JITTER DETERMINATION METHOD AND MEASUREMENT INSTRUMENT

Abstract

A jitter determination method for determining at least one random jitter component of an input signal is described, wherein the input signal is generated by a signal source, comprising: receiving the input signal; determining a time interval error associated with the random jitter component; determining at least one statistical moment of the time interval error based on the determined time interval error, wherein the order of the statistical moment is two or larger; at least one of determining an impulse response based on the input signal and receiving the impulse response, the impulse response being associated with at least the signal source; and determining the standard deviation of the random jitter component based on at least one of the determined statistical moment and the determined impulse response. Moreover, a measurement instrument is described.

Description

FIELD OF THE DISCLOSURE

Embodiments of the present disclosure generally relate to a jitter determination method for determining at least one jitter component of an input signal. Further, embodiments of the present disclosure generally relate to a measurement instrument.

BACKGROUND

For jitter analysis, the components of jitter such as Data Dependent Jitter (DDJ), Periodic Jitter (PJ), Other Bounded Uncorrelated Jitter (OBUJ) and Random Jitter (RJ) must be separated and the bit error rate (BER) must be calculated.

So far, techniques are known that exclusively relate on determining the Time Interval Error (TIE) of the Total Jitter (TJ). In fact, the causes of the different jitter types lead to a distortion of the received signal and they, therefore, have an influence on the TIE via the received signal. Accordingly, the respective components of jitter are calculated once the Time Interval Error (TIE) of the Total Jitter (TJ) is determined.

However, it turned out that the measurement time is long if a high accuracy is to be achieved. Put another way, the signal length of the signal to be analyzed is long resulting in a long measuring duration if high precision is aimed for.

Moreover, the respective components of jitter are obtained by averaging operations. For instance, the Data Dependent Jitter (DDJ) is estimated by averaging the Time Interval Error (TIE) of the Total Jitter (TJ), namely a DDJ eye diagram or the DDJ worst case eye diagram. Moreover, certain components of jitter cannot be determined in a reliable manner.

Accordingly, there is a need for a fast and reliable possibility to determine a jitter component of an input signal, particularly the Random Jitter (RJ).

SUMMARY

Embodiments of the present disclosure provide a jitter determination method for determining at least one random jitter component of an input signal, wherein the input signal is generated by a signal source. In an embodiment, the jitter determination method comprises the following steps:

receiving the input signal;

determining a time interval error associated with the random jitter component;

determining at least one statistical moment of the time interval error based on the determined time interval error, wherein the order of the statistical moment is two or larger;

at least one of determining an impulse response based on the input signal and receiving the impulse response, the impulse response being associated with at least the signal source; and

determining the standard deviation of the random jitter component based on at least one of the determined statistical moment and the determined impulse response.

Accordingly, the standard deviation of the RJ is determined based on the determined statistical moment and/or based on the determined impulse response. In some embodiments, the standard deviation of the RJ is determined based on both the determined statistical moment and the determined impulse response.

As the RJ is usually normal-distributed and has an expected value of zero, the standard deviation is enough information to completely define a probability density function of the RJ. Accordingly, the RJ component of the TJ can be completely determined by determining the standard deviation of the RJ.

In some embodiments, a step response being associated with at least the signal source is determined based on the input signal. The impulse response can then be determined by differentiating the determined step response, for example with respect to time or frequency, depending on the domain under consideration.

The input signal may be received via a transmission channel attached to the signal source. Thus, the impulse response may be associated with both the signal source and the transmission channel.

More specifically, the signal source generates an electrical signal that is then transmitted via the transmission channel and may be probed by a probe, for example a tip of the probe. In fact, the electrical signal generated by the signal source is forwarded via the transmission channel to a location where the probe, for example its tip, can contact the device under test in order to measure the input signal.

Thus, the electrical signal may generally be sensed between the signal source and the signal sink assigned to the signal source, wherein the electrical signal may also be probed at the signal source or the signal sink directly. Accordingly, the impulse response may be associated with the signal source and with the transmission channel up to the probe.

The signal source and/or the transmission channel exhibit a certain memory range, which is a number of sequential symbols that can interact with one another, for example perturb one another. When determining the at least one statistical moment, the correlation of the time interval error evaluated at a certain time and the time interval error evaluated at a shifted time may be determined, wherein the functions describing the two time interval errors each have a memory range. These functions represent the random jitter component of the input signal. For the calculation of the statistical moment, only an overlap region of the two memory ranges of the two time interval errors needs to be taken into account.

According to an embodiment of the present disclosure, a linearized mathematical relation is used to determine the statistical moment, wherein the mathematical relation correlates the time interval error with a random jitter component of the input signal and to a slope of a data dependent jitter signal.

In another embodiment of the present disclosure, the at least one statistical moment comprises at least one of an autocorrelation function of the time interval error and a power spectral density.

According to an aspect of the disclosure, at least one conditional autocorrelation function is determined in order to determine the autocorrelation function. Therein, the conditional autocorrelation function is the autocorrelation function for a particular realization of the overlap region of the memory ranges of the functions describing the time interval error and the shifted time interval error. In other words, a particular autocorrelation function is calculated under the constraint that the functions describing the time interval errors have a particular bit sequence in the overlap region of their memory ranges.

According to another aspect of the disclosure, the at least one conditional autocorrelation function is determined by calculating an expected value of a product of a function representing the random jitter component with the same function shifted by a shifting parameter. As described above, these functions each exhibit a certain memory range, and only an overlap region of these two memory ranges contributes to the expected value and thus to the autocorrelation function.

In a particular embodiment of the disclosure, at least two conditional autocorrelation functions are summed with appropriate joint probability factors being the coefficients of the conditional autocorrelation functions in order to determine the autocorrelation function. The joint probability factors represent the probability for the respective particular realization of the overlap region of the memory ranges of the functions describing the time interval error and the shifted time interval error to occur. The joint probability factors may also be called joint probability densities. Thus, the at least two conditional autocorrelation functions are weighted with the appropriate probabilities and summed up. In some embodiments, a sum over all conditional autocorrelation functions weighted with the appropriate joint probability factors may be performed.

The joint probability factors may be at least one of preset and calculated based on possible permutations of a symbol sequence contained within the input signal. In some embodiments, the joint probability factors can be calculated combinatorically.

According to a further aspect, at least one of a data dependent jitter signal and a slope of the data dependent jitter signal is determined in order to determine the time interval error. More precisely, the time interval error associated with the random jitter can be determined as the negative inverse of the slope of the data dependent jitter signal in time domain times a function representing the amplitude perturbation that is caused by the temporal RJ.

According to another embodiment of the disclosure, the at least one of the data dependent jitter signal and the slope of the data dependent jitter signal is evaluated at a clock time associated with a clock signal or at an edge time associated with a signal edge. Evaluating the data dependent jitter signal and/or the slope of the data dependent jitter signal at the edge time associated with the signal edge can lead to a higher accuracy of the determined autocorrelation function. Accordingly, the edge time or rather the several edge times being associated with several signal edges may be detected beforehand.

According to a further aspect, the input signal is PAM-n coded, wherein n is an integer bigger than 1. Accordingly, the method is not limited to binary signals (PAM-2 coded) since any kind of pulse-amplitude modulated signals may be processed. In some embodiments, the input signal is established as a bit stream, i.e. as a PAM-2 coded signal.

The input signal may be decoded, thereby generating a decoded input signal. In other words, the input signal is divided into the individual symbol intervals and the values of the individual symbols (“bits”) are determined.

Thereby and in the following the term “decoding” is understood to mean that a symbol sequence contained in the input signal is determined based on the input signal. For example, if the input signal is PAM-n coded, the input signal contains a sequence of symbols having a certain average symbol duration and one of n different signal level values. For the case of a PAM-2 coded signal, the symbol sequence contained in the input signal is a bit sequence, i.e. each symbol is a bit having one of two possible values. Thus, the decoded input signal contains the determined symbol sequence.

Based on the determined symbol sequence comprised in the input signal, the impulse response being associated with at least the signal source is determined.

In some embodiments, a step response being associated with at least the signal source is initially determined based on the decoded input signal. The impulse response can then be determined by differentiating the determined step response with respect to time or frequency, depending on the domain under consideration.

The step of decoding the input signal may be skipped if the input signal comprises an already known bit sequence. For example, the input signal may be a standardized signal such as a test signal that is determined by a communication protocol. In this case, the input signal does not need to be decoded, as the bit sequence contained in the input signal is already known.

