Research Journal of Recent Sciences ________________________________________________ ISSN 2277-2502Vol. 2(9), 95-100, September (2013) Res.J.Recent Sci. International Science Congress Association 95 Review Paper Comparative Survey on Time Interleaved Analog to Digital Converter Mismatches Compensation TechniquesHafiz Allah Nawaz, Abida Sharif2 and Muhammad SharifNational University of Computer and Emerging Sciences, Islamabad, PAKISTAN COMSATS Institute of Information Technology, Islamabad, PAKISTAN COMSATS Institute of Information Technology, Wah Cantt, PAKISTANAvailable online at: www.isca.in Received 4th December 2012, revised 28th February 2013, accepted 7th June 2013 AbstractThis survey investigates recent “time interleaved analog to digital converters (TI-ADCs)” mismatches correction techniques. It includes each technique performance evaluation on the basis of SNDR, SFDR, Frequency and Input signal. Many algorithms have been developed over the past two decades, mainly emphasizing on the digital calibration of TI-ADCs mismatches. The aim of this survey is to review a number of more promising techniques which are offering efficient mismatch corrections. Keywords: Inherent, Compensate, Frequency Mismatch, SFDR, SINDR Introduction Today communication systems are going towards digital form to offer more correctness and flexibility. For this purpose communication systems need high speed and larger bandwidth which can be achieved with TI-ADCs1-3. TI-ADCs means “time-interleaved analog to digital converters” which utilize parallel low rate ADCs to increase the sampling rate4-5. Figure 1 shows M channel TI-ADCs where a low rate channel of ADCs operates with an fs/M sampling frequency6-7. TI-ADCs generate output each time and each channel periodically has a sampling time MTs where as overall system (TI-ADCs) sampling time is Ts.. X (t) is the input going to all ADCs and Y(nT) is the output after combining the low rate ADCS through mux. Figure-1 channel time interleaved ADCs TI-ADCs comprise of parallel low rate channels which take analog singles at input and generate output with different parameters among the channels. These parameters are gain, offset, trimmingsand bandwidth which generate mismatch errors. Mismatches occur among the channels components generating mismatch error in single channel. The mismatch error is different at different channels. Gain mismatch is difference among the slope of ideal transfer curve and transfer curve with gain errors. Similarly offset, bandwidth and timing mismatches are difference among the offset, bandwidth and timing errors respectively10. Also there are a number of techniques utilized for minimizing mismatch errors to achieve compensated output. The compensated output includes improved SINDR which means signal to noise-and-distortion ratio and improved SFDR which means spurious free dynamic range. In this paper different techniques are reviewed under SNDR, SFDR, frequency and input signal. Mismatch Compensation Techniques There are several techniques to tackle mismatch errors classified as according to figure 2. TI-ADCs primarily have two categories; static and dynamic10. Static mismatch includes offset and gain mismatch which means fixed mismatch. Similarly dynamic mismatch includes timing mismatch which means variable mismatch with time. Timing mismatch techniques are poly-phase, transfer characteristic tuning, LMS, Multichannel filters, background, blind, adaptive blind, on blind etc. Gain mismatch correction techniques include adaptive blind, sigma delta and on blind. Similarly offset mismatches are sigma delta and nonwhite etc. Among the mismatches as timing, gain and offset, timing mismatches are crucial because it decreases the performance more as compared to other mismatches. Timing Mismatch Compensation Techniques The time-interleaved analog to digital converters performance primarily suffers through timing mismatches which are Research Journal of Recent Sciences ______________________________________________________________ ISSN 2277-2502Vol. 2(9), 95-100, September (2013) Res. J. Recent Sci. International Science Congress Association 