ES2332497B2 - MONOBIT ADAPTIVE SYSTEM AND METHOD OF CORRECTION OF PHASE AND EXTENSION BALANCES IN IQ DEMODULATORS. - Google Patents

Abstract

Adaptive monobit system and method of correction of phase and amplitude imbalances in IQ demodulators, comprising:

- a correction block (14) that corrects the samples (I, Q) using a multiplier (5) with coefficient β and a multiplier (6) with coefficient α, obtaining a corrected signal (S_ {corr});

- two complex filters (8, 9) that filter the corrected signal (S_ {corr}) obtaining filtered signals (S_ {f1}, S_ {f2});

- a coherence calculation block (10) that receives the filtered signals (S_ {f1}, S_ {f2}) and performs the estimation of spectral coherence;

- a coefficient update block (11) that it receives said estimate and, depending on it and the value of an update step \ mu, update the value of the α and β coefficients;

- a block modifying the step of update (12) responsible for modifying the step value of update \ mu based on the squared module value of the estimated spectral coherence.

Description

Adaptive monobit system and method of correction of phase and amplitude imbalances in demodulators IQ

Technical sector

The invention falls within the sector technological communication systems, more specifically in the receivers that I-Q demodulators use. The presented system is applicable in all those equipment and devices that need to perform an IQ demodulation for their functioning. Naming any of these systems: terminals public mobile communications, telephone terminals wireless, WIFI systems, WIMAX, receivers defined by program, among others.

Background of the invention

In international literature you can find a large number of articles in which the problem of correcting the imbalances of a demodulator I-Q In the document "A novel I / Q compensation scheme for a low-if receiver front-end "(Jérémie Chabloz and Christina Enz) [1], the proposed solution requires injection, at the entrance of the receiver, of a reference signal from which they get the correction coefficients, which complicates not only the circuitry but also the control of the receiver, since in a way periodic adjustment is necessary. In the document "Self tuned fully integrated high image rejection low IF receivers: Architecture and performance "(Zheng Yuanjin and Tear, Chin Boon Terry) [2], the solution provided is based on using the correlation between the signal in the image band and the signal in the desired band, so that the need for a external calibration signal. The main problem posed by this solution is that the correlation level depends on the signal level, so the adaptive algorithm used will have times of convergence dependent on the level of the input signal.

In the present invention, the estimation of the spectral coherence between the signal desired and the image [4]. The value of said spectral coherence is between 0 and 1, being independent of the level of signal (it will take the value 0 when there is no correlation between the signal and image, and the value 1 when the correlation is maximum). In addition, instead of estimating the value of spectral coherence in All receiver bandwidth, focus on bands around the image and desired signal through the use of two complex pass-band filters.

Additionally, another important improvement introduced by the present invention, is derived from the use of filters complexes whose coefficients can only be worth 1, -1, j or -j, [5], simplifying, in an important way, the circuit that makes said filtered, since it is only necessary to perform addition and subtraction.

Bibliographic references

[1] Jérémie Chabloz and Christina Enz , "A novel I / Q compensation scheme for a low-if receiver front-end", Proceedings of the 2004 international symposium on Circuits and Systems, ISCAS'04, Volume IV, pp: 453- 456

[2] Zheng Yuanjin and Tear, Chin Boon Terry , "Self tuned fully integrated high image rejection low IF receivers: Architecture and performance", Proceedings of the 2003 international symposium on Circuits and Systems, ISCAS'03, Volume II, pp: 165 -168.

[3] M. Valkama & M. Renfors , "Advanced DSP for I / Q imbalance compensation in a low-IF receiver", IEEE International Conference on Communications , ICC 2000 , Volumne II, pp: 768-772

[4] G. Clifford Carter , Charter H. Knapp & Albert H. Nuttall . "Estimation of the Magnitude-Squared Coherence Function via Overlapped Fast Fourier Transform processing", IEEE Transactions on Audio and Electroacoustics , Vol. AU-21, pp: 337-344, August 1973

[5] DS Pok , CH Chen , JJ Schamus , CT Montgomery , JBY Tsui , "Chip Design for Monobit Receiver", IEEE Transactions on Microwave Theory and Tehniques , Vol. 12, pp: 2283-2295, December 1997.

