IMPROVEMENTS IN OR RELATING TO RF RECEIVERS
The present invention relates to RF receivers. More particularly, the invention relates to the cancellation, or at least reduction, of distortion caused in components of a direct conversion receiver arrangement.
Radio frequency (RF) receivers are widely incorporated into modem communications equipment such as that used, for example, in mobile telecommunications systems, where RF receivers are typically employed in both mobile terminal units and base stations.
Most conventional receivers are manufactured using a superheterodyne receiver architecture, and such receivers are widely adopted in mobile telecommunications systems. Due to manufacturing pressure, however, there has been increasing interest in alternative architectures, such as the so-called direct conversion receiver architecture.
While superheterodyne receivers exhibit good selectivity and sensitivity, direct conversion receivers require fewer components, and direct conversion can also be implemented on-chip. Superheterodyne receivers translate all received signals within a frequency band, fc, to a fixed intermediate frequency (IF) band carrier frequency, f;. The frequency translation is performed by a mixer (alternatively referred to as a frequency converter). The fixed carrier frequency is said to be up-converted if the resulting IF is greater than the carrier frequency, or down-converted if the IF is lower than the carrier frequency. In general the frequency terms are related by
Ij — |imix -*- Ic|-
The mixing signal is generally supplied by a tuneable local oscillator (LO) and the output of the mixer is then filtered by a bandpass filter to select the component at the new frequency, £. Frequency mixing is also referred to as "heterodyning". Direct conversion (alternatively called zero-IF or homodyne) may be considered as a special case of the superheterodyne receiver architecture wheein the LO is set to the same frequency as the desired frequency band, fc and thus the IF, f, is set to zero. Filtering and amplification can therefore take place at dc, where gain is easier to achieve with low power. Filtering can be achieved with on-chip RC circuit components rather than bulky and expensive SAW (surface acoustic wave) filters.
The placement of the filter at baseband (usually split between the analogue and digital domains) permits multiple filter bandwidths to be included without a concomitant penalty in board area, since the filtering is accomplished on-chip. For this reason, direct conversion is expected to become increasingly important in the development of multimode receivers of the future. Multimode operation using a superheterodyne configuration requires separate IF filters for each mode, thus increasing the number of components required and their complexity.
Direct conversion is associated with some problems, however, and although an ideal linear receiver component would operate on an input signal in a linear manner, the use of non-ideal components can lead to intolerable levels of amplitude and/or phase distortion. For example, non-linear behaviour in a mixer may cause an amplitude modulation to phase modulation (AM/PM) transfer characteristic, whereby phase variations in the output amplified signal
are dependent upon amplitude variations in the input signal. Distortion may also be purely or partly AM/AM in nature.
Distortion is a consequence of the fundamental non-linearity of components, and may additionally be influenced by other physical factors, including changes in the operational characteristics of the component (i.e. operating frequency), temperature variations, power supply fluctuations and load mismatches.
In the absence of components with perfect linear transfer characteristics, non-linear distortion effects can appear as spurious signals having frequencies which are generally in simple arithmetic relation to input frequencies; for example harmonic distortion and intermodulation distortions (IMD).
Intermodulation and harmonic distortions are important classes of effects generally termed "mixing products". Intermodulation distortion (IMD) products can usefully be characterised in terms of their origins. The "order" of a mixing product, f, is given by the sum:
O(f) = |m| + |n| + . . . + |z| where f = mfi + nf2 + . . . + z - Thus 3fι, the third harmonic of fi, is of order three; so too is the
IMD product (2^ - f2). This 3rd order intermodulation distortion product will be referred to hereinafter as IM3. Likewise, the 2nd order intermodulation distortion product will be referred to as IM2.
IMD is present in both superheterodyne and direct conversion receiver architectures. In most superheterodyne RF receivers, IM3 dominates. In homodyne systems, on the other hand, even-order distortion (including IM2) can be severe. IM2 distortion is characterised
by a value known as the second-order (intermodulation) intercept point (IP2), which is defined as the intersection point of the fundamental frequency component gain curve and the second-order intermodulation product gain curve. The presence of even-order (IM2) products effectively demodulates amplitude modulated signals to baseband; this demodulation effect being known as 'direct detection' which, considered more generally, is caused by even order non-linear behaviour in mixers. In all RF communication systems employing direct conversion receiver architectures direct detection is unwanted and presents problems since, if a receiver tuned to a first base station is unable to reject IM2 products, the receiver will be prone to unwanted steps in dc output due to signals from other nearby base stations and/or mobile terminals.
It is known to seek to compensate for non-linear transfer characteristics in electronic components, such as power amplifiers, or briefly PAs, by applying a range of digital processing and analogue techniques, and apparatus effecting such compensation is, variously termed 'predistorter', inearizer' and 'equaliser'. The difference between terms is one of emphasis and order of application; a predistorter being an apparatus for applying a predistortion that seeks to complement any distortion introduced by the component before the input signal is fed to the component. On the other hand, linearizers attempt to bring the combined Iinearizer and PA arrangement as close as possible to an ideal linear PA and equalisers aim to flatten the distortion across an operating spectrum.
