Libatskaya Olga Nikolaevna
Recearch of crosstalk in cabling as a cause of jitter in digital systems of transfer.
Scientific adviser:
Dr.Sci.Tech. Vorontsov Alexander
Grigorjevich vag@fcita.dn.ua
Every year more active transition from analog to digital methods of transfer and switching is observed, that besides advantages has caused occurrence of new problems connected with principles of signaling in a digital kind. The important class of tasks are to digitize and restore signals, synchronize a network and its stability. Jitter is one of the reasons of unsynchronization and infringements of signal structure. Occurrence of jitter is caused by a number of reasons, one of which is crosstalk in a cable and as a special case a near end crosstalk (NEXT). Research NEXT and attenuations for various types of cables allows to make conclusions on suitability of their use (for the certain frequencies we find maximum length of a researched cable at which the ratio signal\handicap is comprehensible). NEXT characteristics received during experiments allow to estimate an opportunity of jitter occurrence.
In the abstract the task is to develop a technique that can estimate opportunities of using cable in digital systems. Estimation is based on an experimentally received data (attenuation and interferences), by the given technique the ratio of signal\handicap is calculated (fixed frequency) for limited by technology lengths. On the basis of the received signal\handicap ratio, which specifies a level of reliability of transfer, we judge an opportunity of using the given cable of the chosen length in digital systems.
The most important task of the work is to estimate influences of crosstalk on jitter occurrence.
Jitter or phase trembling is the phenomenon of phase modulation of an accepted signal (both analog, and digital). In practice two basic approaches to definition of jitter was extended - in terms of a phase and in terms of frequency (jitter as phase trembling, jitter as a variation of frequency of an accepted signal).
Let´s consider function of frequency instability of an accepted digital signal.
Frequency of an accepted signal is characterized by the average value f and deviation f. If modulation, at which change of frequency occurs periodically with period T, take place, then connecting frequency of modulation with the period of frequency changes fd ~ 1/Т, distinguish two types of frequency changes:
Considering jitter influence on signals parameters of quality in modern telecommunications, it is necessary to note, that this influence is shown in two directions.
There are some principal causes of jitter occurrence which influence on its structure and on parameters of quality of communication systems. Occurrence of random and deterministic jitter in system is caused by the various reasons.
Deterministic jitter is usually correlated with transmitted sequences of bites. Processes in multiplexers and regenerators can cause deterministic jitter because delays in scramblers and coders usually depend on type of accepted / transmitted sequence. Infringements in the channel of transfers can be the second reason of deterministic jitter, these infringements connected with presence of crosstalks which also bring deterministic jitter, corrilated with bites sequences. Usually deterministic jitter in this case arises because of misoperation of equalizer or infringements in adjustments of data restoration circuits and it is more common for radio-frequency systems of transfer.
Random jitter is usually caused by electromagnetic influence and an interference with external sources of a signal, such as noise, reflections, crosstalks or an interference with circuits of a feed and other sources of electromagnetic feed. In this case usually the spectrum of a signal gives the information on a source interferenced signal. The interference with circuits of a feed of 50 Hz and other low-frequency signals is usually easily identified, whereas search of an interference from computer and computing systems is connected to the analysis in a range about 60 MHz and represents the certain complexities.[3]
There are many ways of supervision and measurements influencing on the device jitter, each way is capable to clear its origin. Mentally uniting various ways it is possible to receive more full picture of an event which will help you to find the reasons of jitter and to define ways for its reduction or elimination.
There are many ways of jitter definition from eye diagrams and histograms in time area up to the analysis its frequency characteristics, allowing to divide random and determined components of full peak jitter.
The easiest and intuitively clear
way is the estimation of jitter on eye diagram. Eye diagram is a total kind of all bit periods laied on each other. It is when the image of a signal from the beginning of
the period 2 to the beginning of the period 3 is imposed on the image of a signal from
the beginnings of the period 1 to the beginning of the period 2, and so on for all bit periods. In figure 1 the typical eye diagram is shown that is equal enough and symmetric with smooth transitions (right and left
points of crossing), the big widely open "eye" giving
place for exact identification of a bit. If the trial point is located in
the center of an eye where the signal reaches the maximum or a minimum it is low probability of a bit mistake occurrence. Distance between left and right
points of crossing name an unit
interval.
The kind of eye diagram gives a lot of information about jitter and about it parameters. For example, the set of separate fronts and edges specify probable jitter presence that dependent on the data.
Eye diagram not simply gives set of information, it is convenient in use because of its simplicity and that it can be applied to measurements in circuits with the real data. For eye diagrams is not required presence of special test signal, it is possible to use measuring signal of the pulse generator. It can be effectively applied in research of the casual and pseudo-casual data, and concerns to band measurements.
