Estimation of inchannel-interference to Digital Radio Mondiale (DRM) signals during operation

Transcription

1 TRANS., VOL. X, NO. XX, MAY Estimation of inchannel-interference to Digital Radio Mondiale (DRM) signals during operation Iker Losada, Andreas Schaefer, David Guerra, Gorka Prieto, Javier Morgade and Pablo Angueira Abstract This paper presents two novel interference measurement estimators for the DRM (Digital Radio Mondiale) system. They are based on the PSD (Power Spectral Density) estimation of the received spectrum, and provide interference estimation whilst the DRM transmitter is in operation. The first estimator provides successful detection of in-channel narrowband interference. If the interferer is an AM co-channel signal, the second estimator takes advantage of the specific characteristics of the DRM modulation to provide accurate measurement of the signalto-interference ratio value. This allows protection ratio calculus and failure detection caused by interference. Index Terms DRM, Co-channel interference, Interference I. INTRODUCTION Communication systems can be heavily limited by interference. For that reason, in the study of the QoS of a communication system, it is important to have information about the existence of an interfering system. Furthermore, if there is an interfering system, it is important to know the amount of interfering power. Depending on the interfering power level, the system will be able to work properly or not. Moreover, if the interfering power level is high enough to degrade the quality of the system under study, knowledge of the interference level is necessary to find an appropriate solution. In the band allocated for the DRM (Digital Radio Mondiale) service [1] [4], the frequencies and the power that broadcasters use in their transmitters are regulated by the countries and/or international committees. In the case of medium and long wave, coordination of frequencies, powers and locations is done by the ITU, in the WRC (World Radiocommunication Conference) and RRC (Regional Radiocommunication Conference). The current planning for medium and long wave is based on the final acts of the RRC [5] prepared in Geneva in Although it has been slightly modified by mutual coordination between countries. During daytime, there are not much interference problems between DRM and AM broadcasting. The main reason is that, in LW and MW, during the daytime, the skywave propagation is highly attenuated, and the main propagation mechanism is groundwave propagation [6]. Under this stable scenario, it is easier to limit the propagation inside the intended coverage This work was supported in part by the Ministry of Education of Spain and the UPV/EHU under the scope of grants MCYT TEC and XX-XX. Authors are with the Electronic and Telecommunication Department of the University of the Basque Country. Author is with T-Systems Media&Broadcast GmbH. area, and an appropriate frequency planning can be achieved in order to fulfill almost all situations. During nighttime, when main propagation mechanism is skywave propagation, the situation changes dramatically [7]. This propagation mechanism relies on the reflection of the transmitted signal in the different ionosphere layers to propagate the signal along large distances (even several thousand of kilometers). However, this kind of propagation produces unstable scenarios, where propagation can make the received power levels vary between days. Under these circumstances, interference between transmitters located far away is more likely to happen, and the quality thresholds of the system are reached [8]. Thus, it is more difficult to produce an appropriate frequency planning. In these bands DRM and AM signals must coexist. Therefore, it is really common to have DRM signals interfered by co-channel AM signals during nighttime. The DRM signal provides different code rate protection levels (0.5, 0.6, 0.71 and 0.78) that trade data rate with protection against interference [1], [9]. Under these circumstances, it is important to accurately measure the level of interference present, as it is needed to choose the most convenient solution in each situation (e.g. select the optimum value for the code rate protection level or inform the regulatory body). The traditional method to measure the AM co-channel interferer level consists on switching down the DRM transmitter. Due to the varying propagation, in order to properly characterize the interference levels that appear during night time, it must be monitored during several nights. Consequently, this solution is not acceptable from an operational point of view. The novel method presented in this paper, apart from interference detection, allows accurate measurement of the AM cochannel interference level during normal DRM transmission operation. In cases where there is strong interference between services, noise requirements like the ones presented in [10], [11], are no longer the most restrictive cause. Therefore, this method allows to monitor the interference values during long periods of time, while the service is in operation. This is specially relevant for measurement campaigns [12] [14] and propagation studies [15], [16], so that interference limitation can be detected, and properly taken into account in the processings. In section II, the specifics of the DRM system that allow usage of the proposed measurement are presented. The available sources of data needed to calculate this measurement are discussed in section III. The interference detector and measurement are presented in section IV. The power correction needed in order to compare protection ratios and signal to interference