It is known to improve the IP2 performance of mixers used in mobile terminal applications through a combination of calibration
(trimming) and high dynamic range analogue-to-digital (A/D) conversion. Such techniques are however not generally suitable for implementation in base stations.
It is therefore an object of the invention to obviate or at least mitigate the aforementioned problems.
A more specific object of the present invention is to provide an apparatus and a generic method for the cancellation of the second (or higher) order intermodulation products, combined with an appropriate control system. In general, the present invention seeks either to apply a predistortion to a signal supplied to a non-ideal mixer device or to correct for distortion introduced by the mixer device adaptively. The predistortion or adaptive correction mainly seeks to compensate for IM2 products thereby improving IP2 performance of mixers used in direct conversion receiver architectures. It is further noted that receivers with this improved performance can readily be manufactured on bipolar or CMOS integrated circuits.
In accordance with one aspect of the present invention, there is provided apparatus for compensating for non-linear behaviour of an RF mixer device, the apparatus comprising: a first multiplier for squaring an input RF signal; a scaling multiplier, for multiplying the squared input signal by a scaling coefficient, and an adder, for adding the scaled, squared signal to a further signal; wherein the apparatus is coupled to the mixer device and
wherein said scaling coefficient is dimensioned and configured to adjust the RF signals in a sense tending to compensate for the non-linear behaviour of said mixer device.
In accordance with another aspect of the present invention, there is provided a method for compensating for non-linear behaviour of a mixer device (in a direct-conversion receiver architecture), the method comprising: receiving an input RF signal, X; squaring said input RF signal to generate a squared signal, X2; multiplying the squared signal, X , by a scaling coefficient, K2; and adding the scaled, squared signal, K2 X2, to a further signal; coupling the apparatus to the mixer device; and utilising said scaling coefficient to adjust the RF signals in a sense tending to compensate for the non-linear behaviour of said mixer device. In one preferred embodiment, the adder is disposed to receive said input RF signal as said further signal, and the output of said adder is coupled to the mixer device In this way, it is arranged that the adjustment of said RF signals is effected by predistorting them to allow for distortions that will be introduced by the mixer. It is further preferred in such circumstances that said scaling coefficient is derived from a store of said coefficients, which store conveniently comprises a look up table.
Moreover, the selection of said scaling coefficient from the store is preferably influenced by operational criteria affecting said mixer; said criteria including the temperature to which the mixer is subjected and/or its frequency of operation.
In another preferred embodiment, the mixer receives said RF input signals directly and is coupled to said apparatus so as to provide said further signal to said adder, and the scaling factor is generated by means of a correlation stage. By this means, the adjustment of said RF signals comprises an adaptive correction applied to the mixed signals.
Preferably, the correlation stage comprises, in series connection, a low pass filter, a correlator and an integrator; the correlator stage being connected to receive the squared input signal and to provide the scaling factor as its output. The first preferred embodiment of the invention provides apparatus as a part of an open loop arrangement whereby the apparatus serves as a predistorter.
The second preferred embodiment of the invention provides apparatus as a part of a closed loop arrangement whereby the apparatus serves as a Iinearizer, through the application of adaptive correction.
The mixer device is advantageously a Gilbert mixer, which is often used in direct conversion, an example of which is shown in Figure 3, and a description of which is provided at this point for convenience. The invention is, however, applicable to all mixers and frequency converters and is not in any way restricted to usage with Gilbert mixer devices.
The Gilbert mixer consists of two sections: an RF input section (Q5 & Q6), connected as an emitter-coupled differential pair, and a switching stage (Qι, Q2, Q3 & Q4). In this example, the emitters are degenerated by resistors RES & EO and the differential pair generates differential output currents i5 & iβ which are fed into the polarity switching stage of the mixer to effect the desired frequency conversion.
The principle of the switching stage is that the polarity of the differential currents i5 & 1$ are inverted with every half-cycle of the local oscillator signal. Ideally, at any time, two of the four switching transistors (Q Q2, Q & Q4) are "on" (saturated) and two are "off. The output Vrp+ is generated by i5*Ru3 during the positive half-cycle of VL0+; and by i6*RLi3 during the positive half-cycle of VLo- Similarly the output VIF- is generated by -6*RL24 during the positive half-cycle of VLO+, and by is*RL24 during the positive half-cycle of VLo-
The IF output signals can thus be represented as: VIF+ = VLO+*i5*RLi3 + VLO- *i6*RL13
VIF. = VL0+ *i6*RL24.+ VLO-*i5*RL24
As with all mixers, the Gilbert mixer exhibits significant nonlinear behaviour. A simple technique for modelling a non-linear system is to represent the transfer function by a Taylor series expansion. V0(t) = ko + kιVi(t) + k2Vi(t)2 + k3Vj(t)3 +
This model is simplified and is only valid for a memoryless system, but it serves to illustrate the mechanisms involved in intermodulation in a non-linear system.