Embodiments of the present disclosure also provide a measurement instrument for determining at least one random jitter component of an input signal, comprising at least one input channel and an analysis circuit or module being connected to the at least one input channel. The measurement instrument is configured to receive an input signal via the input channel and to forward the input signal to the analysis module. The analysis module is configured to determine a time interval error associated with the random jitter component. The analysis module is configured to determine at least one statistical moment of the time interval error based on the determined time interval error, wherein the order of the statistical moment is two or larger. The analysis module is configured to at least one of determine an impulse response based on the input signal and receive the impulse response, the impulse response being associated with at least the signal source. The analysis module further is configured to determine the standard deviation of the random jitter component based on at least one of the determined statistical moment and the determined impulse response.

Accordingly, the standard deviation of the RJ is determined by the measurement instrument based on the determined statistical moment and/or based on the determined impulse response. In some embodiments, the standard deviation of the RJ is determined by the measurement instrument based on both the determined statistical moment and the determined impulse response.

As the RJ is usually normal-distributed and has an expected value of zero, the standard deviation is enough information to completely define a probability density function of the RJ. Accordingly, the RJ component of the TJ can be completely determined by the measurement instrument by determining the standard deviation of the RJ.

In some embodiments, a step response being associated with at least the signal source is determined by the measurement instrument based on the input signal. The impulse response can then be determined by the measurement instrument by differentiating the determined step response, for example with respect to time or frequency, depending on the domain under consideration.

The measurement instrument is configured to determine the impulse response being associated with at least the signal source based on the determined symbol sequence comprised in the input signal.

In some embodiments, the measurement instrument is configured to perform the jitter determination method described above.

According to one embodiment of the present disclosure, the at least one statistical moment comprises at least one of an autocorrelation function of the time interval error and a power spectral density.

According to an aspect of the disclosure, the analysis module is configured to determine at least one conditional autocorrelation function in order to determine the autocorrelation function. Therein, the conditional autocorrelation function is the autocorrelation function for a particular realization of the overlap region of the memory ranges of the functions describing the time interval error and the shifted time interval error. In other words, the measurement instrument is configured to calculate a particular autocorrelation function under the constraint that the functions describing the time interval errors have a particular bit sequence in the overlap region of their memory ranges.

In a particular embodiment of the disclosure, the analysis module is configured to determine the at least one conditional autocorrelation function by calculating an expected value of a product of a function representing the random jitter component with the same function shifted by a shifting parameter. As described above, these functions each exhibit a certain memory range, and only an overlap region of these two memory ranges contributes to the expected value and thus to the autocorrelation function.

In another embodiment of the disclosure, the analysis module is configured to sum at least two conditional autocorrelation functions with appropriate joint probability factors being the coefficients of the conditional autocorrelation functions in order to determine the autocorrelation function. The joint probability factors represent the probability for the respective particular realization of the overlap region of the memory ranges of the functions describing the time interval error and the shifted time interval error to occur. The joint probability factors may also be called joint probability densities. Thus, the at least two conditional autocorrelation functions are weighted with the appropriate probabilities and summed up by the analysis module. In some embodiments, a sum over all conditional autocorrelation functions weighted with the appropriate joint probability factors may be performed by the analysis module.

The analysis module may be configured to calculate the joint probability factors based on possible permutations of a symbol sequence contained within the input signal.

In some embodiments, the measurement instrument comprises an interface via which a user may preset the joint probability factors.

According to a further aspect, the analysis module is configured to determine at least one of a data dependent jitter signal and a slope of the data dependent jitter signal in order to determine the time interval error. More precisely, the analysis module may be configured to determine the time interval error associated with the random jitter to be the negative inverse of the slope of the data dependent jitter signal in time domain times a function representing the amplitude perturbation that is caused by the temporal RJ.

In another embodiment of the disclosure, the analysis module is configured to evaluate the at least one of the data dependent jitter signal and the slope of the data dependent jitter signal a clock time associated with a clock signal or at an edge time associated with a signal edge. Evaluating the data dependent jitter signal and/or the slope of the data dependent jitter signal at the edge time associated with the signal edge can lead to a higher accuracy of the determined autocorrelation function. Accordingly, the analysis module may be configured to determine the edge time or rather the several edge times being associated with several signal edges beforehand.

The measurement instrument may be established as at least one of an oscilloscope, a spectrum analyzer and a vector network analyzer. Thus, an oscilloscope, a spectrum analyzer and/or a vector network analyser may be provided that is enabled to perform the jitter determination method described above for determining at least one jitter component of an input signal.

The analysis module may be configured to decode the input signal such that a decoded input signal is generated. In other words, the analysis module divides the input signal into the individual symbol intervals and determined the values of the individual symbols (“bits”).

The step of decoding the input signal may be skipped if the input signal comprises an already known bit sequence. For example, the input signal may be a standardized signal such as a test signal that is determined by a communication protocol. In this case, the input signal does not need to be decoded, as the bit sequence contained in the input signal is already known.

The analysis module can comprise hardware and/or software portions, hardware and/or software modules, etc. Some of the method steps may be implemented in hardware and/or in software. In some embodiments, one or more of the method steps may be implemented, for example, only in software. In some embodiments, all of the method steps are implemented, for example, only in software. In some embodiments, the method steps are implemented only in hardware.

DESCRIPTION OF THE DRAWINGS

The foregoing aspects and many of the attendant advantages of the claimed subject matter will become more readily appreciated as the same become better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings, wherein:

FIG. 1 schematically shows a representative measurement system with an example measurement instrument according to an embodiment of the disclosure;

FIG. 2 shows a tree diagram of different types of jitter and different types of noise;

FIG. 3 shows a representative flow chart of a jitter determination method according to an embodiment of the disclosure;

FIG. 4 shows a representative flow chart of a signal parameter determination method according to an embodiment of the disclosure;

FIGS. 5A-5D show example histograms of different components of a time interval error;

FIG. 6 shows a representative flow chart of a method for separating random jitter and horizontal periodic jitter according to an embodiment of the disclosure;

FIGS. 7A and 7B show diagrams of jitter components plotted over time;

FIG. 8 shows a schematic representation of a representative method for determining an autocorrelation function of random jitter according to an embodiment of the disclosure;

FIG. 9 shows an overview of different autocorrelation functions of jitter components;

FIG. 10 shows an overview of different power spectrum densities of jitter components;

FIG. 11 shows an overview of a bit error rate determined, a measured bit error rate and a reference bit error rate;

FIG. 12 shows an overview of a mathematical scale transformation of the results of FIG. 11; and

FIG. 13 shows an overview of probability densities of the random jitter, the other bounded uncorrelated jitter as well as a superposition of both.

DETAILED DESCRIPTION

The detailed description set forth below in connection with the appended drawings, where like numerals reference like elements, is intended as a description of various embodiments of the disclosed subject matter and is not intended to represent the only embodiments. Each embodiment described in this disclosure is provided merely as an example or illustration and should not be construed as preferred or advantageous over other embodiments. The illustrative examples provided herein are not intended to be exhaustive or to limit the claimed subject matter to the precise forms disclosed.

FIG. 1 schematically shows a measurement system 10 comprising a measurement instrument 12 and a device under test 14. The measurement instrument 12 comprises a probe 16, an input channel 18, an analysis circuit or module 20 and a display 22.

The probe 16 is connected to the input channel 18 which in turn is connected to the analysis module 20. The display 22 is connected to the analysis module 20 and/or to the input channel 18 directly. Typically, a housing is provided that encompasses at least the analysis module 20.

Generally, the measurement instrument 12 may comprise an oscilloscope, as a spectrum analyzer, as a vector network analyzer or as any other kind of measurement device being configured to measure certain properties of the device under test 14.

The device under test 14 comprises a signal source 24 as well as a transmission channel 26 connected to the signal source 24.

In general, the signal source 24 is configured to generate an electrical signal that propagates via the transmission channel 26. In some embodiments, the device under test 14 comprises a signal sink to which the signal generated by the signal source 24 propagates via the transmission channel 26.