2 TRANS., VOL. X, NO. XX, MAY Fig. 1. Different parts of AM interfered DRM signal spectrum (not to scale) ratio measurements are considered in section V. In section VI, the laboratory test performed in order to validate the accuracy of this interference measurement is explained. The results of the laboratory tests are presented in section VII. Finally, in section VIII, the conclusions are summarized. II. DRM SYSTEM The DRM system was designed with special care on the real scenario of deployment [1] [3]. In these bands, most analog transmissions are AM signals. Under these circumstances, the center carrier of an OFDM system would be heavily degraded by such interferer. In order to avoid this situation, and to achieve better protection ratios against the existing analog services, the DRM system does not have a carrier in the center of the transmission channel [1], [9]. In figure 1 a graphical description of the spectrum of a typical 9 khz MW channel, containing a DRM signal interfered by a co-channel AM signal is presented. The DRM signal consists of a lower DRM signal part (L) and an upper DRM signal part (U). The space between both parts was intentionally left unused in the DRM standard [1]. The system was defined in such a way, that if a co-channel AM signal was present, the carrier of such interfering signal would appear inside both DRM signal parts. The interference measurement method proposed in this paper, needs an estimation of the power spectral density (PSD) of the received signal. In the DRM system, this PSD estimation can be obtained from two different sources, the RSCI (Receiver Status and Control Interface) [17], or directly calculated from the incoming signal. The RSCI is a monitoring protocol included in the DRM system professional receivers. This protocol provides important information about the reception, like the Delay or Doppler levels of the channel, every DRM frame (400 ms). A. RSCI PSD Estimation III. PSD ESTIMATION The RSCI PSD estimation provides a PSD estimation in decibels of the received input signal with sampling resolution of Hz and resolution bandwidth of Hz (values standardized in [17]). The range of this estimation is from -8 khz to 8 khz for half and full bandwidth DRM transmission modes (4.5 khz, 5 khz, 9 khz and 10 khz nominal bandwidth), and from -8 khz to 18 khz for double bandwidth DRM modes (18 and 20 khz nominal bandwidth). This PSD estimation provides 85 values for half and full bandwidth modes and 139 values for double bandwidth ones. This data source has the advantage of being directly provided by the RSCI of the DRM receiver, and it can be successfully used to detect in-channel strong narrowband interferers. However, this PSD estimation source does not provide the necessary resolution bandwidth to accurately measure the signal to AM interference ratio. The main reason is that resolution bandwidth of this PSD estimation is too large and the interferer power measurement takes into account not only the energy of the AM carrier of the co-channel interferer, but also a large amount of DRM signal power. This fact will be further studied in section VII. B. IQ PSD Estimation If IQ signal is available, the PSD can be directly calculated from the baseband input IQ signal. The PSD of this complex band limited signal is calculated every one DRM frame using well-known Welch spectral estimation with Hann window [18] and 50% overlap. The input DRM baseband IQ signal is usually provided sampled at 12 khz or 48 khz. Applying the Welch estimation method with 50% overlap, 512 (2048) point FFT and 512 (2048) sample Hann window for 12 khz (48 KHz), the PSD samples are spaced Hz. Taking into account the Hann window spread, the resolution bandwidth of each sample of the PSD is Hz. Using these parameters, a single FFT is obtained every ms, however single FFT spectrum are averaged according to Welch estimation method to provided a value every 400 ms (18 averages with 50% window overlap). The 400 ms estimation time is set because in DRM, RF and quality parameters, are also provided every 400 ms, and channels usually remain stationary during this period. However, it should be remarked that the estimation time could be decreased if the stationarity of the channel requires it. As in all PSD estimation methods, the selection of the FFT size is a trade-off between frequency resolution and estimation variance. For efficiency reasons only radix 2 FFT sizes will be considered. In order to select the previous parameters, the lowest useful frequency resolution was selected, in order to minimize the estimation variance. The minimum acceptable frequency resolution is the one capable of measuring the co-channel AM signal inside the DRM signal gap (82 Hz as defined in subsection IV-A). The smallest radix-2 FFT size that allows this is the one presented in the previous paragraph, that with 3 samples integrating the power of Hz. Therefore, these values are an agreement between frequency resolution, estimation variance and method accuracy. The overlap window value of 50% is a good agreement between reduction of estimation variance and increase of computational effort.

