S Simulation program SEAMCAT

Transcription

1 S Post Graduate Course in Radiocommunications Spring 2001 Simulation program SEAMCAT (The Spectrum Engineering Advanced Monte Carlo Analysis Tool) Pekka Ollikainen Page 1

2 Simulation program SEAMCAT (The Spectrum Engineering Advanced Monte Carlo Analysis Tool) 1. Introduction The statistical methodology described here and used for the tool development is best known as Monte-Carlo technique. The term "Monte-Carlo" was adopted by von Neumann and Ulan during World War II, as a codename for the secret work on solving statistical problems related to atomic bomb design. Since that time, the Monte-Carlo method has been used for the simulation of random processes and is based upon the principle of taking samples of random variables from their defined probability density functions. The method may be described as the most powerful and commonly used technique for analysing complex statistical problems. Monte Carlo approach is seen not to have an alternative in development of a methodology for analysing unwanted emission interference. The approach is: Generic - Flexible - A diversity of possible interference scenarios can be handled by single model. The approach is very flexible, and may be easily devised in a such way to handle the composite interference scenarios. In the figure 1 the signals An(t) with known statistical properties create an output signal Y(t) which has some yet unknown statistical properties. The result Y(t) depends on the model and the input signal An(t) statistics. One field of application is in the radio communication system simuation based on Monte Carlo method. The Monte Carlo simulation is based on random signal generation of appropriate type with the chosen properties. The result of interest in the simulation is the statistical properties at the output of the communications system. The systems blocks can be taken into account by inputting blocks performance variations which have also affect the result of the simulation. The simulation basically consist of the generation of the signals An(t), the processing the samples in the system model block(s) and the estimation of the output signal Y(t) properties. In pure Monte Carlo simulation all the input signals affecting the output should be considered. A1(t) A2(t) A3(t).. An(t) System Model Y(t) Figure 1. Model for Monte Carlo simulation Page 2

3 The simulation model need not to take into account all the random processes. Only those major signals that are considered to affect the results of the output are normally taken into account, depending on the accuracy the estimate of Y(t) needs to be considered. In a simple additive Gaussian noise radio channel the signal samples will have the same Gaussian noise properties at the receiver decision circuit and the result can be calculated analytically. Ideally the Monte Carlo simulation model correspond to a real system within reasonable limits. The expected output value from the system model is estimated in Monte Carlo similation according to E N 1 { g( Yˆ( t)) } = N n= 1 Equation 1. Expected output g( Y ( n)) where the ^ indicates the estimated value and N is the number of the samples. In digital communication systems the signal input can be generated by binary bit sequence {A(k)} having values of either +1 or -1. The binary signals can be generate by maximal sequence shift registers having a long repetition properties enough. The noise is typical random signal where the samples {N(k)} have for example Gaussian distribution G(σ). The faded radio signal may have a Rayleigh distribution R(σ). The generation of these signals and the respective output cannot be easily evaluated. G( σ ) = 2 2 2σ 2 1 exp πσ Equation 2. Gaussian distribution R( σ ) 2 r r exp σ 2σ = 2 2 x 2 Equation 3. Rayleigh distribution Page 3

4 1.1 ILLUSTRATIVE EXAMPLE ON UNWANTED EMISSIONS For interference to occur, it has been assumed that the minimum carrier-to-interference level (C/I) is not satisfied at the receiver input. In order to calculate the C/I experienced by the receiver, it is necessary to establish both the wanted signal and unwanted signal levels. Unwanted emissions considered in this simulation are assumed to result from active transmitters. Moreover, only spurious signals falling into the receiving bandwidth have been considered to contribute towards interference. For the mobile to fixed interference scenario, an example is shown in figure 2 below Many potential mobile transmitters are illustrated. Only some of the transmitters are actively transmitting and still fewer emit unwanted energy in the victim receiver bandwidth. It is assumed that interference occurs as a result of unwanted emissions from the most influent transmitter with the lowest path loss (median propagation loss + additional attenuation variation + variation in transmit power) to the receiver. An example of Monte-Carlo simulation process as applied to calculating probability of interference due to unwanted emission. For each trial, a random draw of the wanted signal level is made from an appropriate distribution. For a given wanted signal level, the maximum tolerable unwanted level at the receiver input is derived from the receiver's C/I figur. For the many interferers surrounding the victim, the isolation due to position, propagation loss (including any variations and additional losses) and antenna discrimination is computed. The lowest isolation determines the maximum unwanted level which may be radiated by any of the transmitters during this trial. From many trials, it is then possible to derive a histogram of the unwanted levels and for a given probability of interference, then to determine the corresponding unwanted level. By varying the values of the different input parameters to the model and given an appropriate density of interferers, it is possible to analyse a large spectra of interference scenarios. Figure 2. Illustrative example of unwanted interference Page 4

