ANNUAL JOURNAL OF ELECTRONICS, 2013, ISSN 1314-0078 Design of an FPGA based TV-tuner test bench using MFIR structures Jean-Jacques Vandenbussche, Peter Lee and Joan Peuteman Abstract - The paper shows how Multiplicative Finite Impulse Response (MFIR) filter structures can be used to implement digital linear phase low pass filters with a very narrow transition band and a high stop band attenuation in state of the art low-end FPGA technology. A digital Intermediate Frequency (IF) filter for a TV-tuner test bench is used as a reference design. The design procedure and the hardware requirements are presented. The paper concludes that the MFIR filter structures allow the design of demanding filters, while traditional linear phase FIR filters with equivalent characteristics would require significantly more hardware resources. Keywords – MFIR, FPGA, TV-tuner, IF filter I. INTRODUCTION In analog TV transmissions the signals at the front-end of a TV receiver typically consist of a desired signal accompanied by strong neighbor channels and interfering signals [1] [2]. The received signal is down converted to a standardized Intermediate Frequency (IF), but before demodulation can take place, the neighbor channels must be strongly attenuated to exclude any interference. In analog TV sets, high performance surface acoustic wave IF filters developed by specialized suppliers were used to achieve this high attenuation. Although high frequency Analog to Digital Convertors (ADC) exist, present state of the art digital TV-receivers still down convert the received signal in the analog domain. The resulting IF frequency is then sampled and ready for further digital signal processing [2] [3]. Before digital demodulation can take place, possible interference from neighboring channels must be avoided using high quality band pass filtering. This band pass filtering is realized in the digital domain, implying TV designers can design the IF filter without the need for specialized filter technology. Moreover, the frequency where demodulation can take place is now programmable. Unfortunately, in cable TV reception, the neighbor channel of a desired digital TV channel can be an analog TV channel with an amplitude which is 40 dB higher than the desired signal [4]. This puts a very stringent requirement on the digital IF filter. In addition, digital modulation schemes (QAM, QPSK etc) carry the information in the phase of the carrier, requiring linear phase filters to avoid inter symbol interference (ISI). In state of the art TV receivers, a low IF of 4 MHz is chosen resulting in an IF filter with a low pass characteristic and a required attenuation of at least 75 dB immediately outside the pass band [4]. In a TV design lab a high performance reference IF low pass filter is used in a test jig to allow benchmarking of tuner samples from different suppliers. The minimum requirements for this IF filter are: • linear phase low pass filter • 80 dB stop band attenuation • Sampling frequency f s = 50 MHz • transition band between 0.28 f s /2 and 0.31 f s /2 A filter with these specifications has been realized in an FPGA using the Multiplicative Finite Impulse Response (MFIR) approximation [5] [6] and is presented in this paper. Section II introduces the principle of the MFIR filter structure. Section III presents the actual implementation and hardware requirements of the filter. The paper ends with some conclusions. II. THE MFIR FILTER STRUCTURE MFIR filters are based on the basic expression [7] ( ) 2 0 0 1 for 1. i i i i x x x ∞ ∞ = = = + <  ∏ (1) The infinite sum in (1) can be approximated by a finite sum [5] ( ) 1 2 1 2 0 0 1 for 1. P i P i i i x x x − − = = = + <  ∏ (2) For the case where H(z) is the transfer function of a stable real pole (|| < 1) IIR filter, the application of (2) yields () ( ) () ( ) ( ) () () 1 1 0 1 1 2 1 2 1 1 0 0 0 1 1 1 . P i i i P P i i i i i H z z z H z z z M z M z λ λ λ λ ∞ − − = − − − − − = = = = = −   ≈ = + = =       ∏ ∏ (3) Where M(z) indicates the entire MFIR filter structure and M i (z) is called a stage of the MFIR structure. It is clear that for a good approximation of the pole filter, a sufficiently large number of stages must be chosen. Analysis has shown [6] that even for poles extremely close to the unit circle (|| = 0.99 for example) ten stages are sufficient to approximate the magnitude response of the approximated pole filter with a maximum deviation of just |0.01| dB. The first major advantage of the MFIR structure is the ability to replace a possible unstable IIR filter by a cascade of always stable FIR filter structures. An IIR filter with a transfer function H(z) having a conjugate pole pair  and  * where  = re +j ,  * = re -j and |r| < 1, can be approximated with a cascade MFIR structure given by [5] J-J. Vandenbussche is with the Faculty of Industrial Engineering Sciences, KULeuven, KULAB Brugge, Belgium, e- mail: jeanjacques.vandenbussche@kuleuven.be P. Lee is with the School of Engineering and Digital Arts, University of Kent, Canterbury, Kent CT2 7NT, UK, e-mail: P.Lee@kent.ac.uk J. Peuteman is with the Faculty of Industrial Engineering Sciences, KULeuven, KULAB Brugge, Belgium, e-mail: joan.peuteman@kuleuven.be