Description of the invention

The present invention relates to a system adaptive monobit correction of imbalances between branches I-Q of a demodulator by means of a cell correction whose coefficients are calculated continuously and adaptively

The updating of the coefficients is based on the estimation of the spectral coherence between the signal desired and the image signal, for which it is necessary to perform a filtered, with two complex filters, centered one by one frequency of the desired signal and the other at the frequency of the image signal Since IQ demodulation is basically a frequency translation, if the desired signal is transferred to positive frequencies (around FI), the image signal will remain moved to negative frequencies (around -FI) and vice versa, with what filters should be tuned, one focused on FI and the other one focused on -FI.

The coefficient update responds to the following expression (1):

one

where C xy is the spectral coherence, µ is the update step (the change allowed for α and β from one iteration to the next depending on the spectral coherence) and k indicates the iteration of the algorithm.

The advantage obtained by using consistency spectral is that its value is independent of the signal level of input, so it will not affect the speed of convergence or The stability of the algorithm.

Spectral coherence can be estimated using the following expression (2)

2

where \ gamma_ {x} ( f ) is the estimation of the power in a certain bandwidth around the desired signal γ_ {y} ( f ) is the estimation of the power in a certain bandwidth around the signal image and N is the number of samples. Bandwidths used for the estimates of \ {x} gamma_ (f) and \ gamma_ {y} (f) be secured by filters tuned to the frequency of the desired signal and the image signal frequency.

Spectral coherence will take a zero value. when there is no correlation between the desired signals and image (when the imbalance has been corrected or does not exist), and a value greater than zero, but generally less than 1, when there is a high correlation between the desired signals and image (when the imbalance has not been corrected).

The most important advance of the present invention is that the coefficients of complex filters are limited to unit values, of the form 1, -j, -1 and -j (where j represents the complex number). In this way the necessary filtering for the calculation of spectral coherence it is very simplified, since the use of the coefficients converts all addition and subtraction operations.

Brief description of the drawings

Then it goes on to describe very brief a series of drawings that help to better understand the invention and that expressly relate to an embodiment of said invention presented as a non-limiting example of is.

Figure 1 shows a diagram of the different IQ demodulation stages, including the system subject to the invention.

Description of a preferred embodiment

As shown in Figure 1 (which represents the different stages of IQ demodulation in a receiver including the adaptive system monobit 15 object of the present invention, which can be executed in the same DSP, FPGA or microcontroller that performs the subsequent signal processing complex IQ), radio frequency signal, input signal s (t), after being moved in frequency with the help of frequency converters or conversion means of frequency 1, multiplying the signal s (t) by a sine signal and by a cosine signal, and filtered by the low pass filters or low-pass filter media 2, obtaining the signal phase I (t) and the quadrature signal Q (t), is directed to analog to digital converters or conversion media analog-digital 3 at whose output you will get the samples that will form the complex signal, whose components are I (n) and Q (n), to be processed by the receiver after the imbalance of the demodulator has been corrected. The samples thus obtained they must be routed to a correction block 14, in concrete to a correction cell 4 formed by multipliers, a first multiplier 5 and a second multiplier 6, whose coefficients α and β will vary over time to correct the imbalance of the IQ demodulator, 1. This cell of correction 4 implements the following expression (3):

3

where I (n) and Q (n) are the samples without the corrected imbalance, \ alpha_ {k} and \ beta_ {k} are the correction coefficients in the k-th step of the iterative algorithm and I_ {c} (n ) and Q_ {c} (n) are the samples with the corrected imbalance once the algorithm has converged. The multiplier 5 is in charge of carrying out the operations in which the coefficient \ beta_ {k} intervenes, while the multiplier 6 is in charge of performing the operations in which the coefficient \ alpha_ {k} intervenes with the output of this cell of correction 4 a complex signal S_ {corr} is formed in a complex signal generation block 7 (within the correction block 14) that will be filtered by two complex filters 8 and 9 tuned at the frequency of the desired signal and the another at the frequency of the image signal, as already mentioned.

