LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009 EE_SP_Mitigation of RFI in Ad Hoc Wireless Receiver Nodes P. Banerjee1, D. K. Basu2, and M. Nasipuri 2 ECE Department, Heritage Institute of Technology, Kolkata, India Email: prabir.banerjee@gmail.com 2 Computer Science & Engineering Department, Jadavpur University, Kolkata, India dipakkbasu@gmail.com;mitanasipuri@yahoo.com 1 Abstract— The Ad-hoc wireless networks are assuming great importance because of the ease of operation and their operation without any infrastructure. On the other hand, it has its own characteristics like operation from low-capacity battery, frequent change of routing and connectivity etc.. Bit Error Rate (BER) of any wireless data network is critically dependent on the signal to noise and distortion(SINAD) ratio or signal to noise ratio(SNR) or signal to interference ratio(SIR) parameters of a receiver. The degradation in receiver performance can be made up theoretically by boosting the transmitter power level but the maximum power level is also highly regulated. Consequently, solution by reducing the interference is more relevant for Ad-hoc nodes. In this paper, we have analyzed one of the most critical internal radio interference factor of a receiver and have tried some circuits successfully to reduce or remove the difficult-to-remove interference problem of such receivers. In this model, the radio receiver, which is an embedded part of the platform, receives a signal which is the sum of the propagated signal from the link and the noises contributed by the noise sources residing in the platform. Platform noise may be added into the receiver system at a number of possible locations and for any widespread wireless network like an Ad Hoc wireless network, the overall effect of this noise can be analysed only if the noise from individual sources are analysed and then summed up. Unfortunately, this is not the model used by most RF designers in the simulations. They generally assume a white Gaussian noise source for predicting the behaviour of the tuned intermediate frequency (IF) stages of receivers and the modulator stage of transmitters. Such models seldom match the real time scenario since the platform-generated noises are typically non-Gaussian. As such, the present radio design methodology does not comprehend the impact of such sources of noise at the time of design of the radio. It may become a cause for more concern in the wireless designs since they will have to perform under the following situations: • Higher processor/controller clock speed • More mobility of the nodes in the ad hoc wireless networks • Increasing density of nodes The following sub-sections deal with the sources of interference, modeling of the noise and analysis of the cause of desensitization of a receiver. Index Terms— Signal to noise and distortion ratio(SINAD), Signal to noise ratio(SNR), Signal to interference ratio(SIR), Phase Locked Loop(PLL) I. INTRODUCTION Almost all the modern Radio equipments are controlled by clock-driven processors or controllers so that the equipments are versatile. This means that radio frequency sources inside the compact radios are more likely to cause electro-magnetic interference (EMI) and electro-magnetic conduction (EMC) problem. Most of the radio-frequency (RF) coupling takes place in a radio in two ways, either through radiation or conduction. The hardware platform, in our paper, refers to those digital circuits which are designed around the embedded radio to provide different functionalities. The EMI and EMC generated by the hardware platform of the radios have become integral sources of interference for all radios. A realistic radio model in which a part of the radio transmission channel is associated with the platform of the wireless radio is depicted in figure 1. A. SOURCES OF INTERFERENCE It has been observed that there are quite a few powerful sources of electro magnetic interference in a modern day radio. Understanding the spectral content of the different digital signals helps one to gain an insight into how they can couple to incidental antennas such as heat sinks, PCB solder points , power planes etc.. Any clock signal (approximately square with 50% duty-cycle) in time domain, shows three distinct phases, namely, rise time, steady-state phase and fall time. To understand the pattern of interference in a radio, it is better to translate the time domain clock signal to its Fourier Transform as one can plan and adjust the signals to produce less interference only in the frequency domain. The solution of Fourier Transform is a function of rise time, fall time, time-period and harmonic order. The relationship between time and harmonic order is not the same. It is observed that the first harmonic is the strongest in terms of amplitude and it varies slowly with