Another way of jitter estimations is the histogram. The histogram represents distribution a set of values given for the measured parameter (time or value which are marked on an axis X), depending on their frequency occurrences (axis Y).
The histogram provides a level of understanding, which is not accessible for eye diagram. During troubleshooting the signal characteristic such as time of front increase and edge recession, the period and fillings factor can be displayed on the histogram. These histograms illustrate distribution of productivity for different operating modes, which can be correlated to conditions of a circuit functioning, for example, kind of transmitted sequence.
Main application of histograms is distribution of frequency for values of time interval error (TIE) for all bit transitions of a measured signal. TIE is a difference in time between the valid and expected points crossings on eye diagram. The histogram of TIE values is the basic data set for procedures of jitter allocation, required various standards of digital trunks.
In figure 2 is shown eye diagram and TIE histogram. Eye diagram is displaced
so that in the center of transition area (a point of crossing) was visible between
two "eyes". On the diagram two separate lines of fronts and edges are traced it means that presence of determined jitter. But, these lines are
dim, that confirm the presence of casual jitter.
The histogram of transition points on eye diagram has two maximum, that
deforms Gauss curve. This also means that the signal has determined and random components of jitter.
The next way of jitter estimation consists on construction of a U-shaped curve (bathtub curve). It represents dependence mistakes on bit (BER) from
positions of a trial point on an unit interval (UI). Usually the curve
represents in logarithmic scale to reduce an inclination of
curve.
When the trial point is near a point of transition, BER=0,5 (equal probability of correct or not correct bit definition). In this area a curve rather flat and here the mechanism of determined jitter prevails. As a trial point goes to the center of unit interval BER decreases. In this area the mechanism of casual jitter prevails and BER is defined as average square-law rejection from Gauss processes, determining casual jitter. It is expected, that optimum position of trial point there will be a center of an unit interval.
The sides of U-shape curve will show borders of correct transfer at chosen allowable level of BER. As further a left side of a curve from right that more stability of jitter in the developed system. These edges correspond with "tails" of Gauss function constructed on the basis of TIE histogram. The U-shaped curve also can be used for division random and determined jitter and definitions of average square-law rejection for random components.
Consideration of jitter in frequency area is one more way of definition its reasons. Sources of determined jitter in frequency area are shown as discrete spectrum. During the frequency analysis of phase noise or a spectrum of jitter phase noise or jitter corresponds with shift of frequency concerning carrying frequency or synchronization.
Measurements of phase noise provide the most exact estimation of jitter due to obviously high frequency of digitization of a signal and management of frequencies range. With their help it is possible to understand the processes occuring in the developed device, using quartz generators and phase frequency trims it is easy to define determined jitter by peaks on a spectrum. They are useful in optimization of signal restoration circuits and detection of internal handicapes and noise sources.
One more method of jitter consideration in frequency area is application of fast transformation (FFT) to values of a mistake in time interval TIE. Method FFt is not acceptable to measure weak phase noise, but it is necessary for fast and simple viewing of obvious processes.
In figure 4 is shown a range of the same
signal representations – synchronizing pulses with frequency of 456 MHz which are shown
on the top oscillogram. On the second line the histogram of a transition point is shown. It is obvious, that the histogram differs from Gauss function that
confirms the presence of both determined and random
jitter. The third line draws behaviour of a time
interval TIE mistake in time; it would be a straight line if jitter was absent.
And, at last, below we can see a jitter spectrum received with the FFT help. The peak in the center means jitter presence in a synchronization circuit on subharmonic frequency of 114 MHz (one fourth from frequencies of clock sequence). Such form of spectrum is caused by determined jitter. Even at infinite clock the peak amplitude on the curve will not grow current time. The same peak bears the responsibility for asymmetry of the histogram and periodicity of TIE curve.
Small "hump" is much less obvious on the left side on frequencies from 0 up to 10 MHz. Eventually it will be grow and, finally, will exceed size of the central peak, that characterizes its nature as random noise.
Jitter division on components is not one way of its measurement, but it very important in practice – as for finding the reasons of malfunctions as for an estimation of development reliability. If you can separate determined jitter and then count behaviour of average square-law rejection of random jitter you can quickly estimate frequency of mistakes on bits (BER) and define borders serviceability of a devise, not resorting to long measurements, which are required the order of 10-12 and 95 % of reliability.
The U-shaped curve gives one more way of division a random and determined jitter. The top part of curve (where prevails determined jitter) falls downwards at BER about 9-10. On points on an inclined site it is possible to make approximation curve and to estimate parameters describing it. One of these parameters will be average square-law rejection of Gauss function.