With the output of the filters is done in a coherence calculation block 10 consistency calculation spectral by averaging N samples. Made the split and square root operations, involved in the calculation of the spectral coherence estimate, the result it is sent both to a coefficient update block 11, as a modification block of update step 12. The coefficient update block 11 ensures that the actual part of the result is used to modify the value of the coefficient? according to the update step? and the part Imaginary be used to modify the coefficient value ? analogously to how it is done with the? and according to the Equation (1).

The step modification block of update 12 will be responsible for modifying the value of \ mu in function of the value of the squared module of the spectral coherence (which can be easily calculated from its real part and its imaginary part) so that at a lower value of the squared module the lower consistency will be the value of the update step, allowing, on the one hand, to reach a rejection value to the image better, and on the other get a great speed of convergence against sudden changes in the elements of demodulator

The preferred embodiment of the presented system It is supported by a reception system with IQ demodulators. The signal demodulated and filtered is sent to two analog converters to digital with the relevant number of bits to improve the receiver parameters

The correction cell 4 applies the α and β coefficients; in this way the signal is obtained corrected complex S_ {corr} as I_ {c} (n) + jQ_ {c} (n).

Said corrected signal S_ {corr} is filtered by two complex filters tuned to the signal frequencies desired and image. The number of coefficients of said filters determines the bandwidth over which the estimate will be made of spectral coherence. Additionally those mentioned coefficients are quantified with a single bit.

Complex filters are characterized basically because the module of its transfer function in frequency has no symmetry with respect to zero frequency, such and as with real filters, which allows designing filters band pass that select only part of the positive frequencies or only part of the negative frequencies.

         \ vskip1.000000 \ baselineskip
      

In this environment, the filters used in the present invention are FIR filters (impulse response filters finite) whose coefficients will be complex. Specifically the filter FIR used is a rectangular window with the answer in frequency transferred to the frequency of interest, desired or image. The coefficients of said filter then respond to:

4

where \ omega is the pulsation at that the frequency response of the filter has been transferred and M the number of stages that make up the FIR filter and that will determine the bandwidth same.

         \ vskip1.000000 \ baselineskip
      

It is about these coefficients that are you can apply the quantification, or monobit coding, attending to the following nonlinear function (5):

5

with which all products are have become sums

The filter output is used to perform the estimation of the powers in the signal bands desired and image. After averaging a set N of estimates, the spectral coherence calculation is carried out. This calculation involves performing two operations computationally expensive, such as division and root square, however these operations are not critical given realize that they have to be done at high time intervals.

The result of the calculation of the spectral coherence is complex, using the real part to update the α coefficient and the imaginary part to update the β coefficient. The update step used, µ, must have a value large enough to ensure rapid convergence and at the same time small enough to ensure the best possible correction of amplitude and phase imbalances. These contradictory requirements impel to choose an approach in which the mentioned step of update is variable and dependent on the degree of correction reached. Thus, when the imbalance is high, the factor [mu] takes on a large value and as the imbalance is corrected, said factor [mu] gradually decreases its value. For this variation of the update step you can use a strategy by sections on the filtered value of the squared module of coherence, so that if the squared module of the spectral coherence is in a certain section, the update step would take the value assigned to said section. As an example, the range 0 to 1 (values between which spectral coherence will always be found) is divided into P intervals not necessarily regular, C_ {p}, (with p = 0..P-1) limited by boundary values m_ {p} such that: 0 < m 0 {<... < m _ {p} <... < m _ {p-1} <1, thus the value of the update step would be determined based on the squared module of the spectral coherence as;

6

It must be fulfilled that: 0 <\ mu_ {0} < ... <\ mu_ {p} <... <\ mu_ {p-1} < \ mu_ {p}.