time. On the other hand, higher order harmonics, though much lower in amplitude, vary more quickly in time and, Figure 1 © 2009 ACADEMY PUBLISHER 115 LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009 therefore, could appear as a strongly varying interference source in a receiver. Another serious interference problem arises because of signal asymmetries i.e., when even and odd harmonic amplitudes are different. A trapezoidal signal radiates odd harmonics only and this is why in radio systems, the square waves are transformed into trapezoidal form by suitable filtering. It also helps to reduce radiated power since with increase in rise-time, the number of significant higher-order harmonics increases. Even then, the symmetry of the waveform has a role in generating only odd or odd and even harmonics. There are multiple sources of radio frequency emissions in a wireless radio, where the functionality of the radio is controlled by high- speed processor or controller. The overall picture can be compared with a situation, where there is a large number of dipole radiators. The radiation pattern is not deterministic and it varies greatly with the frequency of the signals, their shapes and the number of active radiators at a point of time and it complicates the things further, as far as reliable radio performance is concerned. The RFI, as discussed above, may be taken into consideration by modifying the model of equation (1), as suggested in [5]. The modified equation is given below: B. INTERFERENCE MODEL In any wireless data network, including Ad-hoc networks, a key performance metric of any radio link is its BER. It can be described as a decreasing function of the SIR observed at its receiver node. We can, therefore, consider the quality of service (QOS) of a wireless link as an increasing function of its SIR and we have used the parameter SIR in our problem formulation. In any ad-hoc wireless network, the radio link between the source and the destination has to work satisfactorily in spite of interference from other sources. Some of the sources can be exposed and some other may be hidden. Given N interfering links in a particular channel, the SIR of the ith link can be denoted by [1]: C. RESULT OF DESENSITIZATION The causes of internal RF interference in communication receivers can be studied by considering a microcontroller based double super-heterodyne receiver system. Let us consider a channel frequency f CH of this receiver. For such a double super-heterodyne receiver with the first Intermediate Frequency (IF) equal to 10.7 MHz, the translation loop equation for the first mixer is f CH ∼ f LO = 10.7 MHz ------ (3) where, fLO is the Local Oscillator frequency corresponding to the channel frequency, f CH . In any clock-based receiver, due to the non-linear characteristics of the mixer device spurious IF can be generated and this can easily be explained by the following equation: G i i Pi Ri = ---------------------------------N i + ∑ G i j Pj + ∑ Pk j≠I where i, j, k Є {1,2,3,……….., N} and Pk denotes the powers radiated by different clock sources residing in the hardware platform of the receiver. It is true that out of these different spurious, undesirable power sources Pk, the power level of only a few may be strong enough to degrade the SIR factor but it is impossible to foresee and isolate those sources. It is true since the degree of interference is also a function of the frequency. Therefore, it is required to establish that the factor Pk is the main additional source of noise that degenerates SIR and to provide a circuit solution to eliminate this factor of noise and, ideally, the solution should be independent of variables like frequency. G i i Pi Ri = -------------------------------(1) N i + ∑ G i j Pj j≠I m x f CLK ∼ n x f LO = 10.7 MHz --(4) where, both m and n are positive integers and start from 1. If condition stated by equation (4) is met, then spurious IF is generated and it leads to what is known as desensitization of receiver. where i, j Є {1,2,3,……….., N} In this equation, Pi is the power of the transmitter of the desired link i, Ni is the thermal noise power at the receiver of link i, Gi j is the power gain (actually loss) from the transmitter of the jth link to the receiver of the ith. one and Pj is the power of the undesired transmitter j which interferes with the receiver i . Now this well-known equation takes into consideration only the radio sources external to the receiver. It has been observed and established that apart from this cause, internal signals play a critical role in determining the performance in terms of receiver SIR of any microprocessor/microcontroller controlled synthesized radio. The radio frequency (RF) radiated by the internal signals can desensitize the receiver to any extent and the cause of this interference