Digital Predistortion of Non-Linear Devices
The present invention relates to digital pre-distortion of non-linear devices. In particular, the invention relates to digital pre-distortion of power amplifiers in network nodes such as base stations which deal with traffic in bursts. Spectral efficiency is increasingly important in today's mobile communications. The non-constant envelope (high peak-to-average ratio) digital modulation schemes used in many 2.5G and 3G wireless systems make RF power amplifier (PA) linearity and efficiency a crucial design issue. Typically, linearity is achieved either by reducing efficiency or by using linearization techniques. For a Class A PA, simply ‘backing off’ the input can improve linearity, but this reduces power efficiency and increases heat dissipation. When considering the vast numbers of base stations wireless operators need to account for, increased power consumption is not a realistic possibility, and linearization techniques are therefore required. Digital pre-distortion (DPD) has become the most important linearization technique for nonlinear devices such as Base Station Power Amplifiers and frequency mixers. The basic concept of a pre-distorter is to calculate the inverse nonlinearity function of the nonlinear device. The pre-distortion is applied to the baseband signal before modulation, up-conversion, and amplification by the PA, so as to reduce Intermodulation (IM). The nonlinear expression can be implemented with polynomials or Look-up Tables (LUTs). The nonlinearity function is usually a simplified Volterra series that characterizes the inputs together with the “memory effects” of the device for better linearization performance. When a PA is stimulated with signals with low average power variation, the learning of the DPD can reach quite good precision. Unfortunately, in practice, signals are very seldom continuous in power. They vary for each timeslot depending on different power schemes. Every time the power is applied (i.e. there is a transient in power) the DPD and PA will together generate a residual IM which is the mismatch of the pre-distorter spectrum and PA spectrum. The amount of transient IM is dependent on the difference in PA characteristics between the average state and transient state. Consider, for example, a PA in a Base Station operating Time-Division Duplex (TDD). In TDD operation the PA will be turned off when transmission is not taking place, to reduce the noise in the receiver and unnecessary power consumption. This procedure means that the PA will be in a transition state during the whole TX transmitting period. This is illustrated in Existing approaches are often in terms of static DPD. The DPD is adapted with samples from somewhere in the transmit burst 3 Alternatively, the DPD parameters could be remembered from the end of one burst 4 This can be understood with reference to It would be desirable to design a DPD scheme in which the transient IM at the start of each burst is reduced. It is an object of the present invention to address, or at least alleviate, the problems described above. In accordance with a first aspect of the present invention there is provided a signal manipulation unit for use in a network element of a telecommunications network. The signal manipulation unit comprises a non-linear device for manipulating the signal. A digital pre-distortion unit is provided for generating a pre-distortion function using stored parameters and applying it to the signal before it reaches the non-linear device so as to compensate for nonlinearities in the device. An adaptation unit is provided for receiving feedback from an output of the device and adapting the parameters in response to the received feedback. A storage unit is operatively connected to the adaptation unit and pre-distortion unit for storing the parameters. The digital pre-distortion unit comprises a function generator configured to generate a first pre-distortion function on the basis of a first set of parameters shortly before an initial discrete increase in power applied to the device. A signal distorter is configured to apply the first pre-distortion function to the signal. A clock is configured to determine a predetermined time period after the initial power increase. The pre-distortion unit is configured so that, after the predetermined time period, the function generator stops generating the first pre-distortion function and starts generating a second pre-distortion function on the basis of a second set of parameters. The pre-distortion unit is further configured so that, shortly before the start of a subsequent discrete increase in power applied to the device the function generator starts generating the first pre-distortion function on the basis of the first set of parameters. The first pre-distortion function can thus be used to compensate for IM when the device is in a “transient” state, just after a sharp increase in power. Once the device has settled down, the second pre-distortion function can be used to compensate until the next sharp increase in power, when the first pre-distortion function should be re-activated. In other words, there are at least two DPD states in time that represent the beginning and end of the burst that represent the behaviour change of a non-linear device. It will be appreciated that, in most situations, there will be decreases in power between the initial discrete increase and subsequent increases. The signal manipulation unit may be configured to store the first set of parameters in the storage unit when the first pre-distortion function stops being applied after the initial power increase, so that the first set of parameters used to start generating