More specifically, the signal source 24 generates the electrical signal that is then transmitted via the transmission channel 26 and probed by the probe 16, for example a tip of the probe 16. In fact, the electrical signal generated by the signal source 24 is forwarded via the transmission channel 26 to a location where the probe 16, for example its tip, can contact the device under test 14 in order to measure the input signal. Thus, the electrical signal may generally be sensed between the signal source 24 and the signal sink assigned to the signal source 24, wherein the electrical signal may also be probed at the signal source 24 or the signal sink directly. Put another way, the measurement instrument 12, particularly the analysis module 20, receives an input signal via the probe 16 that senses the electrical signal.

The input signal probed is forwarded to the analysis module 20 via the input channel 18. The input signal is then processed and/or analyzed by the analysis module 20 in order to determine the properties of the device under test 14.

Therein and in the following, the term “input signal” is understood to be a collective term for all stages of the signal generated by the signal source 24 that exist before the signal reaches the analysis module 20. In other words, the input signal may be altered by the transmission channel 26 and/or by other components of the device under test 14 and/or of the measurement instrument 12 that process the input signal before it reaches the analysis module 20. Accordingly, the input signal relates to the signal that is received and analyzed by the analysis module 20.

The input signal usually contains perturbations in the form of total jitter (TJ) that is a perturbation in time and total noise (TN) that is a perturbation in amplitude. The total jitter and the total noise in turn each comprise several components. Note that the abbreviations introduced in parentheses will be used in the following.

As is shown in FIG. 2, the total jitter (TJ) is composed of random jitter (RJ) and deterministic jitter (DJ), wherein the random jitter (RJ) is unbounded and randomly distributed, and wherein the deterministic jitter (DJ) is bounded.

The deterministic jitter (DJ) itself comprises data dependent jitter (DDJ), periodic jitter (PJ) and other bounded uncorrelated jitter (OBUJ).

The data dependent jitter is directly correlated with the input signal, for example directly correlated with signal edges in the input signal. The periodic jitter is uncorrelated with the input signal and comprises perturbations that are periodic, particularly in time. The other bounded uncorrelated jitter comprises all deterministic perturbations that are neither correlated with the input signal nor periodic. The data dependent jitter comprises up to two components, namely inter-symbol interference (ISI) and duty cycle distortion (DCD).

Analogously, the total noise (TN) comprises random noise (RN) and deterministic noise (DN), wherein the deterministic noise contains data dependent noise (DDN), periodic noise (PN) and other bounded uncorrelated noise (OBUN).

Similarly to the jitter, the data dependent noise is directly correlated with the input signal, for example directly correlated with signal edges in the input signal. The periodic noise is uncorrelated with the input signal and comprises perturbations that are periodic, particularly in amplitude. The other bounded uncorrelated noise comprises all deterministic perturbations that are neither correlated with the input signal nor periodic. The data dependent noise comprises up to two components, namely inter-symbol interference (ISI) and duty cycle distortion (DCD).

In general, there is cross-talk between the perturbations in time and the perturbations in amplitude.

Put another way, jitter may be caused by “horizontal” temporal perturbations, which is denoted by “(h)” in FIG. 2 and in the following, and/or by “vertical” amplitude perturbations, which is denoted by a “(v)” in FIG. 2 and in the following.

Likewise, noise may be caused by “horizontal” temporal perturbations, which is denoted by “(h)” in FIG. 2 and in the following, and/or by “vertical” amplitude perturbations, which is denoted by a “(v)” in FIG. 2 and in the following.

In detail, the terminology used below is the following:

Horizontal periodic jitter PJ(h) is periodic jitter that is caused by a temporal perturbation.

Vertical periodic jitter PJ(v) is periodic jitter that is caused by an amplitude perturbation.

Horizontal other bounded uncorrelated jitter OBUJ(h) is other bounded uncorrelated jitter that is caused by a temporal perturbation.

Vertical other bounded uncorrelated jitter OBUJ(v) is other bounded uncorrelated jitter that is caused by an amplitude perturbation.

Horizontal random jitter RJ(h) is random jitter that is caused by a temporal perturbation.

Vertical random jitter RJ(v) is random jitter that is caused by an amplitude perturbation.

The definitions for noise are analogous to those for jitter:

Horizontal periodic noise PN(h) is periodic noise that is caused by a temporal perturbation.

Vertical periodic noise PN(v) is periodic noise that is caused by an amplitude perturbation.

Horizontal other bounded uncorrelated noise OBUN(h) is other bounded uncorrelated noise that is caused by a temporal perturbation.

Vertical other bounded uncorrelated noise OBUN(v) is other bounded uncorrelated noise that is caused by an amplitude perturbation.

Horizontal random noise RN(h) is random noise that is caused by a temporal perturbation.

Vertical random noise RN(v) is random noise that is caused by an amplitude perturbation.

As mentioned above, noise and jitter each may be caused by “horizontal” temporal perturbations and/or by “vertical” amplitude perturbations.

The measurement instrument 12 or rather the analysis module 20 is configured to perform the steps schematically shown in FIGS. 3, 4 and 6 in order to analyze the jitter and/or noise components contained within the input signal, namely the jitter and/or noise components mentioned above.

In some embodiments, one or more computer-readable storage media is provided containing computer readable instructions embodied thereon that, when executed by one or more computing devices (contained in or associated with the measurement instrument 12, the analysis module 20, etc.), cause the one or more computing devices to perform one or more steps of the method of FIGS. 3, 4, 6, and/or 8 described below. In other embodiments, one or more of these method steps can be implemented in digital and/or analog circuitry or the like.

Model of the Input Signal

First of all, a mathematical substitute model of the input signal or rather of the jitter components and the noise components of the input signal is established. Without loss of generality, the input signal is assumed to be PAM-n coded in the following, wherein n is an integer bigger than 1. Hence, the input signal may be a binary signal (PAM-2 coded).

Based on the categorization explained above with reference to FIG. 2, the input signal at a time t/T_b is modelled as

$\begin{array}{cc}{x}_{\mathrm{TN}}\left(t/T_b\right)=\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}b\left(k\right)·h\left(t/T_b-k-\varepsilon \left(k\right)/T_b\right)+\sum _{i=0}^{N_{{\mathrm{PN}\left(v\right)}^{-1}}}A_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right)& \left(E\mathrm{.1}\right)\end{array}$

In the first term, namely

$\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}b\left(k\right)·h\left(t/T_b-k-\varepsilon \left(k\right)/T_b\right),$

b(k) represents a bit sequence sent by the signal source 24 via the transmission channel 26, wherein T_b is the bit period.

Note that strictly speaking the term “bit” is only correct for a PAM-2 coded input signal. However, the term “bit” is to be understood to also include a corresponding signal symbol of the PAM-n coded input signal for arbitrary integer n.

h(t/T_b) is the joint impulse response of the signal source 24 and the transmission channel 26. In case of directly probing the signal source 24, h(t/T_b) is the impulse response of the signal source 24 since no transmission channel 26 is provided or rather necessary.

Note that the joint impulse response h(t/T_b) does not comprise contributions that are caused by the probe 16, as these contributions are usually compensated by the measurement instrument 12 or the probe 16 itself in a process called “de-embedding”. Moreover, contributions from the probe 16 to the joint impulse response h(t/T_b) may be negligible compared to contributions from the signal source 24 and the transmission channel 26.

N_pre and N_post respectively represent the number of bits before and after the current bit that perturb the input signal due to inter-symbol interference. As already mentioned, the length N_pre+N_post+1 may comprise several bits, particularly several hundred bits, especially in case of occurring reflections in the transmission channel 26.

Further, ε(k) is a function describing the time perturbation, i.e. ε(k) represents the temporal jitter.

Moreover, the input signal also contains periodic noise perturbations, which are represented by the second term in equation (E.1), namely

$\sum _{i=0}^{N_{{\mathrm{PN}\left(v\right)}^{-1}}}A_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right).$

The periodic noise perturbation is modelled by a series over N_PN(v) sine-functions with respective amplitudes A_i, frequencies f_i and phases ϕ_i, which is equivalent to a Fourier series of the vertical periodic noise.

The last two terms in equation (E.1), namely

+x_RN(v)(t/T_b)+x_OBUN(v)(t/T_b)

represent the vertical random noise and the vertical other bounded uncorrelated noise contained in the input signal, respectively.