can be analyzed in the following manner. © 2009 ACADEMY PUBLISHER (2) D. MITIGATION OF RFI The basic method for getting rid of a RFI source is to remove the radiating mechanism. Conversion of electromechanical contacts, for example, to solid state would remove the arc-generated RFI. Electrostatic discharge (ESD) generated RFI could be reduced by removing the charge generating mechanism or by providing a method to bleed off the charge. If the sources of RFI are global and are identifiable, then it is relatively easier to eliminate the sources. If the RFI from a source, which is essential for the operation of the total radio system, is responsible for the desensitization of the receiver channel, then the solution cannot be a straightforward one. The communication systems must be able to 116 LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009 function in the presence of high-frequency interfering clock signals. There are four basic methods, which can be used to minimize RFI effects: • Eliminate the radiating source; • Shield either the source or the receiver; • Separate distance between the source of radiation and the receiver ; and • Improve circuit design. The following block diagram (Fig.2) represents a double-super-heterodyne FM receiver and it also shows the different functional blocks of a Phase-locked-loop (PLL). The signal picked up by the antenna is fed to the Low Noise Amplifier (LNA), the output of which is mixed with the first Local Oscillator (LO) signal. This local oscillator signal is the locked output of the Voltage Controlled Oscillator (VCO). Shielding is probably the most common means used to reduce the effects of RFI. The effectiveness of the shielding depends on the frequency of the incident waves, the thickness of the shielding plates, the property of the shielding material and slots or apertures in the cage. Efficient shielding in present day wireless radios, particularly ad hoc wireless nodes is very difficult to achieve in the presence of very high frequency clocks. Another alternative is to use absorbers instead of RF shields. The best approach as far as passive mitigation of RF interference is concerned, is to design the layout of the radio with a technique so that the sensitive radio part is located away from the noisy components Administrative controls can also be used to prohibit RFI sources from being operated near sensitive equipment. The first three approaches are not effective in present day designs of receivers because of the requirements of miniaturization and lesser weight. So, removing or reducing the interference by circuit design approach has become the accepted solution. In the present work, the sources of RFI have been analyzed and experiments have been carried out with an existing microcontroller based communication system. Finally, circuit based solutions have been proposed strongly for the reduction of the RFI problem caused by hardware platform of the radio. Figure 2 The first mixer output can be represented mathematically by: IF1 = {fCH + fm} ∼ fLO1 , where, fCH is the channel or carrier frequency; fm is the modulating frequency and fLO1 represents the first local oscillator frequency. For PLL schemes, fLO1 is equal to the voltage controlled oscillator (VCO) output in the locked condition. This frequency is programmed in such a way that the first Intermediate Frequency (IF), IF1 becomes independent of the channel frequency. As an example, if fCH = 80.0 MHz, IF1= 10.7 MHz (for the receiver) and fm= 1 KHz then fLO1 = 90.7 MHz. IF1 signal is then mixed with second local oscillator signal, LO2 which is fixed for a given receiver. For our system, it is 10.245 MHz so that the second intermediate frequency (IF2) becomes (455 KHz +fm). The detector stage extracts the modulating signal fm. For the PLL, IC of Philips numbered SAA 1057 was used. It contains the voltage controlled oscillator (VCO), the mixer, the low pass filter and the reference oscillator. The division ratio, N, is to be supplied to the PLL chip by suitable intelligent circuit which, in turn, is multiplied by the internal reference frequency to generate the channel dependent frequency for the mixer. The value of N in transmit mode (NTX) of operation of PLL is given by: NTx = (fCH / fref) where, fCH represents the transmit channel frequency of the set at a given point of time. In receive mode, the value of N in receive mode (NRX) is calculated by: NRx = (fCH ± 1st. IF / fref ) , where, fCh is the channel frequency, 1st.IF stands for the first intermediate frequency of the super-heterodyne receiver, fref is PLL chip specific reference frequency and is equal to 12.5 KHz.. The NTx as well as NRx are obviously decimal numbers and are to be converted to binary format before loading into memory. In the memory, blocks are to be earmarked for receive and transmit