96 dominant at high frequencies11. Different techniques exist to compensate these timing mismatches. A comparative evaluation of each technique is given in Table 1 under said parameters as SNDR and SFDR. To deal with timing mismatch errors, the least mean squarebasedmismatch compensation approach is presented in which, an input signal isoversampled and band is limited with 6 × 10 samples. In the technique the band of input signal [-0.7, 0.7] with least mean square LMS algorithm achieves SINDR of 59dB. The timing mismatch is identified by energy minimization in mismatch band and based on the identification, LMS compensate the output12. In mixed-signal scheme, the timing mismatch is identified with timing error detection subsystem. The detection subsystem is controlled by the voltage bootstrapped switch. The system takes 200MSps and after compensation the achieved SINDR is 77.4dB and SFDR is 84.4dB13. Another approach is demonstrated known as background calibration for timing mismatch error compensation. The approach utilizes digital interpolation to compensate samples at the output and as a result spurious free dynamic range SFDR is improved from 20dB to 60dB. For ideal system, background calibration technique gets SNR 85dB and SFDR 120dB14. Similarly in blindcalibration technique for timing mismatch error correction, take three sinusoidal components at 0.1, 0.45and 0.75. The method improves 12dB at the output and finally gets the SINDR 46.1dB16. Another approach for minimizing timing mismatches includes finite impulse response filter with one linear-phase cascaded with one multiplier for assisting magnitude response minimization. In the technique narrow and wide band signals have become computationally cheap operations and multiplier changes the signal sign through short tap size FIR filter. The technique is developed for wide band signal and gets the signal to noise-and-distortion ratio 63.23dB after magnitude compensation. Also the approach achieves SINDR of 56.71dB17. Blind calibration algorithm minimizes the timing mismatches adaptively in TI-ADCs converters whereas TI-ADCs poly-phase model requires slight oversampling to parameterize the timing mismatches15,18. Compensating timing and bandwidth mismatch errors, a multichannel filtering based approach is presented. In the approach each sub-channel is down sampled for analog to digital converter when filtered outputs are summed and each channel comprises of coefficients for every one filter bank. Then the weighted least square method calculates the inverse for optimization and input signal is taken at different frequencies19. The technique known as Lagrange polynomial interpolation includes the compensation method comprised of three steps. Initially mismatch parameters are compensated, secondly finite impulse response filtering operation is performed for all ADCs and lastly multiplexer is used at the output. The presented solution is computationally efficient and non sensitive to fluctuation for timing mismatch. The approach at 70MHz input frequency produces 30dB improvement in spurious free dynamic range. Similarly multi rate filter banks compensation method enhances 19dB of SFDR. Also at 40 points for interpolation, the approach achieves 94dB at 320MHz20. Furthermore, another approach which utilizes digital signal processing has been used for sample time error detection and correction. The approach utilizes finite impulse response filters with 29-taps at input frequency 0.1fs and achieves SINDR signal to noise-and-distortion ratio more than 60dB21. Similarly the approach known as back ground using random data includes 800 MS/s for the input at 350MHz frequency and enhances the spurious free dynamic range to 58.1 dB and SNDR 50.1 dB. Also at 5 MHz frequency, the achieved SNDR is 59.6 dB22. Figure-2 Classification of different methodologies to compensate TI-ADCs mismatches Research Journal of Recent Sciences ______________________________________________________________ ISSN 2277-2502Vol. 2(9), 95-100, September (2013) Res. J. Recent Sci. International Science Congress Association 97 Moreover to compensate timing mismatch