5 Page 5 Figure 3. Receiver signal levels and interference

6 2. SEAMCAT Radio Tool SEAMCAT Radio Tool is the implementation of a Monte-Carlo Radio Simulation tool managed by the group of CEPT Administrations, ETSI members and international scientific bodies. SEAMCAT is public domain software distributed by the CEPT European Radiocommunications Office, Copenhagen. This paper describes the implementation aspects of the simulation tool but not the use of the SEAMCAT tool that is explained in the relevant user documentation. 2.1 Use of radio spectrum The radio spectrum is a limited resource and can only be used optimally, if the the radio systems located in the same or adjacent frequency ranges are compatible by chosen criteria. For example, a criterion for radio compatibility is the difference in signal levels between the wanted and unwanted signals in the interfered receiver input. This parameter is used to derive a separation in the space or frequency for the different radio services. In the adjacent bands the most significant interferencing mechanisms are the unwanted emissions from the transmitters, and blocking and intermodulation in the victim receiver. The simulation of real time system takes lot of program time to have all the possible signal variations in time. The The Monte-Carlo simulation method is based upon the principle of taking samples of random variables from their defined probability density functions. The simulation tool user defines distributions of the parameters, and the software uses them to extract samples (also called trial or snapshot). Then, for each trial SEAMCAT calculates the strength of the interfering and the desired signal and stores them as arrays. The software derives the probability of interference taking into account the quality of the receiver in a known environment, and the calculated signals. The Monte Carlo method can used for almost all radio-interference scenarios, like e.g. sharing or compatibility studies. This flexibility is achieved by the way the system parameters are defined. Each random parameter (antenna pattern, radiated power, propagation path, ) is input as a statistical distribution function. It is therefore possible to model even very complex situations by relatively simple elementary functions. Only the statistical distribution of the parameters need to be known. 3 Description of the SEAMCAT approach One of the main requirements is to select such an architectural structure that the simulation tool is flexible. It should be flexible enough to accommodate analysis of composite interference scenarios in which a mixture of radio equipment sharing the same and/or multiple sources of interference (e.g. out-of-band emission, spurious emission, intermodulation..) are involved and can be treated concurrently. Other requirements would be that the proposed architecture lend itself for modular development and is versatile enough to allow treatment of the composite interference scenarios. The chosen Monte Carlo architecture which meets these constraints is presented in Fig. 2. The proposed architecture is basically of a sequential type and consists of four processing engines: Page 6

7 event generation engine (EGE) distribution evaluation engine (DEE) interference calculation engine (ICE) limits evaluation engine (LEE) User Interface System Manager Event Generation Engine Technical Database Manager Distribution Evaluation Engine Results Database Manager Interference Calculation Engine Technical Database Future Calculation Engine Results Database Figure 4. SEAMCAT architecture 3.1 EVENT GENERATION ENGINE (EGE) The event generation engine (EGE) takes the relevant parameters from the chosen interference scenario and generates information on the received signal strength of the desired as well as on the strength for each of the interfering signals included in the composite interference scenario. This process is repeated N times ( the number of trials), which should be large enough to produce statistically significant results. Generated samples of the desired and all interfering signals are stored in separate data arrays of the length N. The trials on parameters being common for desired and interfering radio paths are done concurrently in order to capture possible correlation between desired and interfering signals. Such an implementation will not cover only those seldom cases of interference in which one interference mechanism is excited by another interference (e.g. a strong transmitter mixes with spurious emission of the second transmitter and produce an intermodulation type of interference). Page 7