operation .The NTx(binary) and the NRx(binary) values are to be stored in appropriate blocks. LTX uses 8051family microcontroller to control receive and transmit operation. Since it is a simplex II. RFI MITIGATION BY HARDWARE APPROACH To establish Pk as the major source of noise, we have experimented with a microcontroller based Very High Frequency (VHF) band set of Philips Telecommunication. The model had the following characteristics. • It is a programmable, 16-channel set • It is a synthesized radio using phase locked loop( PLL) technique • The receiver section is double superheterodyne type with first IF equal to 10.7 MHz • It is a low band set working in 66-88 MHz range For generating the reference frequency, the PLL chip requires a stable 4.0 MHz temperature controlled crystal oscillator (TCXO). This reference frequency is multiplied by a factor N(the division ratio) which, in turn, is supplied to the SAA1057 chip. A. DETAILS OF EXPERIMENT © 2009 ACADEMY PUBLISHER 117 LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009 of the receiver but due to the reasons stated in this section itself. transceiver set, the single VCO and PLL are switched between receive and transmit modes. It fetches with the help of the microcontroller the NRx(binary) from receive memory block and sends to the multiplier section of the PLL for locking in receive mode. When the set is switched to the transmit mode, the controller fetches the NTx (binary) string from transmit block of memory and sends to the multiplier of the PLL for locking to the transmit frequency. The very high frequency synthesized set, code-named LTX, has two variants – low-band(68 to 88 MHz) and high-band(146 to 174 MHz). For our experimentation, the low-band LTX set was selected. The justification for this choice is explained below. With reference to equation (4), it is seen that spurious intermediate frequency (IF) is generated if: m x f CLK ∼ n x f LO = 10.7 MHz, as 1st. IF for LTX is 10.7 MHZ. Now, for simplicity, let us consider that there is no received signal at the antenna and let us assume that the microcontroller clock frequency is 4.0 MHz. Hence, putting the values of fCLK = 4.0 MHz and fLO = 90.7 MHz corresponding to the channel frequency of 80.0 MHz, in equation (3), we obtain : m= 20 and n=1. With increase in channel frequency, the value of ‘m’ will increase and this means that the intensity of spurious harmonic emission due to the clock signal will decrease. Therefore, to observe the performance in the worst possible condition, the low-band LTX set was chosen. C. FIRST SOLUTION It may be noted from the experimental results of section III, that only if the interference factor is nil (equation (4)), the SINAD or SNR will determine the receiver performance of a radio receiver. The following approaches were tried in the LTX to achieve the best possible result: • To eliminate the interfering factor, RF-shielding of the microcontroller Printed Circuit Board (P.C.B.) along with the Voltage Controlled Oscillator (V.C.O.) was done. It reduced the intensity of interference but the readings were well below the 12 dB mark for the interfering frequencies. • The microcontroller clock was made separate from the PLL clock and a different value of this clock immediately increased the SINAD values for the frequencies which were earlier giving low SINAD This second approach was studied further with different frequency combinations and some other disturbing data were exposed. It was observed that the different address bus lines(A0 – An) and data bus signals and other digital signal lines can also become sources of RFI .Hence, we thought that it will be best if the microcontroller system could be forced into idle/sleep mode once the receiver channel is locked. The following circuit was developed to sense the locked condition and the circuit was geared to track change in the channel. The schematic is shown here in Fig .3. B. EXPERIMENTS AND ANALYSIS To check the signal to noise and distortion ratio (SINAD expressed as S+N+D/N+D) at different frequencies over the entire band, the LTX was programmed with different frequencies.The SINAD was measured using the method described in CEPT standard MPT 1323 :1996[7]. The method asks for application of 0.3µV (pd) signal strength from a suitable radio frequency(RF) generator at the correct channel frequency to the antenna socket of the receiver. The receiver output has to be measured with the help of a suitable audio distortion analyzer and the (S+N+D/N+D) ratio should be more than or equal to 20dB if measured through a telephone psophometric network or better than 12dB without the network. The SINAD was recorded for all the frequencies at input signal strength of 0.3µV (pd) with the help of Rohde & Scharwz CMS-91 model and the readings are shown in the same table. It