If a two-tone signal is applied to the input, i.e. Vj = Acos(coit) + Bcos(ω
2t) then inserting the equation for V; into the Taylor series expansion above and using standard trigonometric identities, gives:
{A
2 + B
2)k
2 b
ko +
cos((ω1 -ω2)f)
This expression contains second order and third order distortion and intermodulation products. The third order intermodulation components at (ωι-2ω2) and (ω2-2ωι) and at higher frequencies are clearly evident. The second order intermodulation generates output products at second harmonic frequencies of the input signals, at sum and difference frequencies of the input tones. In addition, there is also a "direct detection" term at DC:
{A2 + B2)k2
2 The direct detection term is proportional to the sum-of-squares of the input tone amplitudes. This is independent of signal frequency, thus in a direct conversion receiver it can be extremely problematic, since any input signal that falls within the bandwidth of the front-end filtering will contribute to baseband interference, regardless of its frequency separation from the wanted signal. IM2 distortion (characterised by the IP2) is therefore very important in Gilbert mixers and direct conversion receivers in general.
For a better understanding of the present invention, reference will now be made, by way of example only, to the accompanying drawings in which:-
FIGURE 1 shows a block diagram of apparatus in accordance with one example of the invention, comprising an analogue pre-distorting circuit;
FIGURE 2 shows a block diagram of apparatus in accordance with another example of the invention, comprising an alternative circuit arrangement based upon adaptive correction; and FIGURE 3, to which reference has already been made, shows a block diagram of a Gilbert mixer.
In Figure 1, an RF input signal, X, is applied to a linear multiplier circuit 10 and to an adder circuit 12. The input signal is supplied to two inputs of the linear multiplier circuit 10, thereby generating a squared signal, X , which is then scaled by a predefined coefficient, K2, in a further multiplier circuit 14. The scaled, square signal K2X , is summed with the input signal, X, in the adder circuit 12 and the result of this summation is supplied to a mixer 16. By careful choice of coefficient, K2, a pre-distribution can be applied, substantially cancelling the IM2 products (i.e. the distortions caused by second-order non- linearity of the mixer 16).
The non-linear behaviour of the mixer may generally be described by the Taylor series:
Y=G!*X+G2*X2+G3*X3+ ...+Gn*Xn where X is the input signal, Y is the output signal (shifted in frequency), G\ is the conversion gain and G2 ... Gn are the non-linear terms.
In Figure 1, the choice of predefined coefficient is supplied from a look-up table (LUT) 18; the values stored in the LUT 18 being chosen in accordance with prevailing ambient conditions including the ambient temperature and the carrier frequency (or local oscillator) at which signals are transmitted.
The architecture of Figure 1 is achieved using analogue circuitry. The LUT 18 may however be implemented as a digital memory store.
In the embodiment of the invention shown in Figure 2, the coefficients, K2, are supplied by a correlator stage 20, rather than a look- up table, and the Figure 2 arrangement therefore requires less memory. The calculation of coefficients may conveniently be implemented in software running on a digital signal processing device (DSP).
The embodiment shown in Figure 2 displays a feed forward arrangement for adaptively compensating for IM2 distribution in the mixer wherein the input signal is supplied directly to a mixer 22 and the IM2 distortion caused by the mixer 22 is compensated with adaptive correction using the correlator stage 20.
As in Figure 1, the feed forward arrangement of Figure 2 utilises a linear multiplier 24, connected as before and thereby generating a squared signal, X2. Again, the squared signal is scaled by a coefficient, K2, in a multiplier 26 and the scaled and squared value, K2X , is applied as an input to an adder 28. The adder in the feed-forward arrangements sums the output Xm from the mixer 22 with the scaled squared signal, K2X2 and the summed signal passes through a low pass filter 30 to give a filtered output signal.
The coefficients, K2, are generated as follows by means of the correlator stage 20.
Firstly a sample of the squared signal, X2, is fed to a low pass filter 32. The low pass filter 32 is configured to filter out unwanted harmonics originating in the squaring multiplier; thereby discarding any signals outside the baseband bandwidth. The filtered, squared signal is supplied to a first input of a correlator 34. The filtered output signal from the low pass filter 30 is supplied to a second input of the correlator 34. The correlated signal, generated by the correlator 34, is applied as input into an integrator 36 thereby generating an appropriate coefficient, K , for supply to the scaling multiplier 26. In effect, coefficient K2 substantially cancels out the second order non-linear term G2 in the Taylor expansion of the output of the mixer.
The functions of the correlator 34 and/or the integrator 36 may be implemented in the form of software running on a digital signal processor (DSP). AID components, although not illustrated, are necessary at any interface between the digital and analogue domains. The value of K2 in the arrangement of Figure 1 is calculated to compensate for the IM2 products in a slightly different manner. In this case, the mixer 16 distorts an incoming signal (X + K2X ) giving an outgoing signal (GiX + G,K2X2 + G2X2 + other terms of 3rd order or greater). To cancel out second order terms in the Taylor expansion, GjK2 must be substantially equal to G2.