8 3.2 DISTRIBUTION EVALUATION ENGINE The distribution evaluation engine DEE takes arrays of the data generated by the EGE and processes the data with aims of: 1 assessing whether or not the number of samples is sufficient to produce statistically stable results 2 calculating correlation between the desired signal and interfering signal data and between different types of the interfering signal (e.g. blocking vs. Unwanted emissions) 3 calculating a known "continuous" distribution function as the best fit to the RSS random variable. First and third of the above points can be achieved using well known goodness-of-fit algorithms for general distributions such as the Kolmogorov-Smirnov test. If DEE detect unacceptable variation in discrete distribution parameters estimated in two successive estimations using N and N+dN sample size, the EGE is instructed to generate another dn of additional samples. This test is repeated until a tolerable variation of the parameters is measured over the pre defined number of successive tests. Three different kind of outputs are possible from the DEE engine: 1 data arrays of the wanted and interfering signals. This is the output in the case that a high degree of correlation is detected between the wanted and any of the interfering signals. 2 discrete distributions of the wanted and interfering signals are passed in the case of a weak correlation between the signals or in the case that there was no correlation between the signals but no «continuos» distribution approximation with satisfactory accuracy was possible. 3 continuous distribution functions of the wanted and interfering signals are passed to ICE in the case that signals were decorelated and discrete distributions were successfully approximated with continuos distribution functions 3.3 INTERFERENCE CALCULATION ENGINE The interference calculation engine (ICE) is the heart of the proposed architecture. Here, information gathered by the event generation engine (EGE) and processed by distribution evaluation engine (DEE) are used to calculate probability of interference. Depending on which kind of information was passed from DEE to ICE three possible modes of calculating the probability of interference are identified. Mode 1: data arrays fo passed by DEE to the ICE, and vector representing the composite interfering signal I is calculated as a sum of the data vectors. Mode 2: distribution function for the composite interfering signal is calculated by taking random samples for distributions and linearly adding them up. Page 8

9 Mode 3: The interference is calculated using numerical or analytical integration of the supplied distribution functions for each of the interference sources. Mode 4: All signals are assumed to be mutually independent and the overall probability for interference is identified as the probability to be disturbed by at least one kind of interference. Different criteria for calculation of interference probability can be accommodated within the processing engine. A cumulative probability functions can be calculated for C/N+I or N/N+I random variables. All interfering signal distributions are calculated with respect to reference levels or functions of unwanted (emission mask), blocking (receiver mask) or intermodulation attenuation. Interfering signal distributions for some other reference levels or functions can be derived by first order (unwanted or blocking) or third order (intermodulation) linear translation of the reference distributions 3.4 LIMITS EVALUATION ENGINE The limits evaluation engine (LEE) is envisaged to play a important role in two aspects of the tool development: 1 selection of optimal values for the limits 2 verification of the tool Output from the interference calculation engine (ICE) is presented as a multidimensional surface characterising the dependance of the probability of interference versus the radio parameters.two main features of the probability surface are: 1 the same probability of interference is achieved by different sets of the limit values for the radio parameters under consideration. 2 probability of interference parameter is not used in the radio system design and as such doesn t lend itself nicely for the validation through the system performance measurements. Instead, degradation in system coverage or traffic capacity seems to be more appropriate to understand impact of a particular probability of interference to the radio system performance. The radio variables are transformed from the probabilistic space into a system performance space enabling to evaluate the system performance degradation due to presence of interference. When the inter-system compatibility is analysed (e.g. unwanted emission) the radio coverage and/or traffic capacity can be used to evaluate the impact of the radio parameters limits. For the case of intra-system compatibility study (e.g. out-of-band emission) spectrum efficiency should be used to derive appropriate values for the radio parameters. Page 9

10 4. SYSTEM OPTIMISATION ALGORITHM The limit values are derived by means of an optimisation algorithm. For optimisation to work a criteria needs to be set. The usefuk criteria the «cost» function and optimisation process has for a task to minimise this «cost» function. The «cost» function is a function of all radio parameters and their significance to the «cost» can be altered by means of the weight coefficients. The weight coefficients can be studied in any of the following aspects in the optimisation process: - system availability - traffic capacity - spectrum utilisation - technological limitations - economic constraints The set of radio parameters values for which the «cost» function is minimised represents the optimal solution for the limit values. 5. PROPAGATION MODELS Two propagation models have been included: - Hata modified model for frequencies between 30 MHz and 3GHz - Spherical diffraction model for frequencies above 3 GHz The man-made noise and the antenna temperature noise can be considered as an increase of the thermal noise level, decreasing thus the sensitivity of a receiver, and can be entered in the simulation when the criteria of interference is I/N or C/I+N. 6. CONCLUSIONS SEAMCAT Radio Tool is the implementation of a Monte-Carlo Radio Simulation tool. The program contains integrated simulation tools for Monte Carlo type simulations and can be used variable radio system simulations. There is a possibility to add user defined functions to the tools depending on the special requirement. For the SEAMCAT simulation tool there is a User Documentation and the ERC Report 68 describes the simulation program in more details. Page 10

11 Annexes Annex 1: Annex 2: Annex 3: Annex 4: Available distributions DEE and ICE Flowchart Example of Victim link/general definition Example Application :Short Range Device Operating at 165 MHz interfering with FM broadcasting at 100 MHz. Annex 1: Available distributions Distributions: Uniform distribution Gaussian distribution Rayleigh distribution Page 11