3 TRANS., VOL. X, NO. XX, MAY This PSD estimation has the advantage of being calculated to fill the needs of the interference estimation method. Therefore, it does not only allow strong narrowband interference detection, but also provides accurate estimation of the DRM signal to AM interference ratio. IV. INTERFERENCE MEASUREMENT FIGURES In this section, the input PSD estimation signal x is considered to be in linear scale regardless it was obtained from the RSCI or directly from the IQ signal. A comparison of the novel co-channel interference measurement accuracy using the both PSD sources is presented in section VII. A. Signal description The graphical description of the typical PSD of a DRM signal interfered by a co-channel AM signal is presented in figure 1. This spectrum can be divided into the following parts: Lower DRM signal (L): It is the part of the DRM spectrum under the center frequency of the channel. Upper DRM signal (U): It is the part of the DRM spectrum above the center frequency of the channel. Measurement gap (G): It is the part of the PSD spectrum that is centered in the channel, and defined to have 82 Hz banwidth regardless of the robustness mode and spectrum occupancy used. Excluded gap (X): It is the part of PSD spectrum between the upper and lower DRM parts, and the measurement gap. This part of the PSD contains neither DRM signal power nor co-channel interferer power. The bandwidth of these parameters depend on the spectrum occupancy and the robustness mode used in the DRM transmitter. Bottom and upper limits for the lower DRM signal (L) and upper DRM signal (U) are the limits standardized in [1]. The measurement gap (G) is defined to take into account carriers with power inside the 82 Hz measurement gap. The minimum spacing between the upper and lower DRM signal parts, happens for mode B [1], and is 92 Hz. The selected value is the nearer possible value that do not integrate DRM power into the AM signal measurement, and minimizes PSD estimation variance. Furthermore, the maximum deviation of the carrier frequency of the AM carrier in the transmitter is limited to ±10 Hz [19]. Though in modern AM transmitters the maximum deviation is usually smaller than ±5 Hz in the worst case, when synchronization to the synthesizer is not used [20]. Taking into account the maximum transmitter deviation, the selected measurement bandwidth also assures that all the carrier power is integrated. The excluded gap (X) is determined from the values taken by the rest of the parameters. It should be explained that only the power of the carrier is considered. There are two reasons, the protection ratio as defined in [9] also do not consider them, and the power contribution to the overall signal is small (this will be demonstrated in section V). The proposed values of the interference measurement parameters for the different robustness modes and spectrum occupancies are summarized in table I. It should be remarked that the AM carrier in the excluded gap may be additionally phase modulated, as this fact does not affect the power measurement. If the interferer AM carrier is co-channel and inside the measurement gap, as it is usually the case, the estimator described in subsection IV-C provides accurate measurement results. In the case that the AM carrier interferer is not inside the measurement gap, as it may happen in spectrum occupancies 4-5 (18 and 20 khz), but inside the DRM parts, the strong narrowband interferer may be detected using the estimator described in subsection IV-B. B. In-channel strong narrowband interferers Detection of strong narrowband interferers in the DRM spectrum can be accomplished by considering the PSD estimation signal x, and calculating the equation (1). D i = 10 log x i k L,U x k, i L, U (1) where, x i are values of PSD estimation signal in linear scale. L is the lower DRM signal part (see figure 1). U is the upper DRM signal part (see figure 1). The equation provides a detection value D for each value of the PSD estimation signal. The number of values vary depending on the data source frequency and resolution bandwidth, also with the robustness mode and spectrum occupancy. The detector provides an estimation of the power level of each PSD value with respect to the overall power of the DRM signal. Therefore, high values of this detector mean that a narrowband interferer is present in the frequency that corresponds to that value of the detector. This detector is useful to detect the presence of strong interferer outside the measurement gap. It should be noted that this detection method is only valid for strong AM carriers with power levels higher than the DRM carriers (at least 3 db over the average of the DRM carriers, to avoid detection of overboosted carriers like frequency references). It is still possible that AM carriers located outside the gap affect the QoS, and not being detected by this detector (because their power is sufficient to affect quality, but too small to be detected). In real field trials, if it is known that QoS degradation exist, and no reason can be found, it can be useful to reduce the power of the transmitter by 15 db (to try to detect this fact). Taking into account the protection ratios of ITU-R BS.1615, such interferer, would be detected by the estimator. C. Co-channel AM interferer Under the assumption that co-channel AM interference is present, the obtained signal PSD will be similar to the one depicted in figure 1.