The third way of divisions jitter components is application of FFT to TIE. The lines caused determined jitter are left from received spectrum then the return transformation FFT is made. As a result we shall receive random jitter without determined components. [8]
Quality and length of communication on CTN lines are defined by electric characteristics of elements on a path of transfer: parameters of transfer, mutual influences and noise in a line. Primary parameters of transfer are resistance, electric capacity, inductance and conductivity. Secondary parameters of transfer are a factor of distribution γ and wave resistance Zв. The factor of distribution depends on primary parameters of circuit R, C, L and G, being complex sized values, consists of two components: α- the factor of attenuation of a circuit determining reduction of a voltage, a current or capacity on one kilometer of a circuit, and factor of a phase β - taking into account changes phase, current or voltage per kilometer.
The factor of distribution is defined as
Attenuation of a homogeneous circuit, dB, is possible to express as the attitude of currents, voltage and capacities in the beginning and at the end of a circuit:
Change on a phase:
In previous expressions Uн, Iн, Uк, Iк - complex values of voltage and current in the beginning and the end of a circuit; φ, ψ - phases of a voltage and current in the end and in the beginning of a circuit.
Parameters of influence between cable lines circuits define a degree of transitive distinct conversations and noise in telephone circuits from the next circuits (fig. 1). Distinguish two transitions of electromagnetic energy: on near and far ends. The influence, shown on the end of a circuit where the generator is located, is defined by transitive attenuation, dB, on near end:
where Р10 - capacity of the generator on near end of an influencing circuit; Р20 - capacity of handicapes on the near end of the circuit that is aim of influence. In case of identical circuits (Zв1=Zв2) the size of transitive attenuation on near end, dB, can be expressed with corresponding voltage and currents:
Influence on opposite, removed from the generator, end is defined by transitive attenuation on far end, dB:
where Р10 - capacity of the generator included in the influencing circuit on the nearest end; Р2l - capacity in a circuit that is an aim of influence on the far end.
For identical circuits transitive attenuation on the distant end, dB:
In technics of communication is more used the parameter of influence - the security determined as a difference between capacity levels of useful signal and handicapes: Аз = рc-pn. For circuits with identical parameters security Аз on the far end is equal to a difference between transitive attenuation on the far end and own attenuation of a circuit:
Security between circuits on the far end, dB, can be found from expression:
The reason of influence occurrence between circuits is the cross-section electromagnetic field, i.e. a field which power lines are located in a plane perpendicular to a direction of signaling. The result of electric and magnetic fields which are components of an electromagnetic field is total action of electric and magnetic influences - electromagnetic influence. This is the mechanism of mutual influence between two two-wire circuits.
Electric connection is the attitude of handicapes current in a circuit, which is an aim of influence I2, to a potential difference in the beginning of influencing circuit U1:
Magnetic connection:
Generally С12 and M12, representing the attitude of currents and voltage in complex size, are also complex:
where g12 - an active component of electric connection, Сим; С12 - a jet (capacitor) component of electric connection, Ф; г12 - an active component of magnetic connection, Ohm; m12 - a jet (inductive) component of magnetic connection, Гн.
The standard expressions:
In technics of communication the influence shown on the end where the generator is located, name influence on the near end, and the influence shown on the opposite end - influence on the far end.
From fig. 2 follows that the currents acting on the near end caused by electric and magnetic connections which have an identical direction and develop; the same currents acting on the distant end have an opposite direction and are subtracted. Thus, on the near end action of magnetic and electric connection is summarized, and on the far end operates their difference.
The factor of electromagnetic connection on the near and far end:
The degree of mutual influences between circuits is defined by transitive attenuation on near and far ends. Interrelation between values of transitive attenuation and electromagnetic connections on near and far ends.[4]
Formulas do not take into account the factors influencing on work of a cable in non-standard conditions. So it is inexpedient to carry out modelling on the basis of the theoretical data because of presence of significant errors which are the reason of an inexact estimation. This error then will be transferred in a technique of definition the maximum length and the suitability of using the certain type of a cable on ratio signal / noise. According to this decision to lead an experimental research was accepted. The data received in such way, will contain a tool error, but such error can be taken into account and be made less than set errors.
The technique of attenuation and influence measurement on near end NEXT (near end crosstalk) with use of analyzer HP 8753C has been chosen. The given analyzer of circuits allows to measure S-parameters. Parameter S21 (factor of direct transfer) was measured. Measurements of necessary parameter in analyzer can be carried out in a range of frequencies from 300 kHz up to 6 GHz. The frequencies range for researches was chosen from 300kHz (the minimal opportunity border in analyzer) up to 70 MHz.