errors reordering the number of channels based approach is presented. In the technique the sorting algorithm achieves 61dB SFDR at 8-channels, 31dB SFDR at 128-channels and 10dB SINDR in ENOB at 128-channels23. In low resolution and time varying filter approach, the adaptive structure minimizes the mismatches through time varying finite impulse response filters and an extra low-resolution analog to digital converter. Compensating linear frequency response mismatch errors, a structural based approach is presented. In the approach, the weights of least-mean square (LMS) algorithms are adopted by the time varying filter, input signal at 0.8/2 with random amplitude and phase22. The approach achieves SNR of 57.2dB. Similarly utilizing the frequencydomain, a timing mismatch compensation method is used in the oversampled input signals to identify timing mismatch and it depends on the relation of frequency signals between the continuous input and after sampled output. The approach achieves white noise band limited input signal b = 7/10s with 16384 samples and larger signal to noise-and-distortion ratio with 11 effective number of bits24. Offset Mismatch Compensation Techniques In table 2 different offset mismatch compensation techniques are comparatively analyzed according to SINDR, SFDR, frequency and input signal. In data chopping, theinherent offset mismatches are compensated through nonwhite data chopping approach. The nonwhite sequences generally allow faster compensation of offset and better spectral shaping. Average offset compensation has been developed in many ADCs realizations but the best estimation is achieved when the transformed signal is assumed to be a Gaussian process. The technique achieves SNR of 29.82 db with white chopping and 33.85 with low pass chopping25. Random chopper samplingoffset compensation method is based on PRBS which takes the input and converts it to noise. The batch size is controlled by the randomization process which is used for offset compensation. The approach achieves SFDR of 72dB and SINDR of 59dB26. Similarly in background calibrationapproach, to overcome gain and offset mismatches in time-interleaved ADCs, 2nd order structure is used. Simulation result shows that the approach achieves 12dB better as compared to conventional modulator at the same frequency27. In the technique the offset and gain mismatches are compensated through analog and monolithic background calibration by extra channel and mixed signal integrators. Also with the approach the calibrated signal has peak SINDR of 58dB 28. Channel and Gain mismatches compensation techniques To compensate channel and gain mismatches, there are a number of techniques and are comparatively elaborated in table 3. Compensating gain and timing mismatch errors collectively an approach is presented which utilizes digital blind identification structure to estimate gain and timing mismatches adaptively. In the approach “filtered error least mean square (FxLMS)” algorithm is applied which uses a single high-pass filter. The approach achieves elimination of spurious images from the output spectrum of TI-ADC and enhances the “signal to noise ratio (SNR)” to 69.3 dB with 222 samples29. Similarly in another approach called input band limited “white Gaussian noise (WGN)” signal, identified gain and timing mismatches in a TI-ADC 4-channel shows significant improvement in the SNR. Assuming a “wide-sense stationary (WSS)” input, yielding a WSS output and online calibration can be used without disturbance or any committed signal. In the approach wide and sine input signals reduced the mismatches by 20 dB and 50 dB respectively at 100000 samples30. Moreover, expectation maximize algorithm compensates the gain and time offsets31. Table-1 Timing mismatches compensation techniques comparison Methodology SINDR[dB] SFDR[dB] Frequency Input signal LMS 59 00 [ -0.7, 0.7] 6 × 10 4 Samples Mixed signal 77.4 84.4 00 200MSps Background SNR=85dB Improved 20 to 60 00 00 Blind 46.1 00 0.1, 0.45 and 0.75 00 Finite impulse response filter 63.23 00 00 00 Poly phase 00 improved 20 [-0.875, 0.875] 00 Lagrange Polynomial Interpolation 