4 TRANS., VOL. X, NO. XX, MAY TABLE I INTERFERENCE MEASUREMENT PARAMETERS (L/U/X/G) Spectrum Robustness Mode Occupancy A B C D Unit 0 (4.5 khz) 0/4208/41/82 b 0/4265/5/82 N/D a N/D a 1 (5 khz) 0/4708/41/82 b 0/4828/5/82 N/D a N/D a 2 (9 khz) 4208/4208/41/82 b 4265/4265/5/82 N/D a N/D a 3 (10 khz) 4708/4708/41/82 b 4828/4828/5/ /4704/26/ /4714/65/82 Hz 4 (18 khz) 4041/13041/41/82 b 4078/13078/5/82 N/D a N/D a 5 (20 khz) 4541/14541/41/82 b 4640/14578/5/ /14522/26/ /14464/65/82 a Robustness modes C and D not defined for these spectrum occupancies. b Excluded gap bandwith for mode A is bigger than for mode B because mode A has three unused carriers in the center of the channel instead of one. Taking into account the PSD spectrum parameter values given in table I for each DRM mode, the interference measurement is calculated from the estimated PSD signal (x) using the expression (2). Î = 10 log i G i L,U x i x i (2) where, x i are values of PSD estimation signal in linear scale. L is the lower DRM signal part (see figure 1). U is the upper DRM signal part (see figure 1). G is the measurement gap (see figure 1). The accuracy of the interference measurement has a bottom limit that depends on the noise floor of the PSD estimation. Near this floor the accuracy of the interference measurement will be affected by it. However, it must be remarked that this bottom limit is much lower than the levels of interference that cause quality degradation. This measurement does not take into account the contribution of the power of the sidebands of the AM signal. Therefore, it fully agrees with the protection ratio definition for DRM to AM signals [9]. In order to compare the interference measurement with the protection ratio, the sign of the value in decibels must be reversed P R = Î. This estimator takes advantage of the fact that the AM carrier power is inside the measurement gap in the DRM signal spectrum. It should be noted that for AM carriers out of the gap, only detection is possible using the estimator explained in subsection IV-B. V. POWER CORRECTION FACTOR In the measurement of an AM interferer, it is usual to consider the power of the AM carrier, instead of the power of the whole AM signal [9]. Also the protection ratios are calculated based on the power of the carrier. However, if signal to interferer power ratio measurements and signal to AM carrier power ratio measurements are to be compared, a correction factor must be considered. This is the case of comparing the values calculated using the interference measurement with the values used in a channel simulator. The interference measurement calculates the interferer power only from the carrier whilst the channel simulator generates true signal-to-interferer ratios. In this section theoretical calculus of the correction factor is performed. Let î be the interferer measurement (Î) described in subsection IV-C, in linear scale. Lets define i as the signal-tointerference power ratio in linear scale, and I in decibels. The interferer measurement (î) does not consider the contribution of the sidebands to the AM power, but integrates their power into the DRM power term. Therefore, assuming that the DRM signal and the AM signal are uncorrelated, the linear power of both signals is defined in equations (3) and (4). i = P AM = P Carrier + P Sidebands P DRM P DRM (3) P Carrier î = P DRM + P Sidebands (4) where P AM is the power of the AM signal, P DRM is the power of the DRM signal, P Carrier is the power of the carrier of the AM signal and P Sidebands is the power of the sidebands of the AM signal. The power of the AM carrier and AM sidebands can be related to the overall AM signal power by taking into account the modulation index (eq. 5 and 6). P Carrier = q P AM = m 2 P AM (5) P Sidebands = (1 q) P AM = m2 2 + m 2 P AM (6) where, P Carrier is the power of the AM signal carrier. P Sidebands is the power of the AM signal sidebands. P AM is the power of the AM signal. q is the AM carrier power to AM signal power ratio. m is the AM signal modulation index.

5 TRANS., VOL. X, NO. XX, MAY Fig. 3. Interference measurement setup Fig. 2. AM power to carrier correction factor for different modulation indexes protection ratio that the DRM service needs to operate under the presence of the used AM interferer signal. This protection ratio is the failure threshold for this specific co-channel AM interferer. Using equations (5) and (6) in equation (4), the expression in equation (7) can be obtained for the interference measurement in decibels. ( ) q P AM Î = 10 log = P DRM + (1 q) P ( ) AM q = 10 log 1 (7) i + (1 q) The difference between the estimated interference measurement and the real interference power level to DRM power level ratio, can be expressed as a correction factor (C f ). In operation, the characteristics of AM interferer signals vary a lot. Selection of the type of AM signal is a complex topic. Some broadcasters want to have high modulation factor to increase the coverage area, whilst others prefer higher sound quality. The correction factor has been calculated for the typical operational values of the average modulation index [9], [21], and it is presented in figure 2. Using this graph, along with the relationship of the equation (8), the signal-tointerference and the interference measurement can be obtained. I = Î C f (8) It can be seen that the correction factor varies with the power of the AM interferer, it increases as the power of the interferer increases. Furthermore, the contribution of the correction factor to the final result, depends on the average modulation index of the AM signal. Anyway, it