The function ε(k) describing the temporal jitter is modelled as follows:

$\begin{array}{cc}\varepsilon \left(k\right)/T_b=\sum _{i=0}^{N_{PJ\left(h\right)}-1}a_i/T_b·\sin \left(2\pi ·{\vartheta }_i/f_b·k+{\varphi }_i\right)+{\varepsilon }_{RJ}\left(k\right)/T_b+{\varepsilon }_{\mathrm{OBUJ}}\left(k\right)/T_b.& \left(E\mathrm{.2}\right)\end{array}$

The first term in equation (E.2), namely

$\sum _{i=0}^{N_{PJ\left(h\right)}-1}a_i/T_b·\sin \left(2\pi ·{\vartheta }_i/f_b·k+{\varphi }_i\right),$

represents the periodic jitter components that are modelled by a series over N_PJ(h) sine-functions with respective amplitudes α_i, frequencies ϑ_i and phases φ_i, which is equivalent to a Fourier series of the horizontal periodic jitter.

The last two terms in equation (E.2), namely

ε_RJ(k)/T_b+ε_OBUJ(k)/T_b

represent the random jitter and the other bounded uncorrelated jitter contained in the total jitter, respectively.

In order to model duty cycle distortion (DCD), the model of (E.1) has to be adapted to depend on the joint step response h_s(t/T_b, b(k)) of the signal source 24 and the transmission channel 26.

As mentioned earlier, the step response h_s(t/T_b, b(k)) of the signal source 24 may be taken into account provided that the input signal is probed at the signal source 24 directly.

Generally, duty cycle distortion (DCD) occurs when the step response for a rising edge signal is different to the one for a falling edge signal.

The inter-symbol interference relates, for example, to limited transmission channel or rather reflection in the transmission.

The adapted model of the input signal due to the respective step response is given by

$\begin{array}{cc}{x}_{TN}\left(T/T_b\right)=\sum _{k=-N_{pre}}^{N_{\mathrm{post}}}\left[b\left(k\right)-b\left(k-1\right)\right]·h_s\left(t/T_b-k-\varepsilon \left(k\right)/T_b,b\left(k\right)\right)+x_{-\infty }+\sum _{i=0}^{N_{PN}\left(v\right)-1}A_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right)+x_{\mathrm{RN}\left(v\right)}\left(t/T_b\right)+x_{\mathrm{OBUN}\left(v\right)}\left(t/T_b\right).& \left(E\mathrm{.3}\right)\end{array}$

Therein, x_−∞ represents the state at the start of the transmission of the input signal, particularly the state of the signal source 24 and the transmission channel 26 at the start of the transmission of the input signal.

The step response h_s(t/T_b, b(k)) depends on the bit sequence b(k), or more precisely on a sequence of N_DCD bits of the bit sequence b(k), wherein N_DCD is an integer bigger than 1.

Note that there is an alternative formulation of the duty cycle distortion that employs N_DCD=1. This formulation, however, is a mere mathematical reformulation of the same problem and thus equivalent to the present disclosure.

Accordingly, the step response h_s(t/T_b, b(k)) may generally depend on a sequence of N_DCD bits of the bit sequence b(k), wherein N_DCD is an integer value.

Typically, the dependency of the step response h_s(t/T_b, b(k)) on the bit sequence b(k) ranges only over a few bits, for instance N_DCD=2, 3, . . . , 6.

For N_DCD=2 this is known as “double edge response (DER)”, while for N_DCD>2 this is known as “multi edge response (MER)”.

Without restriction of generality, the case N_DCD=2 is described in the following. However, the outlined steps also apply to the case N_DCD>2 with the appropriate changes. As indicated above, the following may also be (mathematically) reformulated for N_DCD=1.

In equation (E.3), the term b(k)−b(k−1), which is multiplied with the step response h_s(t/T_b, b(k)), takes two subsequent bit sequences, namely b(k) and b(k−1), into account such that a certain signal edge is encompassed.

In general, there may be two different values for the step response h_s(t/T_b, b(k)), namely h_s^(r)(t/T_b) for a rising signal edge and h_s^(f)(t/T_b) for a falling signal edge. In other words, the step response h_s(t/T_b, b(k)) may take the following two values:

$\begin{array}{cc}{h}_s\left(t/T_b,b\left(k\right)\right)=\{\begin{array}{cc}{h}_s^{\left(r\right)}\left(t/T_b\right),& b\left(k\right)-b\left(k-1\right)\ge 0\\ h_s^{\left(f\right)}\left(t/T_b\right),& b\left(k\right)-b\left(k-1\right)<0\end{array}.& \left(E\mathrm{.4}\right)\end{array}$

If the temporal jitter ε(k) is small, equation (E.3) can be linearized and then becomes

$\begin{array}{cc}{x}_{TN}\left(t/T_b\right)\approx \sum _{k=-N_{pre}}^{N_{\mathrm{post}}}\left[b\left(k\right)-b\left(k-1\right)\right]·h_s\left(t/T_b-k,b\left(k\right)\right)+x_{-\infty }+\sum _{i=0}^{N_{PN\left(\nu \right)}-1}A_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right)+x_{\mathrm{RN}\left(v\right)}\left(t/T_b\right)+x_{\mathrm{OBUN}\left(v\right)}\left(t/T_b\right)-\sum _{k=-N_{pre}}^{N_{\mathrm{post}}}\varepsilon \left(k\right)/T_b·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t/T_b-k,b\left(k\right)\right).& \left(E\mathrm{.5}\right)\end{array}$

Note that the last term in equation (E.5), namely

$\sum _{k=-N_{pre}}^{N_{\mathrm{post}}}\varepsilon \left(k\right)/T_b·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t/T_b-k,b\left(k\right)\right),$

describes an amplitude perturbation that is caused by the temporal jitter ε(k).

It is to be noted that the input signal comprises the total jitter as well as the total noise so that the input signal may also be labelled by x_TJ(t/T_b).

Clock Data Recovery

A clock data recovery is now performed based on the received input signal employing a clock timing model of the input signal, which clock timing model is a slightly modified version of the substitute model explained above. The clock timing model will be explained in more detail below.

Generally, the clock signal T_cik is determined while simultaneously determining the bit period T_b from the times t_edge(i) of signal edges based on the received input signal.

More precisely, the bit period T_b scaled by the sampling rate 1/T_a is inter alia determined by the analysis module 20.

In the following, {circumflex over (T)}_b/T_a is understood to be the bit period that is determined by the analysis module 20. The symbol “{circumflex over ( )}” marks quantities that are determined by the analysis module 20, for example quantities that are estimated by the analysis module 20.

One aim of the clock data recovery is to also determine a time interval error TIE (k) caused by the different types of perturbations explained above.

Moreover, the clock data recovery may also be used for decoding the input signal, for determining the step response h(t\T_b) and/or for reconstructing the input signal. Each of these applications will be explained in more detail below.

Note that for each of these applications, the same clock data recovery may be performed. Alternatively, a different type of clock data recovery may be performed for at least one of these applications.

In order to enhance the precision or rather accuracy of the clock data recovery, the bit period {circumflex over (T)}_b/T_a is determined jointly with at least one of the deterministic jitter components mentioned above and with a deviation Δ{circumflex over (T)}_b/T_a from the bit period {circumflex over (T)}_b/T_a.

In the case described in the following, the bit period {circumflex over (T)}_b/T_a and the deviation Δ{circumflex over (T)}_b/T_a are estimated together with the data dependent jitter component and the periodic jitter components. Therefore, the respective jitter components are taken into account when providing a cost functional that is to be minimized.

The principle of minimizing a cost functional, also called criterion, in order to determine the clock signal T_cik is known.

More precisely, the bit period {circumflex over (T)}_b/T_a and the deviation Δ{circumflex over (T)}_b/T_a are determined by determining the times t_edge (i) of signal edges based on the received input signal and by then minimizing the following cost functional K, for example by employing a least mean squares approach:

$\begin{array}{cc}K=\sum _{i=0}^{N-1}[\frac{t_{edge}\left(i\right)}{T_a}-k_{i,\eta }·\frac{{\hat{T}}_b\left(\eta \right)}{T_a}-\frac{\Delta {\hat{T}}_b\left(\eta \right)}{T_a}-\sum _{L_{{\mathrm{ISI}}_{pre}}}^{L_{{\mathrm{ISI}}_{\mathrm{post}}}}{\hat{h}}_{r,f}\left(k_i-\xi ,b\left(k_i\right),b\left(k_i-1\right),b\left(k_i-\xi \right),b\left(k_i-\xi -1\right)\right)-{\hspace{1em}\sum _{\mu =0}^{M_{PJ}-1}{\hat{C}}_{\mu }·\sin \left(2\pi ·{\hat{v}}_{\mu }/T_a·k_i+{\Psi }_{\mu }\right)]}^2.& \left(E\mathrm{.6}\right)\end{array}$

As mentioned above, the cost functional K used by the method according to the present disclosure comprises terms concerning the data dependent jitter component, which is represented by the fourth term in equation (E.6) and the periodic jitter components, which are represented by the fifth term in equation (E.6), namely the vertical periodic jitter components and/or the horizontal periodic jitter components.