00 94 32000 Random Data more than 60 00 0.1fs 00 Back Ground using Random Data 50.1 58.1 350MHz 800 MS/s Spectral Shaping 10 in ENOB 61 00 00 Low resolution and Time varying filter SNR=57.2 00 0.8 s /2 2 22 samples Research Journal of Recent Sciences ______________________________________________________________ ISSN 2277-2502Vol. 2(9), 95-100, September (2013) Res. J. Recent Sci. International Science Congress Association 98 Table-2 Offset mismatches compensation techniques comparison Methodology SINDR [dB] SFDR [dB] Frequency Input signal Data Chopping SNR=29.82 00 00 1000 cycle Random Chopper Sampling 59 72 00 00 Background calibration 58 00 00 00 Table-3 Channel mismatches compensation techniques comparison Methodology SINDR [dB] SFDR [dB] Frequency MHz Input signal Orthogonal digital calibration SNR = 49.04 53.94 00 00 Comprehensive digital cal 80 50 Up to 175 00 Randomization Strategy 00 59.8 to 72 00 00 Mixed-signal 71.7 85.8 00 500 sample Adaptive Blind SNR=66 00 00 131072 sample Digital Filter Bank 71.8963 92.0058 75.3 100 MS/s An approach known as orthogonal digital calibration compensates timing, offset and gain mismatches. It is based on code division multiple access (CDMA) technique. In the technique the calibration can occur at run time (online) and the FFT spectrum of calibrated output shows signal to noise ratio of 49.04dB and spurious free dynamic range of 53.94dB32. In comprehensive digital calibration, the approach utilizes transfer function of channel which includes all linear errors between analog input signal x(t), digital output signal y(n) and series of sinusoidal signals used for measuring each channel linear errors. A parallel finite-impulse response (FIR) filter also provides frequency-dependent mismatch correction weighted by least squares principle. The technique achieves spurious free dynamic range of 80dB at the input frequency of 175MHz33. An approach considered as randomization strategy and the compensation technique investigated the channel non-linearity (DNL) mismatches and calibrated the upper frequency band due to baseband frequency disturbance. The approach achieves SFDR 59.8dB to 72dB34. Similarly the mixed-signal clock-skew calibration method is developed to compensate the linearity of time interleaved analog to digital converters. The approach achieves compensated output signal with spurious free dynamic range of 73dB35 at the 19.9MHz input frequency. Another approach achieves SFDR of 85.8dB and SINDR of 71.7dB over the 500 samples while using hybrid filter bank36. Similarly phase skew compensation analysis is used in the wideband input signals to compensate multiple mismatches37. A technique which utilizesleast mean square algorithm overcomes the offset mismatch, gain mismatch, sample-time mismatch and bandwidth mismatches among TI-ADCs simultaneously. The presented compensation method is used for the baseband and at high speed receiver. All mismatches are compensated by the blocks of digital signal processing. The approach achieves with MSE, -37.2dB Gain, -36.8dB offset, -36.6dB sample time and -36.7dB bandwidth38. Similarly in adaptive blind technique,least mean square algorithm has been used blindly to compensate the timing and gain mismatches. The presented compensation technique is based on arbitrary number of channels and oversampled input signal39. The technique achieves signal to noise ratio of 66dB40 for 131072 samples at the input. In digital filter bank approach, theweighted least square coefficients are used for the compensation of time and gain mismatches errors carried by the poly-phase structure. Each channel operates at 100 MS/s with filter bank order of 40. The approach achieves calibrated output at input frequency of 75.3MHz with 71.8963dB SINDR and 92.0058 SFDR41. ConclusionIn this paper a review of the techniques assisting different mismatches compensations in TI-ADCs is presented. The techniques are evaluated on the basis of how much signal to noise ratio SNDR and SFDR are achieved. Form the literature it can be seen that there are a lot of promising techniques for compensating timing, gain, offset and channel mismatches