should also be remarked that the contribution of the sidebands to the overall AM power is limited to 1 db in worst operational case. This correction can be applied to relate the AM power to DRM power measurement with the protection ratio measurement. In section VII, it will be used to compare both measurements. VI. VALIDATION LABORATORY TESTS The objective of the laboratory test is to validate the accuracy of the interference measurement, and to obtain the A. Laboratory Setup The test setup is depicted in figure 3 and consists on the following professional equipment: Content Server The content server used is the Fraunhofer DRM Content Server TM R4 [22]. This equipment is responsible for generating the digital content of the DRM transmission. It is capable of generating the PRBS (Pseudo Random Binary Sequence) that will be used in reception to calculate the BER (Bit Error Rate). DRM Transmitter The transmitter used in the test is the Transradio DMOD2 DRM signal generator and RF processor [23] [24]. It is a full integrated DRM transmitter, that allow flexible extension of the DRM capabilities by using a content server. This equipment is responsible of receiving the digital content generated by a content server, and generating the DRM RF signal. Channel Simulator The professional channel simulator is the Fraunhofer DT230 [25]. This equipment has been specifically designed to perform channel simulation for the DRM system, and has full support for channel simulation using stationary Watterson model approach [26], AWGN addition and customizable interferer inclusion. Furthermore, it allows full remote control based on scripts. Monitoring Station The DRM monitoring station used is the Fraunhofer DT700 [27]. This equipment provides full DRM demodulation and analysis. Furthermore, it also provides RSCI and input signal baseband IQ recording. Control PC The control PC is responsible for executing the program with the test. It is also used for retrieving the RSCI and IQ files stored in the Monitoring Station. The DRM content server generates the DRM multiplex containing the PRBS, that is fed into the DRM transmitter [28]. The DRM Transmitter receives the multiplex stream and generates the DRM RF signal. This unimpaired RF signal is fed into the channel simulator. The channel simulator executes the program commanded from the control PC and generates

6 TRANS., VOL. X, NO. XX, MAY the impaired output RF DRM signal. This impaired output signal consists of the input signal plus the interfering signal, no multipath signal is generated. This output signal is fed into the monitoring station. The monitoring station records the input signal and RSCI. When the program ends, the recorded files are retrieved via ethernet from the monitoring station to the control PC. The DRM configuration used for this laboratory test is DRM robustness mode B, with spectrum occupancy 3 (9 khz) and SM (Standard Mapping). EEP (Equal Error protection) with protection level 1 (coderate 0.6) was used. The MSC (Main Service Channel) was modulated in 64 QAM and the SDC using 16 QAM. Long interleaving (2 seconds) was also used. B. Test Description The test consists on introducing an AM interferer signal using the channel simulator at different power levels in order to assess different DRM signal to AM interferer ratio levels. The interference measurement will be applied to the input IQ signal recorded by the monitoring station. The accuracy of the interference measurement will be obtained by comparison of these values, with the power ratios introduced by the channel simulator. In order to calculate the protection ratio for this AM interferer, a PRBS is used in the DRM multiplex. The monitoring station is capable of handling this PRBS, and provides the number of total received bits and the amount of error bits. These values are generated for each received DRM frame, and the results are available in the RSCI. The channel simulator is used only to include the interference, the channel selected is a perfect transmission channel (with neither attenuation, delay, nor Doppler). The AM IQ interferer is a typical good quality signal with a moderate average modulation factor (23%) [9]. The power is varied from -25 db to 0 db, increasing 1 db every 30 minutes. The time the power is maintained, is considered to be sufficient to detect isolated errors that could be produced by external causes, and provide accurate BER measurement. Under these conditions, the overall test takes approximately 13 hours to complete. The IQ and the RSCI files are processed using a software tool, that extracts the total number of received bits, and the number of received error bits from the RSCI, and calculates the interference measurement at frame level. This data is stored in a database to allow further analysis and results. VII. RESULTS A. Interference Measurement Accuracy In order to study the accuracy of the interference measurement, the interference measurement will be compared to the programmed values in the channel simulator. The interference measurement is extracted from the database, and its values are corrected using the correction factor curve calculated in section V (see figure 2, m=0.2). In order to compare the estimated power ratios with the real ratio, the median value of the estimated power Fig. 4. Interference Measurement Accuracy ratios within the 30 minutes measurement step is obtained. The 90% measurement spread (difference between the 95 th and the 5 th percentiles of the measured values) of the estimated values within the 30 minutes measurement is also computed. This is done to compute the accuracy of