FREQUENCY CONVERSION OF SIGNALS
The invention relates to methods and apparatus for conditioning a signal, involving several mixing processes operating in succession to frequency convert the signal.
Frequency conversion of a signal is commonly obtained by mixing the signal with a reference signal at a different frequency. Frequency conversion by mixing can suffer from the generation of unwanted image components due to harmonics created in the mixing process. To avoid such problems, it is common to implement a frequency conversion by means of several mixing processes interspersed with filters as appropriate.
In this field, a mixing process can be either a high-side mixing process or a low-side mixing process. In a mixing process, an input signal to the mixing process is converted to a different frequency by mixing with a reference signal (e.g. from a local oscillator) and issues as the output signal of the mixing process. If the reference signal used in an upconversion process is higher than the output signal, the upconversion process is said to be a high-side mixing process. If the reference signal used in an upconversion process is at the same frequency as, or at a lower frequency than the output signal, the upconversion process is said to be a low-side mixing process. Downconversion mixing processes can also be categorised as high-sided or low-sided, although the definitions of the categories are slightly different. A downconverting mixing process is said to be high-sided if the reference signal is of higher frequency than the input signal with which it is mixed to create the output signal. A downconverting mixing process is said to be low-sided where the reference signal is of frequency equal to or less than that of the input signal.
Generally, where one or more frequency conversions are performed on a signal, a frequency discrepancy can arise between the actual resultant signal and its planned characteristics. For example, consider the frequency conversion scheme shown in Figure 1, which allows a digital signal processor (DSP) to be used to predistort (i.e. precorrect) an RF input signal to a radio frequency power amplifier (RF PA). The RF input signal is downconverted in frequency, converted to the digital domain and supplied to the digital signal processor to be predistorted. The predistorted input signal is then converted to the
analogue domain and supplied to an upconverter. The upconverter will attempt to convert the predistorted input signal to a desired frequency but, in practice, the actual frequency of the output of the upconverter can differ from the desired frequency (because of, for example, frequency errors in reference signals to be mixed with a wanted signal). Such frequency discrepancies can be problematic in the case where, for example, the output of the RF PA is to be transmitted under very tight frequency constraints.
An aim of the invention is to ameliorate frequency discrepancies of the kind described above.
According to one aspect, the invention provides apparatus for conditioning a conveyed signal travelling on a signal path, comprising a plurality of mixing means arranged to operate in succession on the conveyed signal, wherein each mixing means has a respective reference signal, all the said reference signals are coherent, each mixing means is arranged to frequency convert the conveyed signal by mixing its reference signal with the conveyed signal and the number of said mixing means that are arranged to perform high-side mixing is even.
The invention also consists in a method of conditioning a conveyed signal travelling on a signal path, comprising a plurality of mixing steps performed in succession on the conveyed signal, wherein each mixing step uses a respective reference signal, all the said reference signals are coherent, each mixing step comprises frequency converting the conveyed signal by mixing its reference signal with the conveyed signal, and the number of said mixing steps that are arranged to perform high-side mixing is even.
By ensuring amongst a group of mixing processes whose mixing reference signals are coherent that the total number of mixing processes that use high-side mixing is kept to an even number (which could be zero), the discrepancy between the wanted and actual frequencies of the final signal is substantially reduced. Moreover in most circumstances, the discrepancy is eliminated.
In one embodiment, the coherency of the mixer reference signals is achieved by deriving them from the same source, e.g. a master oscillator.
Other signal processing operations may be performed alongside the mixing operations. For example, a signal processing operation could be performed before or after the mixing processes, or even between two of the mixing processes. The mixing processes may be used to frequency convert the input signal to, or output signal from, a signal processing operation included alongside the mixing processes. Such signal processing operations may be analogue domain operations or digital domain operations (e.g. digital predistortion of a signal). Furthermore, it is possible that one (or more) of the mixing processes is performed in the digital domain with the clock signal constituting its mixing reference signal being coherent with the mixing reference signals used in the analogue mixing processes.
As discussed above, the invention concerns the frequency conversion of signals and it will be apparent to the skilled person that the invention is applicable to single frequency or narrow bandwidth signals and also to multiple frequency or broadband signals. Where a signal having components at different frequencies undergoes frequency conversion by a given frequency shift, then all of the components are subjected to the same shift.
By way of example, an embodiment of the invention will now be described, with reference to the accompanying figures, in which:
Figure 1 is a block diagram illustrating a frequency conversion scheme allowing a digital signal processor to operate on a RF signal; and
Figure 2 is a block diagram illustrating a mixing scheme in accordance with the invention as applied to the frequency conversion scheme of Figure 1.
Figure 2 illustrates how the invention can be used to ameliorate a frequency discrepancy between the desired and actual frequencies of the output of the upconverter of Figure 1. The downconverter involves two downconverting mixer processes 10 and 12 in series. The upconversion process operating on the output of the DSP comprises four mixing processes 13, 14, 16 and 18 operating in series. Mixer 13 is implemented within the DSP and performs its frequency upconversion in the digital domain.
Each of the mixers 10 to 18 mixes a respective reference signal 20 to 30 into the input signal to the RF PA. All of the reference signals 20 to 30 are derived from the same master
oscillator and are therefore coherent. Various filters are included in the downconversion and upconversion processes. Band pass filters 32, 34 and 36 operate on the outputs of mixing processes 10, 14 and 16 respectively. Low pass filters 38 and 40 operate on the outputs of mixing processes 12 and 13 respectively.
In this example, the downconverter reduces the frequency of the RF input signal to one which the DSP can accept. The DSP then digitally predistorts the input signal destined for the RF PA and emits an analogue version of the predistorted signal. The upconverter (in conjunction with mixer 13 in the DSP) raises the frequency of the predistorted input signal to the desired transmission frequency. The upconverted predistorted input signal is then amplified by the RF PA and transmitted from an antenna (not shown).
As shown in Figure 2, mixing processes 14 and 16 perform high-side mixing whereas mixing processes 10, 12, 13 and 18 perform low-side mixing. Since the reference signals used in mixing processes in Figure 2 are coherent and because the total number of high-side mixing processes used in the whole of the downconversion and upconversion chain is even (including any mixing processes implemented digitally within the DSP), the frequency errors in the downconversion/upconversion chain cancel so that the frequency discrepancy between the desired and actual frequencies of the output of the upconverter is eliminated.
The following table illustrates the error removal when the RF input signal is at 1840 MHz and the reference signals used by the mixing processes are all derived from a master oscillator which has a frequency error of 20 parts per million (ppm). The ideal frequencies are given in the left hand portion of the table and the actual frequencies using the reference oscillator with the 20 ppm frequency error are given in the right hand portion of the table.
As can be seen from the above table, the frequency error in the output of the upconverter is zero.
The following table (which has the same format as the earlier table) illustrates a counter example wherein all the reference signals are derived from a master oscillator with a frequency error of 20 ppm (as before) and the total number of high-sided mixing processes in the entire downconversion and upconversion chain is odd.
As can be seen from the above table, when the number of high-sided mixing processes is odd, the frequency discrepancy between the wanted and actual frequencies of the upconverter output is not zero.