the first pre-distortion function shortly before the subsequent power increase correspond to those at the end of the predetermined time period after the initial power increase. This approach enables the first pre-distortion function for subsequent power increases to “learn” from previous power increases. Rather than returning to the same parameters as used at the beginning of the previous transient, the first pre-distortion function can use the parameters stored at the end of the previous transient period. By comparison, the second set of parameters used when the digital pre-distortion unit starts generating the second pre-distortion function may correspond to the first set of parameters when the digital pre-distortion unit stops generating the first pre-distortion function. In other words, the second pre-distortion function may start from where the first pre-distortion function leaves off. The signal may be in the form of a plurality of discrete bursts, with each power increase corresponding to the start of a burst. The pre-distortion functions may be generated from a LUT, or using polynomials having coefficients populated by the parameters. The pre-distortion unit may be configured to generate and apply a plurality of different pre-distortion functions using a plurality of different sets of parameters to compensate for different power conditions in the non-linear device. The non-linear device may be a PA for amplifying the signal. The signal manipulation unit may be provided in a network element, optionally a Base Station, which may operate in Time Division Duplex mode. In accordance with another aspect of the present invention there is provided a method of compensating for nonlinearities in a non-linear device for manipulating a signal. Shortly before an initial discrete power increase is applied to the device, a first pre-distortion function is generated on the basis of a first set of parameters and applied to the signal before it reaches the device. A predetermined time period after the initial power increase, the first pre-distortion function stops being applied to the signal, and a second pre-distortion function is generated on the basis of a second set of parameters and applied to the signal. Shortly before a subsequent discrete power increase is applied to the device, the first pre-distortion function is generated on the basis of the first set of parameters and applied to the signal. The invention also provides a computer program comprising computer readable code which causes a signal manipulation unit to operate as the signal manipulation unit described above or carry out the method described above. The invention also provides a computer program product comprising a computer readable medium and a computer program as just described stored on the computer readable medium. Some preferred embodiments of the invention will now be described by way of example only and with reference to the accompanying drawings, in which: A Digital Pre-Distortion is generally made up of a Pre-Distorter and an Adaptation algorithm. The Pre-Distorter is usually built of LUT blocks (or a set of polynomial coefficients) that contain the inverse nonlinear information of the PA. PA behaviour changes when the signal drops down to low power levels, with a number of time constants. Similar behaviour happens when a signal jumps to high power (as shown in The DPD update rate is constructed to compensate for the PA Behaviour variation introduced from the power varying signal. In the TDD operation this change is not continuous: instead, the signal is made up of bursts 3 One way to overcome this problem is to divide the time of the transmitted burst into several subsets. Each time subset is associated with a set of DPD parameters designed to model the compensation of the PA behavior for that particular part of the transmitted burst. The DPD parameters for the start of the burst will be different to those for later on in the burst. This can be understood by reference to The advantage of this is that it does not require a time critical algorithm, but the disadvantage is that it requires a lot of memory to represent and write all of the subsets. The most obvious disadvantage of this is the inaccuracy of the compensation due to the variation off the statistical content during the burst. A more sophisticated approach enables residual IM or transient IM to be reduced using a transient DPD set that is adapted to the start of the transmitted burst. The normal DPD operation will then track down the IM during slow PA behavior change until steady state. This can be understood with reference to The transient DPD setting 20 When the next burst is applied, the transient DPD setting 20 This process is then repeated for all bursts. The different types of setting can be considered as long or short term memories. The parameters from the end of one transient DPD setting 20 It will be appreciated that the above description describes two DPD settings, but the concept can be extended to additional sets of parameters, each compensating for a specific time constant of the PA behaviour. This is a technique that gives optimum performance in transient operation and superior precision. The memory requirement is low but the algorithm requirement is high. A predetermined time after the start of the burst 2 It will be appreciated that variations from the above described embodiments may still fall within the scope of the invention. For example, the operation of a PA, generally for use in a base station, has been described. However, the approach can be used for any non-linear device used to manipulate a signal, such as for example a frequency mixer. Furthermore, the approach has generally