the individual interference estimations measured at frame level. Finally, the absolute error is calculated as the difference between these values, and the values that were programmed in the PERL script, as it can be seen in (9). ε i = Î i C f,i I i (9) where, ε i is the absolute error in power step i. Î i is the median value of the frame level interference estimations in the power step i. C f,i is the correction factor for the power step i. I i is the interference to DRM power ratio programmed in the channel simulator for the power step i. This result is depicted in figure 4. In the graph, three different areas of interest for the signalto-interference measurement can be observed: Under -20 db. Saturation zone. In this range the measurement is highly saturated and it does not provide accurate measurements, as it is highly affected by the PSD noise floor. In this zone, the measurement error increases fast as the power ratio decreases. Also the 90% measurement spread of the measurements at frame level increases fast. -20 db to -16 db. Affected zone. In this range, the measurement is still affected by the noise floor. However, the influence of floor noise is lower, and a reduced accuracy of 2 db is achieved. The 90% measurement spread of the individual measurements at frame level is limited to 1 db. -16 db to 0 db. Precision zone. In this range, the measurement is not affected by the noise floor. The

7 TRANS., VOL. X, NO. XX, MAY Fig. 5. Interference measurement test using different PSD estimations. Fig. 6. Interference Measurement Threshold measurement accuracy in this zone is better than 1 db. The 90% measurement spread of the individual measurements at frame level is limited to 0.2 db. Taking into account that the DRM protection ratio (and consequently the interference failure threshold) for co-channel AM interferers, for the different DRM spectrum occupancies, modulation and protection ratios is between 0.6 and 10.7 db [9], all of them can be measured inside the precision zone. Therefore, it can be asserted that this novel interference measurement method provides an accuracy higher than 1 db. It should be remarked that in this zone the 90% of the measured values at frame level, only deviate 0.1 db from the median value used for the error calculus. It must be remarked that the average modulation index of the interfering AM signal does not affect the accuracy of the interference measurement, as only the carrier power is measured. This definition fully agrees with [9]. AM signals with the same carrier power, but different average modulation indexes, would provide similar interference measurements. Of course, their effect on the quality of the DRM signal will differ. That is the reason why the protection ratios are defined for high compression signals. Average modulation index must only be taken into account when comparisons with real signal to interference definition is applied. In this cases, the correction factor provides the necessary power to avoid the measurement being affected by the average modulation index of the AM signal. Anyway, it should be remarked that in the worst case, this correction factor is smaller than 1 db. As explained before, usage of the PSD obtained from the RSCI provides inaccurate results. This is mainly caused by the large resolution bandwidth of the PSD estimation. In figure 5, the interference estimated using both PSD sources is compared. It can be clearly seen that, in order to obtain accurate results with the proposed novel co-channel interference measurement method, IQ based PSD estimation with the parameters provided in section IV must be used. B. Protection Ratio Using the same results of the laboratory tests, the protection ratio for this AM interferer is calculated. The BER limit for the DRM system is 10-4 [9]. As it can be seen in figure 6, this corresponds to 5.2 db. This value can be compared with the protection ratio recommended in [9], that is 7.3 db. This protection ratio was calculated with the colored noise interfering signal described in [21], and it was obtained for high compression AM signals [9]. As the value provided by the recommendation takes into account the worst case, it makes sense to obtain lower protection ratio values for less stressing interfering signals. It should also be noted, that the protection ratio threshold is also receiver dependent, and a professional receiver was used for these measurements. C. Field trials The proposed interference estimator method has been successfully used in the field trials performed in Germany from August 2006 to August 2007 in order to test DRM MW NVIS Nighttime propagation. The commercial program from OldieStar radio was broadcasted from a MW transmitter located near Burg operated by T-Systems Media and Broadcast GmbH, using a triple horizontal dipole antenna fed with 100 kw r.m.s power. The transmission under study was broadcasted daily from 0:00 to 6:00 MET (Middle European Time). The DRM system was configured to provide a 9 khz signal with robustness mode B, with a modulation depth of 64 QAM for the MSC and 16 QAM for the SDC. The protection level used was 1 (coderate 0.6) and long interleaver (2 seconds). Two types of measurements were performed, long-term measurements at fixed locations at different distances from the transmitter and mobile measurements around Germany. Each receiver recorded the IQ file and the RSCI profile D file. At the processing stage, the interference measurement method was calculated for every DRM frame from the IQ file, and this value was incorporated to a database along with the most important information of the RSCI.