Claims (10)

1. Monobit adaptive system for correction of phase and amplitude imbalances in IQ demodulators, characterized in that said system (15) comprises: - a correction block (14) responsible for receive and correct samples I and Q, obtained by applying a analog-to-digital conversion stage over phase I (t) and quadrature Q (t) signals, by means of a first multiplier (5) with β coefficient and a second multiplier (6) with coefficient α, obtaining a signal corrected (S_ {corr}); - a first (8) and a second (9) filters complex, responsible for filtering the corrected signal (S_ {corr}) to get a first S_ {f1}, and a second S_ {f2} signals filtered, the coefficients of said complex filters being (8.9) encoded with a bit, which is worth any one of the following values: 1, -1, j, -j, where j is the complex number; - a coherence calculation block (10) responsible for receiving the filtered signals (S_ {f1}, S_ {f2}) and from these signals, calculate the estimate of spectral coherence; - a coefficient update block (11) responsible for receiving the spectral coherence estimate and, based on said estimate and the value of a step of update \ mu, update the value of the coefficients α and β; - a block modifying the step of update (12) responsible for modifying the step value of update \ mu based on the squared module value of the estimated spectral coherence, being the update step \ mu the maximum allowable change in? and? iteration? to iteration based on spectral coherence.

           \ vskip1.000000 \ baselineskip
        

2. System according to claim 1, characterized in that the correction block (14) comprises a complex signal generation block (7) responsible for generating a complex signal, corresponding to the corrected signal S_ {corr}. 3. System according to any of the preceding claims, characterized in that the complex filters (8,9) are of the pass-band type. System according to any one of the preceding claims, characterized in that the complex filters (8,9) are tuned to the frequency of the desired signal and the other to the frequency of the image signal. System according to any of the preceding claims, characterized in that the coherence calculation block (10) is configured to perform the calculation of the spectral coherence estimate by accumulating N samples. System according to any of the preceding claims, characterized in that the coefficient update block (11) is configured to update the value of the coefficients α and β depending on the value of the real and imaginary part, respectively, of the estimated spectral coherence. System according to any of the preceding claims, characterized in that the coefficient update block (11) is configured to update the value of the coefficients α and β, depending on the value of the spectral coherence, adaptively. System according to any of the preceding claims, characterized in that the update step µ is variable according to a strategy by sections on the filtered value of the squared module of the coherence. 9. Monobit adaptive method of correction of phase and amplitude imbalances in IQ demodulators, characterized in that it comprises: - receive and correct samples I and Q, obtained by applying a conversion stage analog-digital over phase signals I (t) and quadrature Q (t), using a first multiplier (5) with β coefficient and a second multiplier (6) with coefficient α, obtaining a corrected signal (S_ {corr}); - filter the corrected signal (S_ {corr}) to get a first S_ {f1}, and a second S_ {f2} signals filtered; - receive the filtered signals (S_ {f1}, S_ {f2}) and perform from these signals the calculation of the estimation of spectral coherence; - receive the consistency estimate spectral and, depending on that estimate and the value of a step update \ mu, update the value of the coefficients α and β. - modify the update step value µ depending on the value of the squared coherence module estimated spectral, the update step being the maximum allowable change in? and? from iteration to iteration in Spectral coherence function. 10. IQ demodulator with imbalance correction phase and amplitude, comprising: - frequency conversion means (1) and low-pass filter media (2) configured to obtain the phase signal from the input signal s (t) I (t) and the quadrature signal Q (t); - conversion media analog-digital (3) configured to obtain the samples I and Q; Y - the adaptive monobit correction system of phase and amplitude imbalances according to any of the claims 1-8.