Therein, L_ISI, namely the length L_ISIpre+L_ISIpost, is the length of an Inter-symbol Interference filter (ISI-filter) hĝ_r,f(k) that is known from the state of the art and that is used to model the data dependent jitter. The length L_ISI should be chosen to be equal or longer than the length of the impulse response, namely the one of the signal source 24 and the transmission channel 26.

Hence, the cost functional K takes several signal perturbations into account rather than assigning their influences to (random) distortions as typically done in the prior art.

In fact, the term

$\sum _{L_{{\mathrm{ISI}}_{pre}}}^{L_{{\mathrm{ISI}}_{\mathrm{post}}}}{\hat{h}}_{r,f}\left(k_i-\xi ,b\left(k_i\right),b\left(k_i-1\right),b\left(k_i-\xi \right),b\left(k_i-\xi -1\right)\right)$

relates to the data dependent jitter component. The term assigned to the data dependent jitter component has several arguments for improving the accuracy since neighbored edge signals, also called aggressors, are taken into account that influence the edge signal under investigation, also called victim.

In addition, the term

$\sum _{\mu =0}^{M_{PJ}-1}{\hat{C}}_{\mu }·\sin \left(2\pi ·{\hat{v}}_{\mu }/T_a·k_i+{\hat{\Psi }}_{\mu }\right)$

concerns the periodic jitter components, namely the vertical periodic jitter components and/or the horizontal periodic jitter components, that are also explicitly mentioned as described above. Put it another way, it is assumed that periodic perturbations occur in the received input signal which are taken into consideration appropriately.

If the signal source 24 is configured to perform spread spectrum clocking, then the bit period T_b/T_a is not constant but varies over time.

The bit period can then, as shown above, be written as a constant central bit period T_b, namely a central bit period being constant in time, plus a deviation ΔT_b from the central bit period T_b, wherein the deviation ΔT_b varies over time.

In this case, the period of observation is divided into several time slices or rather time sub-ranges. For ensuring the above concept, the several time slices are short such that the central bit period T_b is constant in time.

The central bit period T_b and the deviation ΔT_b are determined for every time slice or rather time sub-range by minimizing the following cost functional K:

$\begin{array}{cc}K=\sum _{i=0}^{N-1}{\left[\begin{array}{c}\frac{t_{edge}\left(i\right)}{T_a}-k_{i,\eta }·\frac{{\hat{T}}_b\left(\eta \right)}{T_a}-\frac{\Delta {\hat{T}}_b\left(\eta \right)}{T_a}-\\ \sum _{L_{{\mathrm{ISI}}_{\mathrm{pre}}}}^{L_{{\mathrm{ISI}}_{\mathrm{post}}}}{\hat{h}}_{r,f}\left(k_i-\xi ,b\left(k_i\right),b\left(k_i-1\right),b\left(k_i-\xi \right),b\left(k_i-\xi -1\right)\right)-\\ \sum _{\mu =0}^{M_{PJ}-1}{\hat{C}}_{\mu }·\sin \left(2\pi ·{\hat{v}}_{\mu }/T_a·k_i+{\hat{\Psi }}_{\mu }\right)\end{array}\right]}^2,& \left(E\mathrm{.7}\right)\end{array}$

which is the same cost functional as the one in equation (E.6).

Based on the determined bit period {circumflex over (T)}_b/T_a, and based on the determined deviation Δ{circumflex over (T)}_b/T_a, the time interval error TIE (i)/T_a is determined as

TIE(i)/T_a=t_edge(i)/T_a−k_i,η·{circumflex over (T)}_b(η)/T_a−Δ{circumflex over (T)}_b(η)/T_a

Put another way, the time interval error TIE(i)/T_a corresponds to the first three terms in equations (E.6) and (E.7), respectively.

However, one or more of the jitter components may also be incorporated into the definition of the time interval error TIE (i)/T_a.

In the equation above regarding the time interval error TIE(i)/T_a, the term k_i,η, {circumflex over (T)}_b(η)/T_a+Δ{circumflex over (T)}_b(η)/T_a represents the clock signal for the i-th signal edge. This relation can be rewritten as follows {circumflex over (T)}_clk=k_i,η·{circumflex over (T)}_b(η)/T_a+Δ{circumflex over (T)}_b(η)/T_a.

As already described, a least mean squares approach is applied with which at least the constant central bit period T_b and the deviation ΔT_b from the central bit period T_b are determined.

In other words, the time interval error TIE(i)/T_a is determined and the clock signal T_cik is recovered by the analysis described above.

In some embodiments, the total time interval error TIE_TJ(k) is determined employing the clock data recovery method described above (step S.3.1 in FIG. 3).

Generally, the precision or rather accuracy is improved since the occurring perturbations are considered when determining the bit period by determining the times t_edge (i) of signal edges based on the received input signal and by then minimizing the cost functional K.

Decoding the Input Signal

With the recovered clock signal T_cik determined by the clock recovery analysis described above, the input signal is divided into the individual symbol intervals and the values of the individual symbols (“bits”) b(k) are determined.

The signal edges are assigned to respective symbol intervals due to their times, namely the times t_edge(i) of signal edges. Usually, only one signal edge appears per symbol interval.

In other words, the input signal is decoded by the analysis module 20, thereby generating a decoded input signal. Thus, b(k) represents the decoded input signal.

The step of decoding the input signal may be skipped if the input signal comprises an already known bit sequence. For example, the input signal may be a standardized signal such as a test signal that is determined by a communication protocol. In this case, the input signal does not need to be decoded, as the bit sequence contained in the input signal is already known.

Joint Analysis of the Step Response and of the Periodic Signal Components

The analysis module 20 is configured to jointly determine the step response of the signal source 24 and the transmission channel 26 on one hand and the vertical periodic noise parameters defined in equation (E.5) on the other hand, wherein the vertical periodic noise parameters are the amplitudes A_i, the frequencies f_i and the phases ϕ_i (step S.3.2 in FIG. 3).

Therein and in the following, the term “determine” is understood to mean that the corresponding quantity may be computed and/or estimated with a predefined accuracy.

Thus, the term “jointly determined” also encompasses the meaning that the respective quantities are jointly estimated with a predefined accuracy.

However, the vertical periodic jitter parameters may also be jointly determined with the step response of the signal source 24 and the transmission channel 26 in a similar manner.

The concept is generally called joint analysis method.

In general, the precision or rather accuracy is improved due to jointly determining the step response and the periodic signal components.

Put differently, the first three terms in equation (E.5), namely

$\sum _{k=-N_{pre}}^{N_{\mathrm{post}}}\left[b\left(k\right)-b\left(k-1\right)\right]·h_s\left(t/T_b-k,b\left(k\right)\right)+x_{-\infty }+\sum _{i=0}^{N_{PN\left(\nu \right)}-1}A_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right),$

are jointly determined by the analysis module 20.

As a first step, the amplitudes A_i, the frequencies f_i and the phases ϕ_i are roughly estimated via the steps depicted in FIG. 4.

First, a clock data recovery is performed based on the received input signal (step S.4.1), particularly as described above.

Second, the input signal is decoded (step S.4.2).

Then, the step response, particularly the one of the signal source 24 and the transmission channel 26, is roughly estimated based on the decoded input signal (step S.4.3), for example by matching the first term in equation (E.5) to the measured input signal, for example via a least mean squares approach.

Therein and in the following, the term “roughly estimated” is to be understood to mean that the corresponding quantity is estimated with an accuracy being lower compared to the case if the quantity is determined.

Now, a data dependent jitter signal x_DDJ being a component of the input signal only comprising data dependent jitter is reconstructed based on the roughly estimated step response (step S.4.4).

The data dependent jitter signal x_DDJ is subtracted from the input signal (step S.4.5). The result of the subtraction is the signal x_PN-FRN that approximately only contains periodic noise and random noise.

Finally, the periodic noise parameters A_i, f_i, ϕ_f are roughly estimated based on the signal x_PN-FRN (step S.4.6), for example via a fast Fourier transform of the signal x_PN+RN.