is to be noted that for professional voice-grade communication equipment, the minimum requirement for receiver sensitivity is 12 dB SINAD for 0.3µV (pd) RF signal input. After some abruptly changing pattern with frequency, it was analysed that: • the SINAD has become a function of the receive frequency • the SINAD degrades sharply when the receive frequency is a multiple of 4 To establish the points, the set was re-programmed with a new set of frequencies and readings were taken. It proved that the degradation was not due to the detuning © 2009 ACADEMY PUBLISHER Channel Sensor Circuit Control Module Micro- controller Block PLL Block Figure 3 This scheme worked satisfactorily and the system became very stable and RFI was totally eliminated. The final results obtained are tabulated in the table I below. SERIAL NO. 1 2 3 4 5 6 7 8 9 10 118 TABLE I FREQUENCY (MHz) 66.5 67 68 71 74 77 79 80 84 87.5 MEASURED SINAD(dB) 16 15.6 16.1 15.9 15.3 16.1 15.5 16.2 15.8 16.2 LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009 The first plot shown in figure 4 shows the wide variation in SINAD ratios when the internal controller clocks are running, marked especially by low SINAD at frequencies like 68 and 80 MHz. When the internal clock is stopped by the control circuit, the SINAD for the entire band of frequencies improves remarkably and the consistency is indicated by a very narrow band of SINAD variation in the figure 5.The shortcoming of this approach is that the microcontroller cannot execute any other operation once the frequency is locked. SINAD in dB 20 Figure 6 15 One of the most important output of such versatile design is the Radio Signal Strength Indicator (RSSI) output from the IF processor IC. This is a dc signal and is proportional to the IF signal strength. Therefore, this signal can be used to measure the level of platform noise in the receiver, which includes interference from the controller clocks also. In our scheme, we have utilized the RSSI output from the IF processor IC, NJM2537. The operation of the circuit is controlled by a micro-controller 80C51. The entire sequence of operation is as follows. The RSSI output of NJM2537 is connected to the noninverting input of an OP-AMP, functioning in comparator mode. The inverting input of the same is held at 0.6 volts dc. This is to ensure that the selection of the crystal frequency gives the best possible result since RSSI=0.6 Volts is generated only for IF signal strength of 60dBµV(emf)[9].When the RSSI level equals or exceeds the reference voltage, the comparator output becomes more or less equal to the positive supply voltage of the OP-AMP IC. The reference voltage varies with IC types[8]. This output is used to control the Enable pin of 4066,a CMOS analog transmission gate which transfers +5V dc to the Port 2.4 of 80C51.On the controller side, instead of the conventional practice of connecting a fixed frequency crystal across the pins X1 and X2 of the controller, our circuit uses 4 crystals, all of different frequencies. It is better if the frequencies are not whole numbers. We used two crystals of frequencies 3.58 MHz and 4.0 MHz to check whether our hardware and software schemes work properly. One end of the two crystals are shorted and connected to the terminal X1 of the controller. The other ends of the crystals are connected to the Y inputs of the individual section of analog transmission gates(4066 IC).The outputs(Z terminals) of all the 4066 sections are shorted and the shorted terminal is connected to the terminal X2 of the controller IC. This is possible since the 4066 is a tri-stated device and each of the gate will be enabled individually under software control. Port lines P2.0 to P2.3 are utilized to control the individual enable pins of 4066 gates.The software algorithm works in the following manner. In the power up condition, P2.0 enables crystal XL1 by default. If this frequency ensures +5V to P2.4, the controller does not change the active crystal for the controller. Otherwise, the software with the help of port lines P2.1 to P2.3, switches the crystal selection till P2.4 10 5 66 .5 67 68 68 .0 68 25 .0 5 74 77 80 80 .0 2 87 5 .5 0 Freq. in MHz 80 84 87 .5 77 79 66 . 