUBED t t t 0 for OFDM-type signals this may include, for each channel, the And, number of samples used when implementing the Fast Fourier        2 2 2 2 2 2 P ( a b ) ( a b ) ... ... ( a b )   Transform ( FFT) and how many of these samples are “active”, TOTAL 0 0 1 1 N 1 N 1 (4)    t N 1   2 2 the number of samples dedicated to the Cyclic Prefix (CP), the P ( a b ) TOTAL  t t t 0 modulation type and the associated bandwidth/number of resource blocks, as well as knowing the way in which the channels themselves are formatted and how they are combined, V. PAPR & CM C OMPARISON together with considering other system aspects such as how In order to observe the accuracy of PAPR and CM back-off power control is implemented. estimations, signals representative of those that would occur in practice might be simulated and the CM calculation performed, Ref [2] indicates that the (WCDMA) CM has the required making use of equation (2). Comparison of calculated CMs of accuracy for determining what the PA back-off should be for other system signals is then undertaken. Using the empirical any given channel format and combination for current, modern, results from ref [2] in the form of a graph the calculated CM cellular systems when considering WCDMA. Ref[10] provides values can be mapped onto the required PA power back-off a similar (CM) approach for LTE. values. A similar procedure can be carried out for back-off as determined by only PAPR measurement. Both sets of results can IV. CM FOR LTE then be compared to PA back-off measurements determined The modulator output/input to the PA can be represented as a through experimentation with appropriate PA hardware. complex stream of N samples, {v t }, as indicated in Figure 1; Considering LTE, the standard error for PAPR back-off prediction, is found to be 1.56dB which was also found to be, on average, more than 1.33dB poorer than the CM prediction Ref [10]. Approximated Back-off estimation plots for PAPR estimation and CM estimation can be seen in Figure 2 and Figure 3 respectively. PAPR Back-Off Estimation Error v. Measurement No. Figure 1 Modulation Output 2.5 From the literature, e.g. [9], the CM is given by; 2   3 3 (1) CM CEIL ([ 20 . log (( v ) 20 . log ( v ) )] / K ) 10 10 Re WCDMA norm rms norm f rms Back-Off Error 1.5 Where, 1 v is the normalized voltage waveform of the input signal. norm 0.5 v Re is the normalized voltage waveform of the reference norm f 0 signal (12.2kbps AMR Speech) and 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21  3 20 . log ( v ) 1 . 52 dB Measurement No. . 10 norm Re f rms Figure 2 PAPR Back-Off Estimation Error for a number of measurements K is a normalizing constant depending on the Spreading Factor (for WCDMA) or the bandwidth/number of resource blocks (for LTE), see ref[2] and ref[10] CM Back-Off Estimation Error v. Measurement No. 0.6 T he CM can be found using the modulation output samples‟ cubed power series and the power series itself. 0.5 0.4 Back-Off Error An alternative, more explicit, expression for the CM, ref[1], is 0.3 thus; 0.2 1 / 2   P   0.1  (2) CUBED c N .   rms 3  P  0 TOTAL 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Where, N is the number of samples considered, Measurement No. Figure 3 CM Back-Off Estimation Error for a number of measurements