In the following, these roughly estimated parameters are marked by subscripts “0”, i.e. the rough estimates of the frequencies are f_i,0 and the rough estimates of the phases are ϕ_i,0. The roughly estimated frequencies f_i,0 and phases ϕ_i,0 correspond to working points for linearizing purposes as shown hereinafter.

Accordingly, the frequencies and phases can be rewritten as follows:

f _i /f _b =f _i,0 /f _b +Δf _i /f _b

ϕ_i=ϕ_i,0+Δϕ_i   (E.8)

Therein, Δf_i and Δϕ_i are deviations of the roughly estimated frequencies f_i,0 and phases ϕ_i,0 from the actual frequencies and phases, respectively. By construction, the deviations Δf_i and Δϕ_i are much smaller than the associated frequencies f_i and phases ϕ_i, respectively.

With the re-parameterization above, the sine-function in the third term in equation (E.5), namely

$\sum _{i=0}^{N_{PN\left(\nu \right)}-1}A_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right),$

can be linearized as follows while using small-angle approximation or rather the Taylor series:

$\begin{array}{cc}{A}_i·\sin \left(2\pi ·f_i/f_b·t/T_b+{\phi }_i\right)=A_i·\sin \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}+2\pi ·\Delta f_i/f_b·t/T_b+\Delta {\phi }_i\right)=A_i·\left[\sin \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right)·\cos \left(2\pi ·\Delta f_i/f_b·t/T_b+{\mathrm{\Delta \phi }}_i\right)+\cos \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right)·\sin \left(2\pi ·\Delta f_i/f_b·t/T_b+\Delta {\phi }_i\right)\right]\approx A_i·\sin \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right)+A_i·\cos \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right)·\left[2\pi ·\Delta f_i/f_b·t/T_b+{\mathrm{\Delta \phi }}_i\right]=p_{i,0}·\sin \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right)+p_{i,1}·2\pi ·t/T_b·\cos \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right)+p_{i,2}·\cos \left(2\pi ·f_{i,0}/f_b·t/T_b+{\phi }_{i,0}\right).& \left(E\mathrm{.9}\right)\end{array}$

In the last two lines of equation (E.9), the following new, linearly independent parameters have been introduced, which are determined afterwards:

p _i,0 =A _i

p _i,1 =A _i ·Δf _i /f _b

p _i,2 =A _i·Δϕ_i   (E.10)

With the mathematical substitute model of equation (E.5) adapted that way, the analysis module 20 can now determine the step response h_s(t/T_b,b(k)), more precisely the step response h_s^(r)(t/T_b) for rising signal edges and the step response h_s^(f)(t/T_b) for falling signal edges, and the vertical periodic noise parameters, namely the amplitudes A_i, the frequencies f_i and the phases ϕ_i, jointly, i.e. at the same time.

This may be achieved by minimizing a corresponding cost functional K, for example by applying a least mean squares method to the cost functional K. The cost functional has the following general form:

K=[A(k)·{circumflex over (x)}−x_J(k)]^T·[A(k)·{circumflex over (x)}−x_J(k)]   (E.11)

Therein, x_L (k) is a vector containing L measurement points of the measured input signal. ŝ is a corresponding vector of the input signal that is modelled as in the first three terms of equation (E.5) and that is to be determined. A(k) is a matrix depending on the parameters that are to be determined.

In some embodiments, matrix A(k) comprises weighting factors for the parameters to be determined that are assigned to the vector x_J(k).

Accordingly, the vector x_L(k) may be assigned to the step response h_s^(r)(t/T_b) for rising signal edges, the step response h_s^(f)(t/T_b) for falling signal edges as well as the vertical periodic noise parameters, namely the amplitudes A_i, the frequencies f_i and the phases ϕ_i.

The least squares approach explained above can be extended to a so-called maximum-likelihood approach. In this case, the maximum-likelihood estimator {circumflex over (x)}_ML is given by

{circumflex over (x)} _ML=[A^T(k)·R_n⁻¹(k)·A(k)]⁻¹·[A^T(k)R_n⁻¹(k)·x_L(k)]   (E.11a)

Therein, R_n(k) is the covariance matrix of the perturbations, i.e. the jitter and noise components comprised in equation (E.5).

Note that for the case of pure additive white Gaussian noise, the maximum-likelihood approach is equivalent to the least squares approach.

The maximum-likelihood approach may be simplified by assuming that the perturbations are not correlated with each other. In this case, the maximum-likelihood estimator becomes

{circumflex over (x)} ≈[A^T(k)·((r_n,i(k)·1^T)ºA(k))]⁻¹·[A^T(k)·(r_n,i(k)ºx_L(k))]   (E.11b)

Therein, 1^T is a unit vector and the vector r_n,i(k) comprises the inverse variances of the perturbations.

For the case that only vertical random noise and horizontal random noise are considered as perturbations, this becomes

$\begin{array}{cc}{\left[{\underset{\_}{r}}_{n,i}\left(k\right)\right]}_l={\left(\begin{array}{c}\frac{{\sigma }_{ϵ,\mathrm{RJ}}^2}{T_b^2}\sum _{m=-N_{\mathrm{post}}}^{N_{p\tau e}}{\left[b\left(k-l-m\right)-b\left(k-l-m-1\right)\right]}^2·\\ {\left(h\left(m,b\left(m\right)\right)\right)}^2+{\sigma }_{RN\left(v\right)}^2\end{array}\right)}^{-1}& \left(E\mathrm{.11}c\right)\end{array}$

Employing equation (E.11c) in equation (E.11b), an approximate maximum likelihood estimator is obtained for the case of vertical random noise and horizontal random noise being approximately Gaussian distributed.

If the input signal is established as a clock signal, i.e. if the value of the individual symbol periodically alternates between two values with one certain period, the approaches described above need to be adapted. The reason for this is that the steps responses usually extend over several bits and therefore cannot be fully observed in the case of a clock signal. In this case, the quantities above have to be adapted in the following way:

{circumflex over (x)}=[({circumflex over (h)}_s^(r)))^T({circumflex over (h)}_s^(f))^T{circumflex over (p)}_3NPj^T]^T

A (k)=[b_L,N^(r)(k)−b_L,N^(r)(k−T_b/T_a)b_L,N^(f)(k)−b_L,N^(f)(k−T_b/T_a)t_L,3NPJ(k)]

x _L(k)=[b_L,N^(r)(k)−b_L,N^(r)(k−T_b/T_a)]·h_s^(r)+[b_L,N^(f)(k)−b_L,N^(f)(k−T_b/T_a)]·h_s^(f)   (E.11d)

Input Signal Reconstruction and Determination of Time Interval Error

With the determined step response and with the determined periodic noise signal parameters, a reconstructed signal {circumflex over (x)}_DDJ+PN(v)(t/T_b) containing only data dependent jitter and vertical periodic noise can be determined while taking equation (E.5) into account.

Thus, the reconstructed signal {circumflex over (x)}_DDJ+PN(v)(t/T_b) is given by

$\begin{array}{cc}\begin{array}{c}{\hat{x}}_{\mathrm{DDJ}+\mathrm{PN}\left(v\right)}\left(t/T_b\right)={\hat{x}}_{\mathrm{DDJ}}\left(t/T_b\right)+{\hat{x}}_{\mathrm{PN}\left(v\right)}\left(t/T_b\right)\\ =\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}\left[\hat{b}\left(k\right)-\hat{b}\left(k-1\right)\right]·{\hat{h}}_s\left(t/T_b-k,\hat{b}\left(k\right)\right)+\\ {\hat{x}}_{-\infty }+\sum _{i=0}^{N_{\mathrm{PN}\left(v\right)}-1}{\hat{A}}_i·\sin \left(2\pi ·{\hat{f}}_i/f_b·t/T_b+{\hat{\phi }}_i\right)\end{array}.& \left(E\mathrm{.12}\right)\end{array}$

Moreover, also a reconstructed signal {circumflex over (x)}_DDJ(t/T_b) containing only data dependent jitter and a reconstructed signal {circumflex over (x)}_PN(v)(t/T_b) containing only vertical periodic noise are determined by the analysis module 20 (steps S.3.3 and S.3.4).