71 74 16.4 16.2 16 15.8 15.6 15.4 15.2 15 14.8 5 67 68 SINAD in dB Figure 4 Freq. in MHz Figure 5 D. SECOND SOLUTION A more dynamic solution can be implemented if the check for spurious IF is made progressive with the help of software and the result is utilized to switch on the noninterfering oscillator frequency for the microcontroller. We have tried this second solution and have obtained almost same result as was obtained with our first approach. However, it involves some extra hardware as shown in Fig.6. This solution may be explained as follows. Most of the Intermediate Frequency (IF) stage operations in a radio receiver is executed by the hybrid IF processor chips in most of the new designs. Apart from the classical modules like the Mixers and the Band Pass Filters, these processors also are designed with signal processing blocks. © 2009 ACADEMY PUBLISHER 119 LETTERS International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009 standard code- named Version Envy which will be more RFI proof. We would like to carry out further work in this direction so that more power sensitive solution can be implemented and transmitter power level can be reduced still further in ad hoc wireless networks. is +5V.One has to remember that in this total scheme of operation of a receiver, the reference crystal for the PLL is a separate entity and this modified scheme is in no way interfering with this crystal. For operations like key-pad scanning and display interfacing, the change in source frequency of oscillation is not a factor. For communication softwares, the change in oscillator frequency is to be taken care of properly by designing suitable divider chain or by modifying the software. The two solutions are not strictly comparable since the second solution requires more circuit and is reactive. The advantage of this solution is that the microprocessor/microcontroller can be kept alive for other operations like display multiplexing. Therefore, the choice of the solution will depend on the versatility of the ad hoc network node and we are confident that this solution for RF interference mitigation will give much better and stable result from the point of view of consistent SINAD value which, in turn, is related to bit error rate in data communication. ACKNOWLEDGMENT The authors are indebted to CMATER and SRUVM projects of the Computer Science and Engineering (CSE) department, Jadavpur University for allowing them to use the laboratory and other facilities. Prabir Banerjee is grateful to the authorities of Heritage Institute of Technology, Kolkata for allowing him to work on the research paper, and Dr. D. K. Basu to AICTE, New Delhi for providing him with an Emeritus fellowship. REFERENCES [1] A. Ephremides, “Energy concerns in wireless networks”, IEEE Wireless Communication Mag., vol. 9, no. 4, pp.4859, August 2002 [2] A. Pal, Mehmet Akar, Michael G. Safonov, U. Mitra, “Adaptive Power Control for Wireless Networks Using Multiple Controllers and Switching”. [3] J. Rabaey, J. Ammer, J. L. da Silva Jr., D. Patel, “Ad-Hoc Wireless Networking of Ubiquitous Low-Energy Sensor/Monitor Nodes”. [4] M. Ibnkahla, “Signal Processing for Mobile Communications Handbook” [5] Intel StrongARM SA-1110 Microprocessor Developer’s Manual, June 2000 [6] xapp 347-application note of m/s Xilinx [7] Performance specification; Angle modulated radio equipment for use at fixed and mobile in the private mobile radio service operating in the VHF band, Revised and reprinted November 1996 [8] www.datasheetcatalog.org/datasheet/rohm/bh4127fv.pdf [9] http://www.datasheetcatalog.org/datasheet/newjapanradio/ be04010.pdf [10] C.C. Goodyear, Signals and Information, WileyInterscience,1971 [11] R. P. Feynman, R. B. Leighton and M. Sands, Feynman Lectures on Physics, Vol.2,Addison-Wesley 2006 [12] B. Skalar, Digital Communication, Prentice Hall PTR,2001 [13] K. Slattery, J. Muccioli, T. North, Modeling the Radiated Emissions from Microprocessors and other VLSI devices, IEEE 2000 EMC Symposium, Washington DC [14] H. W. Ott, Noise Reduction Techniques in Electronic Systems (2nd. Ed.),Wiley,1988 [15] A. F. Molisch, Wireless Communications, Wiley,2005 [16] M. Mardiguian, Controlling Radiated Emissions by Design, Springer, 2001 III. COMPARISON The solutions which have been described in this paper, have been compared with the following patents and commercial radio product. The comparisons with each of these have been enumerated. 1. US Patent number 7116958 2. US Patent number 7197291 3. US Patent number 6148039 4. Diversity UHF receiver UCR205D It has been established that these solutions have some weaknesses, which render those unsuitable for use as ad hoc wireless network nodes. IV. CONCLUSION In many Ad Hoc wireless networks, the nodes need not be proactive and it is sufficient if they are reactive only. This is true since power consumption is a critical issue for ad hoc network nodes or sensors. Hence, the first solution can be the solution for RFI mitigation inmost of the Ad Hoc wireless network nodes. However, depending on the roles the nodes play in a dynamic situation like changing the routes, transmission power level control etc., the second solution will be more acceptable. Lots of research work is going on to reduce or eliminate RFI from wireless receivers. In the Bluetooth standard , the receiver sensitivity level is kept as low as 70 dBm to take care of RFI and work is on for a new © 2009 ACADEMY PUBLISHER 120