12 Annex 2: DEE and ICE Flowchart DEE and ICE Flow-Charts RSS Arrays calculated by the EGE are processed by the DEE. Is there a big correlation? Is there one of the interferer link whose strength is dominant with respect to the others? NO YES There is a weak correlation and The «continuous» distribution which approximates the signals is not accurate enough. YES NO The arrays are still in arrays. Use Complete 1 algorithm the arrays are turn into discrete distributions. Use Complete 1 algorithm The arrays are turn into continuous distribution. Use Quick or complete 2 algorithm Complete 1 algorithm Complete 2 algorithm Quick algorithm Need explanations on integration et MC If one of the irss is in array the irss composite is calculated as a sum of the irss vectors and is interpolated in continuous distribution. If the drss is also in arrays, it is interpolated in continuous distribution otherwise it is already in distribution. Then, Seamcat calculatse the probability of interference. If RSS are in discrete distribution, we use the monte-carlo technique. SEAMCAT calculates the irss composite thanks to MC technique or integration. Computing time is the criteria to choose between the two method. Integrating :The irss composite is found by integrating all the irss distribution fonction of each It. MC sampling : the irss composite is calculated by the MC technique. Then, it derives the global probability. SEAMCAT takes N samples of the RSS continous distribution, looks if the signal is interfered, and derive the probability of being interfered. The average is 1/N. Each probability of interference is calculated for each Interfering Link independently. Then it multiplies it to obtain the whole probability. Page 12 Figure 3, DEE and ICE flow-chart

13 Annex 3: Example of Victim link/general definition Description Name: name of the victim link Description: comments on the link Victim receiver: choose in the menu a receiver already defined in the library Wanted transmitter: choose a transmitter already defined in the library User defined drss: define a distribution of the desired Received Signal Strength Notati on drss Type Unit Comments D or S (if consta nt) dbm/vr receptio n bandwi dth Call a receiver already defined in the Library, otherwise type the inputs directly. If Wanted transmitter is checked: Call a transmitter already defined in the Library otherwise type the input directly. drss is calculated taking into account all the Victim link parameters. If User-defined drss is checked: the user defines the drss distribution. tabsheets Wanted transmitter and Wt to Vr Path disappear. The Power control max threshold option is not available in this case in the Victim receiver tabsheet. Frequency f wt S MHz Center frequency of the victim link Page 13

14 Annex 4: Example Application Short Range Device Operating at 165 MHz interfering with FM broadcasting at 100 MHz. Spectrum Engineering Advanced Monte Carlo Analysis Tool Spectrum Engineering Advanced Monte Carlo Analysis Tool 1. Define Victim link FM Broadcast transmitter: 100 MHz, 1 kw Mobile receiver characteristics: C/I, Blocking, Sensitivity 2. Define Interfering link Interfering transmitter: 165 MHz, ETS Receiver definition not needed (no power control) Simulation radius: - Function of density of active transmitters and duty cycle Victim (Vrx) Wanted transmitter (Wtx) Victim (Vrx) Wanted transmitter (Wtx) Interferers (Itx) Spectrum Engineering Advanced Monte Carlo Analysis Tool Spectrum Engineering Advanced Monte Carlo Analysis Tool 3. Event Generation Model the victim link Evaluate desired receive signal strength (drss) Random location of interferers within simulation radius Evaluate interfering signal strength (unwanted/blocking/intermodulation (irss)) 4. Interference Calculation Effect of interfering transmitters unwanted emissions Effect of victim receiver blocking Intermodulation may also be evaluated Victim (Vrx) Wanted transmitter (Wtx) Victim (Vrx) Wanted transmitter (Wtx) Interferers (Itx) Interferers (Itx) Page 14

15 Homework Ollikainen For analytical evaluation of interference distribution: Consider in Fig. two similar active mobile radio systems operating on adjacent channels on 446 and MHz. The output eirp is 10 W, the receiver nominal signal level C is 59 dbm, required C/I ratio is 14 db, adjacent channel attenuation 25 db. The wanted signal path is 5 km, the unwanted interferers are uniformly distributed within 2.5 km distance from the victim receiver. 1. What is the allowed interfering signal level (in dbm). 2. Calculate at which distance the interference from a single source exceeds the wanted signal. 3. Calculate and sketch the interfering signal distribution based on uniform distance distribution from one mobile. 4. What about is the interfering signal level has a Rayleigh distribution. A= log(d)+20log(f)