Based on the reconstructed signals {circumflex over (x)}_DDJ(t/T_b) and {circumflex over (x)}_DDJ+PN(v)(t/T_b), the time interval error TIE_DDJ(k) that is associated with data dependent jitter and the time interval error TIE_DDJ+PJ(v)(k) that is associated with data dependent jitter and with vertical periodic jitter are determined (steps 5.3.3.1 and S.3.4.1).

Histograms

The analysis module 20 is configured to determine histograms of at least one component of the time interval error based on the corresponding time interval error (step S.3.5).

Generally speaking, the analysis unit 20 is firstly configured to determine the time interval error TIE_Jx associated with a jitter component Jx. The analysis module 20 can determine a histogram associated with that jitter component Jx and may display it on the display 22.

FIGS. 5A-5D show four examples of histograms determined by the analysis module 20 that correspond to total jitter, data dependent jitter, periodic jitter and random jitter, respectively.

In the cases of total jitter and data dependent jitter, rising signal edges and falling signal edges are treated separately such that information on duty cycle distortion is contained within the histogram.

Of course, a histogram corresponding only to certain components of the periodic jitter and/or of the random jitter may be determined and displayed, for example a histogram corresponding to at least one of horizontal periodic jitter, vertical periodic jitter, horizontal random jitter and vertical random jitter.

Note that from FIG. 5D it can readily be seen that the time interval error associated with the random jitter is Gaussian-distributed.

Moreover, the deterministic jitter and the random jitter are statistically independent from each other. Thus, the histogram of the total jitter may be determined by convolution of the histograms related to deterministic jitter and random jitter.

The measurement instrument 12 may be configured to selectively display one or more of the determined histograms on the display 22.

In some embodiments, the user may choose which of the jitter components are selectively displayed.

Thus, the histogram corresponding to the time interval error associated with at least one of the vertical periodic jitter, the horizontal periodic jitter, the vertical random jitter, the horizontal random jitter, the data dependent jitter and the other bounded uncorrelated jitter may be selectively displayed on the display 22.

Separation of Random Jitter and Horizontal Periodic Jitter

The analysis module 20 is configured to determine the time interval error TIE_RJ that is associated with the random jitter and the time interval error TIE_PJ(h) that is associated with the horizontal periodic jitter (step S.3.6).

As shown in FIG. 3, the total time interval error TIE_TJ(k) and the time interval error TIE_DDJ+PJ(V) that is associated with data dependent jitter and with vertical periodic jitter are determined firstly as already described above.

Then, TIE_DDJ+PJ(V) is subtracted from the total time interval error TIE_TJ(k) such that the time interval error TIE_RJ+PJ(h) is obtained that only contains random jitter, horizontal periodic jitter and other bounded uncorrelated jitter. In this regard, reference is made to FIG. 2 illustrating an overview of the several jitter components.

Note that in the following, the other bounded uncorrelated jitter component is neglected. However, it may also be incorporated into the analysis described below.

Analogously to the joint analysis method described above (step S.3.2), also the horizontal periodic jitter defined by the first term in equation (E.2), particularly its time interval error, can be determined by determining the corresponding amplitudes α_i, frequencies ϑ_t and phases φ_t. A flow chart of the corresponding method is depicted in FIG. 6.

For this purpose, the amplitudes α_t, frequencies ϑ_t and phases φ_t are roughly estimated at first (step S.6.1).

Then, at least these parameters are determined jointly (step S.6.2).

The time interval error TIE_PJ(h) that is associated with horizontal periodic jitter is then reconstructed (step S.6.3). The result is given by

$\begin{array}{cc}T\hat{I}E_{PJ\left(h\right)}\left(k\right)=\sum _{i=0}^{{\hat{N}}_{PJ\left(h\right)}-1}{\hat{a}}_i·\sin \left(2\pi ·{\hat{\vartheta }}_i/f_b·k+{\hat{\varphi }}_i\right).& \left(E\mathrm{.13}\right)\end{array}$

From this, also the time interval error TIE_RJ being associated only with random jitter is calculated by subtracting TÎE_PJ(h) from TIE_RJ+PJ(h).

Determination of Random Jitter

Generally, the analysis module 20 is configured to determine a statistical moment that is associated with the temporal random jitter ε_RJ. Therein, the statistical moment is of second order or higher.

In some embodiments the analysis module 20 is configured to determine the variance σ_εRJ² that is associated with the temporal random jitter ε_RJ (step S.3.7). This step is explained in more detail below.

The approach is based on determining an autocorrelation function r_TIE,TIE(m) of the time interval error that is defined by

$r_{\mathrm{TIE},\mathrm{TIE}}\left(m\right)=\frac{1}{N_{ACF}\left(m\right)}\sum _{k=0}^{N_{ACF}\left(m\right)-1}\mathrm{TIE}\left(k\right)·\mathrm{TIE}\left(k+m\right),m=0,1,\dots ,L_{ACF}-1$

wherein N_ACF (m) is the number of elements that are taken into account for the calculation of the autocorrelation function. As shown, the number of elements depends on displacement parameter m.

Further, L_ACF corresponds to the length of the autocorrelation function. The length may be adjustable by the user and/or may be equal to or bigger than the maximum of the maximal period of the periodic jitter and the length of the impulse response of the signal source 24 and the transmission channel 26.

In general, the analysis module 20 may be configured to selectively determine the respective autocorrelation function r_TIEJxTIEJx(m) of any jitter component Jx.

Generally, the measurement instrument 12 may be configured to selectively display the autocorrelation function r_TIEJx,TIEJx(m) obtained on the display 22.

Accordingly, the approach for determining the variance

${\sigma }_{{\varepsilon }_{\mathrm{RJ}}}^2,$

namely me variance of the temporal random jitter ε_RJ, is based on determining the autocorrelation function r_TIERJ,TIERJ(m) of the random jitter.

A component x_DDJ+RJ(t/T_b)≈x_DDJ(t/T_b)+n_RJ(t/T_b) of the input signal contains the data dependent jitter signal

$\begin{array}{cc}{x}_{DDJ}\left(t/T_b\right)=\sum _{k=-N_{pre}}^{N_{\mathrm{post}}}\left[b\left(k\right)-b\left(k-1\right)\right]·h_s\left(t/T_b-k,b\left(k\right)\right)+x_{-\infty }& \left(E\mathrm{.14}\right)\end{array}$

and the perturbation

$\begin{array}{cc}{n}_{RJ}\left(t/T_b\right)=-\sum _{k=-N_{pre}}^{N_{\mathrm{post}}}{\varepsilon }_{RJ}\left(k\right)/T_b·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t/T_b-k,b\left(k\right)\right)& \left(E\mathrm{.15}\right)\end{array}$

that is caused by the random jitter ε_RJ(k)/T_b. As shown above, the periodic jitter is not taken into account in the following. However, it might be taken into account if desired.

As can be seen from FIG. 7A, the time interval error TIE_RJ that is associated with the random jitter ε_RJ(k)/T_b is given by

$\begin{array}{cc}\frac{{\mathrm{TIE}}_{RJ}\left(t_{\mathrm{edge}}/T_b\right)}{T_b}\approx -{\left[\frac{dx_{DDJ}\left(t_{\mathrm{edge}}/T_b\right)}{d\left(t/T_b\right)}\right]}^{-1}·n_{RJ}\left(t_{\mathrm{edge}}/T_b\right).& \left(E\mathrm{.16}\right)\end{array}$

In this approach the times t_edge of the signal edges of the data dependent jitter signal x_DDJ(t/T_b) are determined by the analysis module 20, for example based on the reconstructed data dependent jitter signal {circumflex over (x)}_DDJ(t/T_b).

Alternatively, as depicted in FIG. 7B the clock times t_CLK can be used that are known from the clock data recovery explained above (step S.4.1). In this case, the time interval error TIE_RJ is given by

$\begin{array}{cc}\frac{{\mathrm{TIE}}_{RJ}\left(t_{\mathrm{CLK}}/T_b\right)}{T_b}\approx -{\left[\frac{dx_{DDJ}\left(t_{CLK}/T_b\right)}{d\left(t/T_b\right)}\right]}^{-1}·n_{RJ}\left(t_{CLK}/T_b\right)& \left(E\mathrm{.17}\right)\end{array}$

In some embodiments the clock times t_CLK can be used provided that the slopes of the respective jitter signals, namely the data dependent jitter signal x_DDJ(t/T_b) as well as the component x_DDJ+RJ(t/T_b) of the input signal, are substantially equal as indicated in FIG. 7B.