of TI-ADCs. Some techniques are compensating single mismatch while some are compensating multiple mismatches. In the paper the techniques are comparatively analyzed under said parameters as SNDR and SFDR. Although a number of promising TI-ADCs mismatches compensation techniques exist but still there is a need of better compensation techniques for TI-ADCs which could give maximum compensated output and to tackle any single mismatch or multiple mismatches of TI-ADCs. References 1.Christian V. and Heinz K., Behavioral Modeling of Time-Interleaved ADCs using MATLAB, Proceedings of the Austrochip, 45- 48 (2003) 2.K. E.S., A A. and M S., New Sampling Method to Improve the SFDR of Time Interleaved ADCs, circuits and systems, ISCAS, , I -833 (2003) 3.Shahzad S. and Christian V., Adaptive Compensation of Frequency Response Mismatches in High-Resolution Time-Interleaved ADCs using a Low-Resolution ADC and Research Journal of Recent Sciences ______________________________________________________________ ISSN 2277-2502Vol. 2(9), 95-100, September (2013) Res. J. Recent Sci. International Science Congress Association 99 a Time-Varying Filter, IEEE Int. Symp. Circuits Sys, 561 – 564 (2010) 4.Micheal S. and Ronan Farrell., Methodology for Mismatch Reduction in Time-Interleaved ADCs, Circuit theory and design, 352 - 355 (2007)5.Christian V. and Hakan J., Time-Interleaved Analog-To- Digital Converters: Status and Future Directions, ISCAS, 4, 3389 (2006)6.Tsung H. T., Paul. J. H. and Stephen. H. L, Bandwidth Mismatch and Its Correction in Time-Interleaved Analog- to-Digital Converters, circuits and systems, 53, 1133 – 1137 2006) 7.Antonio P. and Sanjit K. M., Analysis of Mismatch Effects Among A/D Converters in a Time-Interleaved Waveform Digitizer, IEEE Transactions on instrumentation and measurement, 40, 831 – 835 (1991) 8.Christian V., The Impact of Combined Channel Mismatch Effects in Time-Interleaved ADCs, IEEE transactions on instrumentation and measurement, 54, 415 – 427 (2005) 9.David. G. N., Time-Interleaved Analog-to-Digital Converters, IEEE Custom Integrated Circuits Conference, 289 – 296 (2008) 10.Christian V., Modeling, Identification, and Compensation of Channel Mismatch Errors in Time-Interleaved Analog-to-Digital Converters, (2005) 11.Ralf v. O., Timing Correction of Time-Interleaved ADCs, Graduation Symposium, (2009) 12.Shahzad. S and Christian V., LMS-Based Identification and Compensation timing Mismatches in a Two-Channel Time- Interleaved Analog-to-Digital Converter, Norchip, 1 - 4 (20007) 13.P Z., F Y., B Y., L. L. and J. R., Mixed-signal calibration of sample-time error in time interleaved ADCs, ELECTRONICS LETTERS, 47, 533 - 535 (2011)14.Huawen J. and Edward K. F. L., A Digital-Background Calibration Technique for Minimizing Timing-Error Effects in Time-Interleaved ADC’s, IEEE Transactions on circuits and systems, 47, 603 - 613 (2000) 15.Oscar E. A. and Venugopal G., Background Calibration of Interleaved Analog to Digital Converters for High-Speed Communications Using Interleaved Timing Recovery Techniques, circuits and systems, , 1390 - 1393 (2005) 16.Steven H. and Bernard. C. L., Blind Calibration of Timing Offsets for Four-Channel Time-Interleaved ADCs, IEEE Transactions on circuits and systems , 54, 863 - 876 (2007)17.Stefan M. and Christian V., A Compensation Method for Magnitude Response Mismatches in Two-channel Time- interleaved Analog-to-Digital Converters, 712 - 715(2006) 18.Patrick S., Bernard C. L and Paul. J. H., A Parametric Polyphase Domain Approach to Blind Calibration of Timing Mismatches for M-Channel Time-Interleaved ADCs, circuits and systems, ISCAS, 4053 - 4056 (2010) 19.Yong C. L.,Fellow., Yue X. Z., Jun W. L. and shing C. C., Time-Interleaved Analog-to-Digital-Converter Compensation Using Multichannel Filters, IEEE Transactions on circuits and systems, 56, 2234 - 2247(2009) 20.Yue X. Z., Shang L. Z., Yong C. L., Fellow and Xiao C., Timing Mismatch Compensation in Time-Interleaved ADCs Based on Multichannel Lagrange Polynomial Interpolation, IEEE Transactions on instrumentation and measurement, 60, 1123 - 1131 (2011) 21.Shafiq M. J., Diahong