been described in the context of a “bursty” signal where the power applied approximates a square wave. It will be appreciated that the same approach can be used in any situation where there is a sharp increase in power after a period in a steady state, so that temperature, charge etc. has stabilised. Thus it may apply to devices having a number of different power levels, for example. A method and apparatus for compensating for nonlinearities in a non-linear device for manipulating a signal is described. Shortly before an initial discrete power increase is applied to the device, a first pre-distortion function is generated on the basis of a first set of DPD parameters and applied to the signal before it reaches the device. A predetermined time period after the initial power increase, the first pre-distortion function stops being applied to the signal, and a second pre-distortion function is generated on the basis of a second set of DPD parameters and applied to the signal. Shortly before a subsequent discrete power increase is applied to the device, the first pre-distortion function is generated on the basis of the first set of DPD parameters and applied to the signal. 1-18. (canceled) 19. A signal manipulation unit for use in a network element of a telecommunications network, comprising:
a non-linear device for manipulating a signal; a digital pre-distortion unit adapted to generate a pre-distortion function, using a set of stored parameters, and to apply the pre-distortion function to the signal before it reaches the non-linear device so as to compensate for nonlinearities in the device; an adaptation unit adapted to receive feedback from an output of the non-linear device and to adapt the parameters in response to the received feedback; and a storage unit for storing the parameters used by the pre-distortion unit to generate the pre-distortion function; wherein the digital pre-distortion unit comprises:
a function generator configured to generate a first pre-distortion function on the basis of a first set of parameters shortly before an initial discrete increase in power applied to the non-linear device; a signal distorter for applying the pre-distortion function to the signal; and a clock for determining a predetermined time period after the initial power increase; and wherein the pre-distortion unit is configured so that, after the predetermined time period, the signal distorter stops generating the first pre-distortion function, and starts generating a second pre-distortion function generated on the basis of a second set of parameters; and shortly before the start of a subsequent discrete increase in power applied to the device, the signal distorter starts generating the first pre-distortion function on the basis of the first set of parameters. 20. The signal manipulation unit of 21. The signal manipulation unit of 22. The signal manipulation unit of 23. The signal manipulation unit of 24. The signal manipulation unit of 25. The signal manipulation unit of 26. The signal manipulation unit of 27. A network element comprising the signal manipulation unit of 28. The network element of 29. The network element of 30. A method of compensating for nonlinearities in a non-linear device for manipulating a signal, the method comprising:
shortly before an initial discrete power increase is applied to the non-linear device, generating a first pre-distortion function on the basis of a first set of parameters and applying the first pre-distortion function to the signal before it reaches the non-linear device; a predetermined time period after the initial power increase, stopping applying the first pre-distortion function, generating a second pre-distortion function on the basis of a second set of parameters, and applying the second pre-distortion function to the signal; and shortly before a subsequent discrete power increase is applied to the non-linear device, generating the first pre-distortion function on the basis of the first set of parameters and applying the first pre-distortion function to the signal. 31. The method of adapting the first set of parameters in response to feedback from an output of the device while the first pre-distortion function is applied to the signal; when the first pre-distortion function is stopped at the predetermined time period after the initial power increase, storing the first set of parameters; when the first pre-distortion function is generated again shortly before the subsequent power increase, starting such generation using the stored set of parameters. 32. The method of 33. The method of 34. A non-transitory computer-readable medium comprising a computer program stored thereupon, the computer program comprising instructions that, when executed by a processor associated with a non-linear device, cause the processor to manipulate a signal by:
shortly before an initial discrete power increase is applied to the non-linear device, generating a first pre-distortion function on the basis of a first set of parameters and applying the first pre-distortion function to the signal before it reaches the non-linear device; a predetermined time period after the initial power increase, stopping applying the first pre-distortion function, generating a second pre-distortion function on the basis of a second set of parameters, and applying the second pre-distortion function to the signal; and shortly before a subsequent discrete power increase is applied to the non-linear device, generating the first pre-distortion function on the basis of the first set of parameters and applying the first pre-distortion function to the signal.TECHNICAL FIELD
BACKGROUND
SUMMARY
BRIEF DESCRIPTION OF THE DRAWINGS
DETAILED DESCRIPTION