Measurements of attenuations and interferences were carried out for two types of cables: UTP 5 (not shielded twisted pair) and telephone ТПП in the lengths 21,5м and 25м accordingly. This cables have structure of symmetric twisted pairs and 100 Ohm wave resistance while the analyzer has asymmetrical inputs, resistance 50 Ohm. According to this we decided to create a devices that help to solve the problem. For coordination the resistance in 50 and 100 Ohm the ratio 1:1,41 (a square root of 2) is necessary.
The given ratio has been chosen on the basis of the following reasons and calculations. The network analyzer with the cable connected to it and loading can be presented as the following equivalent circuit.
Where Zн, Zi complex values of loadings and internal resistance. The condition of the coordination in this case is submitted by mentioned below conditions:
If xi< The condition of the coordination will be the following after the accepted simplifications. (Ri=Rн) The condition of the coordination: The agreeing device represents the transformer working in this range of frequencies as a line with distributed parameters. The given agreeing device is executed on a basis of ferrite cores 30ВЧ K 7*5*2 (a working range of frequencies up to 120 MHz). On a ferrite ring 20 coils of a wire which represents 5 densely twisted with each other wires ПЭЛШО-0,1 are reeled up, such dense twist provides interaction between separate wires, as in a line with the distributed parameters. By virtue of design features (the quantity of twisted wires is limited by the attitude of integers) exact maintenance of necessary factor of transformation 1:1,41 is not obviously possible. The agreeing device, showed below, provides a ratio 1:1,5, that can be a reason of an error. Discrepancy of the necessary and received ratio gives the basis for the assumption that the received results will contain periodicity, characteristic for a mismatch in lines with the distributed parameters. where К1-К5-the one ends of twisted wires;
Н1-Н5-the another ends of othertwisted wires.
Before agreeing devices were used in research we have estimated their properties in the set range of frequencies. Were researced the attenuation brought on pair by identical agreeing devices. Researches carried out under the mentioned below scheme. As a result the following dependence is received. Using the analyzer HP 8753C, the received agreeing device, pieces of researched cables, carried out the measurements.
In result the following dependences on frequency are received: the attenuation brought by agreeing devices; attenuation of a cable; crosstalk from one pair to another. As it was mentioned above the use of agreeing device with characteristics distinct from necessary cause a periodicity that specifies presence of a small mismatch. Next the attenuation brought by agreeing devices were taken into account. For each cable were found the average value with the certain degree of reliability. Through the maximal points of a deviation from the received average curve pass allowable levels of a deviation. On frequencies over 60 MHz, levels of a deviation are great so they can't be used. After the average attenuation in a cable (length is 21,5 m) is determined it is necessary to define attenuation in a cable of the same mark for length 1m, that further will allow to define attenuation in the given cable for lengths limited by technology. Attenuation for 1m in 21,5 times less than the received result, we translate the given value in logarithmic scale and receive required attenuation. The same calculations were done for cable ТПП . Was defined crosstalks for two various cables. Further we carry out the actions described above to define crosstalks for 1m of a cable, that further will allow to carry out recalculation for lengths of this type of a cable. Next was finding the necessary data for definition ratio signal / noise that will allow to define the maximal building length of a cable (at the certain range of frequencies).
Was chosen frequency equal 20 MHz. For the given fixed frequency, using the received curves of attenuations in a cable (for 1m), was constructed a curve which represents dependence of attenuation in a cable from length.
Similar curve was received for dependence of crosstalk in a cable from its length. On the basis of the received data it is possible to receive a ratio signal / noise using mentioned below formulas. The given ratio will allow to define: what maximal length of UTP 5 is possible for using on 20 MHz. The given value allows to estimate suitability of using the certain type of a cable with length l. For each researched cable for the certain frequency it is possible to calculate a limit of used lengths, cables with long lengths will have unacceptable ratio signal / noise and therefore will be not suitable. The same curves was received for cable ТПП.[1] During the experimental researches the following data have been received: attenuation and crosstalk for chosen cables. These characteristics were calculated for each cable (length 1m) using received data and technique. The received data gave an opportunity to calculate for the fixed frequency the same characteristics, but for various lengths of a cable. On the basis of these data it was possible to calculate a ratio signal / noise that help to make a conclusion on suitability of a cable for using in channels. The received technique allows to find maximum length of a cable for the certain frequency at which the ratio signal / noise will be comprehensible.
http://www.iol.unh.edu/training/tokenring/
http://www.unitest.com/theory/jitter-pr.html
http://www.ecolan.ru/alien.htm
http://www.gsc.com.ua/pages/?pid=ttwpair