F., Mahendra P. S., Paul J. H. and Stefan. H. L., Calibration of Sample-Time Error in a Two Channel Time-Interleaved Analog-to-Digital Converter, IEEE Transactions on circuits and systems, 51 130 - 139 (2004) 22.Afshin H.and Kenneth W. M., A Background Sample-Time Error Calibration Technique Using Random Data for Wide-Band High-Resolution Time-Interleaved ADCs, IEEE Transactions on circuits and systems , 55 234 - 238 (2008) 23.Christian V., Dieter D. and Gernot K., Spectral Shaping of Timing Mismatches in Time-Interleaved Analog-to-Digital Converters, circuits and systems , ISCAS, , 1394 - 1397 2005) 24.Christian V., A Frequency Domain Method for Blind Identification of Timing Mismatches in Time-Interleaved ADCs, IEEE Norchip, 45 - 48 (2006)25.Wen J. S and Bo Y., Calibration of Gain and Time Mismatches for Time-Interleaved ADCs Based on Digital Filter Bank, International Conference on Mechatronic Science, 1836 - 1839 (2011) 26.Stefano V., Giampaolo C., Riccardo R. and Gianluca., S, Algorithmic ADC Offset Compensation by Nonwhite Data Chopping: System Model and Basic Theoretical Results, IEEE transactions on circuits and systems, 55 1615 - 1627 (2008) 27.Ramin K. P., Lysander B. L.and David A. J. Time-Interleaved Oversampling A/D Converters: Theory and Practice IEEE Transactions on circuits and systems, Analog and Digital Signal Processing, 44 634 - 645 (1997) 28.Kenneth C. D., Daihong F., Stephan H. L. and Paul J. H., An Analog Background Calibration Technique for TimeInterleaved Analog-to-Digital Converters, IEEE Journal of solid state circuits, 33, 1912 - 1919 (1998)29.Shahzad S. Christian V., On Blind Identification of Gain and Timing Mismatches in Time Interleaved Analog-to-Digital Converters, 33rd International Conference on Research Journal of Recent Sciences ______________________________________________________________ ISSN 2277-2502Vol. 2(9), 95-100, September (2013) Res. J. Recent Sci. International Science Congress Association 100 Telecommunications and Signal Processing, TSP, 151-155 2010) 30.Munkyo S., Mark J. W. R. and Upamanyu M., Blind Correction of Gain and Timing Mismatches for a Two-Channel Time-Interleaved Analog-to-Digital Converter, circuits and systems, ISCAS, (2006) 31.Vijay D. and Gregory W., Signal Recovery in Time –Interleaved Analog -To-Digital Converters, ICASSP, (2004) 32.G F., B L. G., L B., M J., D D. and P. C., Orthogonal Correction Implementation for Time Interleaved Analog -To-Digital Converters: Real Time Application, EUSIPCO, (2008) 33.Munkyo S., Mark J. W. R., Upamanyu. M., Comprehensive Digital Correction of Mismatch Errors for a 400-Msamples/s 80-dB SFDR Time-Interleaved Analog-to-Digital Converter, IEEE Transactions on microwave theory and technique, , 1072 - 1082 (2005) 34.Christian V. and Gernot K., Analysis and Compensation of Non-Linearity Mismatches in Time Interleaved ADC Arrays, ISCAS, , I - 593-6 (2004) 35.David C., Karim B. K., Jean F. N. and Patrick L., Mixed-Signal Clock-Skew Calibration Technique for Time-Interleaved ADCs, IEEE Transactions on circuits and systems , 55, 3676 - 3687 (2008) 36.Yu S. L. and Qi A., Calibration of Time-Skew Error in a Channel Time-Interleaved Analog-to-Digital Converter, ICEMI, 3-968 - 3-971 (2005) 37.Manar E. C. and Boris M. General Analysis on the Impact of Phase-Skew in Time Interleaved ADCs IEEE Transactions on circuits and systems, 56, 902 - 910 (2009) 38.Tsung H. T. Paul J. H.and Stephen H. L., Correction of Mismatches in a Time-Interleaved Analog-to-Digital Converter in an Adaptively Equalized Digital Communication Receiver IEEE Transactions on circuits and systems , 56, 307 - 319 (2009) 39.Christian V., Comprehensive Error Analysis of Combined Channel Mismatch Effects in Time-Interleaved ADCs, IMTC2003 - instrumentation and Measurement Technology Conference VAIL, CO, USA, 54 415 – 427 (2003) 40.David. G. N., Time-Interleaved Analog-to-Digital Converters, IEEE Custom Integrated Circuits Conference, 289 – 296 (2008) 41.Wen J. S and Bo Y., Calibration of Gain and Time Mismatches for Time-Interleaved ADCs Based on Digital Filter Bank, International Conference on Mechatronic Science, 1836 - 1839 (2011)