8 TRANS., VOL. X, NO. XX, MAY There are slight differences between the laboratory test procedure and the field trial measurements. In the laboratory tests, a PRBS is transmitted and BER quality objective is used according to the procedure described in ITU-R BS This laboratory test provides proof of the interference measurement validity. However, in real field trials during operation it is not possible to use a PRBS in the transmitter, the real audio data must be transmitted. Under these circumnstances, the AudioQ quality criteria is used instead of the BER one [8]. This quality assesment criteria is calculated at minute level. The AudioQ is defined as the ratio of correctly decoded audio frames to the transmitted audio frames within one minute. Only minutes with AudioQ values higher or equal than a 98% are considered to be excellent, while AudioQ values under that threshold are considered to have poor quality. The interference estimator is calculated for every DRM frame (400 ms), while the quality criteria can only be applied at minute level. In order to relate observed interference levels to quality, a representative statistic of the observed interference values at minute level is needed. Depending on AM and DRM propagation channels, the interference can vary strongly within one minute. For this reason, usage of first order statistics like median value is not representative of the failure, as the interference only needs to affect sufficiently the DRM signal for a few seconds to make the quality drop from excellent to poor quality level. While combined usage of first and second order statistics is possible, like median and IQR (Inter-Quartile Range) or variance, a more convenient way is to make use of the percentile statistic. This way, a one to one relation between interference and quality at minute level is possible. The statistic selected for this study is the 90 th percentile of the interference measurements at frame level. It may seem that a more natural choice would be to use the 98 th percentile, as it seems to be more related to the 98% quality threshold. However, this statistic usually overestimates the failure. The statistic does not take into account how the failure is distributed within the minute. If the high values are distributed in the minute, the protection mechanisms (channel coding and interleaving) of the DRM system, may be able to prevent quality degradation. The 90 th percentile provides convenient uniform estimation using only one value and enough robustness against the interference overestimation. Therefore, the 90 th percentile of the interference measurements at frame level, and the AudioQ are calculated for the frames database. In the database, minutes failed by different causes (low field strength, delay, Doppler and interference) are present. In order to compare interference and quality levels directly, the minutes affected by the remaining causes are excluded. In figure 7, the distribution of the interference of the minutes by quality are presented for this subset of minutes extracted from the different locations and dates is presented. It can be seen that the quality of the transmission and the values provided by the interference estimator are highly related. It should also be noted that the threshold obtained from the field trials (6.95 db) is closer to the 7.3 db of the ITU-R BS.1615 recommendation than the 5.2 db obtained in the laboratory tests. Apart from the measurement procedure differences, it Fig. 7. Probability Density Function Excellent Poor Interference Estimator (db) Distribution of the interference estimator values by quality must be noted that the AM signal used at the laboratory tests and the one present in the measurement campaign may differ in compression level, thus providing different protection ratio thresholds. VIII. CONCLUSIONS In this paper novel methods for interference measurement in operation using the DRM system are presented. The detection method can detect strong narrowband co-channel interferers. Furthermore, in the case of co-channel AM interference, the method takes advantage of the cleared carrier(s) in the center of the DRM channel to accurately measure the signal to AM Carrier interference ratio. This is the most common interference scenario in the DRM bands. Specially in the phase of introduction were AM and DRM signals must coexist. The detection method, only relies on the PSD of the incoming spectrum, and could be used in other technologies affected by strong narrowband interferers. However, the accurate measurement of co-channel AM interference relies on the peculiarities of the DRM systems and it is specific to the DRM system. This measurement allows to detect when the system failure is caused by interference. This is specially useful when failure caused by other factors (e.g. delay, Doppler) must be identified and studied in interfered environments. This usually happens when DRM nighttime transmissions in the MW must be monitored. This co-channel interference estimator has been successfully used for detection of interference failure cause in a nighttime DRM NVIS measurement campaign in the MW performed in Germany. In this measurement campaign the same DRM configuration was used. The campaign consisted of a DRM transmitter and several fixed monitoring stations scattered in the coverage area of the transmitter. Mobile measurements were also performed. All the receivers calculated, along with other parameters, the co-channel interference measurement using the received IQ signal as described in this paper. It must be remarked that, during the processing of the results, there

9 TRANS., VOL. X, NO. XX, MAY was strong correlation between frames received with estimator values higher than the protection ratio and degraded QoS. Furthermore, usage of the AudioQ criteria, along with the 90 th percentile of the interference measurements at minute level, allowed detection of the failure cause and determine the share of minutes affected by interference. The laboratory test demonstrated that this novel method has an accuracy better than 1 db in the precision zone (within -16 and 0 db). The DRM system protection ratio to a co-channel AM interferer, falls within this range for any combination of the DRM transmission parameters. Regarding the protection ratio, it can be concluded that DRM to AM protection ratio was calculated for high compression AM signals. This means that lower failure thresholds can be obtained for more common, moderate compression AM signals. In the case of field trials, usage of the 90 th percentile statistic of the interference measurements, allows direct relation of quality and interference failure cause. ACKNOWLEDGMENT The authors would like to thank T-Systems Media&Broadcast for their support. REFERENCES [1] Digital Radio Mondiale (DRM); System Specification, European Telecommunications Standards Institute Std. ETSI ES v2.1.1, Oct [2] Digital Radio Mondiale (DRM) - Part 1: System specification, International Electrotechnical Commission Std. IEC , Mar [3] System for digital sound broadcasting in the broadcasting bands below 30 MHz, International Telecommunications Union ITU-R Recommendation BS , Oct [4] F. Hofmann, C. Hansen, and W. Schäfer, Digital radio mondiale (DRM) digital sound broadcasting in the AM bands, IEEE Transactions on Broadcasting, vol. 49, no. 3, pp , [5] Final Acts of the Regional Administrative LF/MF Broadcasting Conference (Regions 1 and 3), ITU-R Std. GE75, [6] J. D. Parsons, The Mobile Radio Propagation Channel, 2nd ed. Baffins Lane, Chichester, West Sussex PO19 1UD, England: John Wiley & Sons LTD, [7] K. Davies, Ionospheric radio, ser. IEE Electromagnetic waves series. Peter Peregrinus Ltd. on behalf of the IEE, 1990, vol. 31. [8] I. Fernández, J. M. Matías, D. Guerra, P. G., D. De la Vega, and J. Ordiales, Subjective evaluation of the reception quality thresholds for the digital radio mondiale broadcast standard, in 56th Annual IEEE Broadcast Symposium, [9] Planning parameters for digital sound broadcasting for frequencies below 30 MHz, International Telecommunications Union ITU-R Recommendation BS.1615, [10] G. Prieto, M. Velez, A. Arrinda, U. Gil, D. Guerra, and D. de la Vega, External noise measurements in the medium wave band, IEEE Transactions on Broadcasting, vol. 53, no. 2, pp , [11] G. Prieto, M. Velez, P. Angueira, D. Guerra, and D. de la Vega, Minimum ON requirements for DRM reception based on field trials, IEEE Communications Letters, vol. 9, no. 10, pp , [12] J. M. Matías, I. Losada, D. Guerra, G. Prieto, U. Gil, and P. Angueira, Local radio coverage using the digital radio mondiale, in IEEE International symposium on broadband multimedia systems and broadcasting, April [13] D. Guerra, U. Gil, D. de la Vega, G. Prieto, A. Arrinda, J. L. Ordiales, and P. Angueira, Medium wave digital radio mondiale (DRM) field strength time variation in different reception environments, IEEE Transactions on Broadcasting, vol. 52, no. 4, pp , [14] J. M. Matias, I. Losada, P. Angueira, U. Gil, J. L. Ordiales, and A. Arrinda, DRM (digital radio mondiale) local coverage tests using the 26 MHz broadcasting band, IEEE Transactions on Broadcasting, vol. 53, pp , [15] D. de la Vega, S. Lopez, D. Guerra, G. Prieto, M. Velez, and P. Angueira, Analysis of the attenuation caused by the influence of orography in the medium wave band, in Vehicular Technology Conference, VTC2007-Spring. IEEE 65th, April 2007, pp [16] D. Guerra, G. Prieto, I. Pena, S. Lopez, D. de la Vega, and P. Angueira, Accuracy analysis of the millington s prediction method in the medium waveband, in Antennas and Propagation, IEEE Transactions on, vol. 55, no. 2, St. Petersburg, Russia, 2007, pp [17] Digital Radio Mondiale (DRM); Receiver Status and Control Interface (RSCI), European Telecommunications Standards Institute Std. ETSI TS V1.1.1, Jan [18] P. D. Welch, The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms, IEEE Trans. Audio and Electroacoustics, vol. 15, no. 2, pp , [19] Electromagnetic compatibility and Radio spectrum Matters (ERM); Transmitting equipment for the Amplitude Modulated (AM) sound broadcasting service; Part 1: Technical characteristics and test methods, ETSI Std. V1.1.1, [20] Transradio SenderSysteme Berlin. AM transmitters. [Online]. Available: [21] Objective measurement of radio-frequency protection ratios in LF, MF and HF broadcasting, ITU-R Std. BS.559-2, Jun [22] Fraunhofer IIS. DRM ContentServer R4. [Online]. Available: [23] Transradio SenderSysteme Berlin. Transradio DRM Products and Services. [Online]. Available: [24] DRM Data Modulator DRM-DMOD Operating Manual, Transradio SenderSysteme Berlin. [25] DRM Test Equipment DT 230 User Manual, 1st ed., Fraunhofer Institut für Integrierte Schaltungen (IIS). [26] Testing of HF modems with bandwidths of up to about 12 khz using ionospheric channel simulators, ITU-R Std. F.1487, May [27] DRM Monitoring Receiver DT 700 User Manual, Fraunhofer Institut für Integrierte Schaltungen (IIS). [28] Digital Radio Mondiale (DRM); Multiplex Distribution Interface (MDI), European Telecommunications Standards Institute Std. ETSI TS v1.1.1, Dec