The respective equations can be easily determined from the respective gradient triangle in FIGS. 7A and 7B.

In the following, the relation of equation (E.16) is used to derive the variance σ_εRJ². However, it is to be understood that the relation of equation (E.17) could be used instead.

Using equation (E.16), the autocorrelation function of the random jitter is given by

$\begin{array}{cc}{r}_{{\mathrm{TIE}}_{RJ},{\mathrm{TIE}}_{RJ}}\left(m\right)=E\left\{TIE_{RJ}\left(t_{edge}/T_b\right)/T_b·{\mathrm{TIE}}_{RJ}\left(t_{edge}/T_b+m\right)/T_b\right\}\approx E\left\{\begin{array}{c}{\left[\frac{dx_{DDJ}\left(t_{\mathrm{CLK}}/T_b\right)}{d\left(\frac{t}{T_b}\right)}\right]}^{-1}·{\left[\frac{dx_{DDJ}\left(t_{CLK}/T_b+m\right)}{d\left(\frac{t}{T_b}\right)}\right]}^{-1}·\\ n_{RJ}\left(t_{CLK}/T_b\right)·n_{RJ}\left(t_{CLK}/T_b+m\right)\end{array}\right\}.& \left(E\mathrm{.18}\right)\end{array}$

Therein and in the following, E{y} indicates an expected value of quantity y.

The method for determining the autocorrelation function r_TIERJ,TIERJ is illustrated in FIG. 8.

The upper two rows in FIG. 8 represent a memory range of the transmission channel 26 with a bit change at time k=0. Accordingly, the lower two rows represent a memory range of the transmission channel 26 with a bit change at time k=m. Note that the example in FIG. 8 is for a PAM-2 coded input signal. However, the steps outlined in the following can readily be applied to a PAM-n coded input signal with appropriate combinatorial changes.

The memory range of the transmission channel 26 is N_pre+N_post+1. Thus, there are 2^Npre+Npost possible permutations {b(k)} of the bit sequence b(k) within the memory range.

The upper rows and the lower rows overlap in an overlap region starting at k=N_start and ending at k=N_end. In the overlap region, the permutations of the bit sequences b(k) in the memory ranges of the upper and the lower rows have to be identical.

Note that only the overlap region contributes to the autocorrelation function.

In order to calculate the number of possible permutations in the overlap region, a case differentiation is made as follows:

The bit change at k=0 may be completely within the overlap region, completely outside of the overlap region or may overlap with the edge of the overlap region (i.e. one bit is inside of the overlap region and one bit is outside of the overlap region).

Similarly, the bit change at k=m may be completely within the overlap region, completely outside of the overlap region or may overlap with the edge of the overlap region (i.e. one bit is inside of the overlap region and one bit is outside of the overlap region).

Thus, there is a total of 3·3=9 cases that are taken into account.

Each permutation {b(k)} has a chance of P(u, v) for occurring and implies a particular slope

${\mathrm{dx}}_{\mathrm{DDJ}}\left(t_{\mathrm{edge}}\text{/}T_b\right)\text{/}d\left(\frac{t}{T_b}\right){|}_{\left(u,v\right)}$

the data dependent jitter signal x_DDJ(t_edge/T_b). Therein, u and v represent a particular realization of the bit sequence inside the overlap region and a particular realization of the bit sequence outside of the overlap region, respectively.

Now, the autocorrelation function of the perturbation n_RJ(t_edge/T_b k) for two particular realizations of the memory ranges at times k=0 and k=m leading to the particular realization u in the overlap region are determined. The conditional autocorrelation function of the perturbation n_RJ(t_edge/T_b) is determined to be

$\begin{array}{cc}E\left\{n_{\mathrm{RJ}}\left(t_{\mathrm{edge}}\text{/}T_b\right)·n_{\mathrm{RJ}}\left(t_{\mathrm{edge}}\text{/}T_b+m\right)\right\}{|}_u=\sum _{k_0=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}\sum _{k_1=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}\left[b\left(k_0\right)-b\left(k_0-1\right)\right]·\left[b\left(k_1+m\right)-b\left(k_1-1+m\right)\right]·h\left(t_{\mathrm{edge}}\text{/}T_b-k_0,b\left(k_0\right)\right)·h\left(t_{\mathrm{edge}}\text{/}T_b+m-k_1,b\left(k_1+m\right)\right)·E\left\{{\varepsilon }_{\mathrm{RJ}}\left(k_0\right)\text{/}T_b·{\varepsilon }_{\mathrm{RJ}}\left(k_1+m\right)\text{/}T_b\right\}& \left(E\mathrm{.19}\right)\end{array}$

The temporal random jitter ε_RJ(k)/T_b is normally distributed, particularly stationary and normally distributed. Thus, the autocorrelation function for the temporal random jitter ε_RJ(k)/T_b can be isolated since the other terms relate to deterministic contributions. In fact, the autocorrelation function for the temporal random jitter ε_RJ(k)/T_b is

$\begin{array}{cc}E\left\{{\varepsilon }_{\mathrm{RJ}}\left(k_0\right)\text{/}T_b·{\varepsilon }_{\mathrm{RJ}}\left(k_1+m\right)\text{/}T_b\right\}=\{\begin{array}{cc}{\sigma }_{{ϵ}_{\mathrm{RJ}}}^2\text{/}T_b^2& k_0=k_1+m\\ 0& \mathrm{else}\end{array}.& \left(E\mathrm{.20}\right)\end{array}$

Hence, the autocorrelation function for the temporal random jitter ε_RJ(k)/T_b has only one contribution different from zero, namely for k₀=k₁+m. Accordingly, equation (E.191 becomes

$\begin{array}{cc}E\left\{n_{\mathrm{RJ}}\left(t_{\mathrm{edge}}\text{/}T_b\right)·n_{\mathrm{RJ}}\left(t_{\mathrm{edge}}\text{/}T_b+m\right)\right\}{|}_u=\frac{{\sigma }_{{ϵ}_{\mathrm{RJ}}}^2}{T_b^2}·\sum _{k=N_{\mathrm{start}}}^{N_{\mathrm{end}}}{\left[b\left(k\right)-b\left(k-1\right)\right]}^2·h\left(t_{\mathrm{edge}}\text{/}T_b-k,b\left(k\right)\right)·h\left(t_{\mathrm{edge}}\text{/}T_b+m-k,b\left(k\right)\right).& \left(E\mathrm{.21}\right)\end{array}$

As already mentioned, only the overlap region has a contribution. Employing equation (E.21), the autocorrelation function of the random jitter is determined to be

$\begin{array}{cc}{r}_{{\mathrm{TIE}}_{\mathrm{RJ}},{\mathrm{TIE}}_{\mathrm{RJ}}}\left(m\right)\approx \sum _uE\left\{n_{\mathrm{RJ}}\left(t_{\mathrm{edge}}\text{/}T_b\right)·n_{\mathrm{RJ}}\left(t_{\mathrm{edge}}\text{/}T_b+m\right)\right\}{|}_u·\sum _v\sum _wP\left(\left(u,v\right)\bigcap \left(u,w\right)\right)·{\left[\frac{{\mathrm{dx}}_{\mathrm{DDJ}}\left(t_{\mathrm{edge}}\text{/}T_b\right)}{d\left(\frac{t}{T_b}\right)}{|}_{\left(u,v\right)}\right]}^{-1}·{\left[\frac{{\mathrm{dx}}_{\mathrm{DDJ}}\left(t_{\mathrm{edge}}\text{/}T_b+m\right)}{d\left(\frac{t}{T_b}\right)}{|}_{\left(u,w\right)}\right]}^{-1},& \left(E\mathrm{.22}\right)\end{array}$

wherein P((u, v)∩(u, w)) is the joint probability density defined by

$\begin{array}{cc}P\left(\left(u,v\right)\bigcap \left(u,w\right)\right)=P\left(u,v\right)·P\left(\left(u,w\right)|\left(u,v\right)\right)=P\left(u,v\right)·\frac{P\left(u,w\right)}{{\sum }_wP\left(u,w\right)·{\sum }_uP\left(u\right)}.& \left(E\mathrm{.23}\right)\end{array}$

As can clearly be seen from equations (E.21) and (E.22), the autocorrelation function r_TIERJTIERJ (m) of the random jitter is linearly dependent on the variance σ_ϵRJ² of the random jitter.

Thus, the variance σ_ϵRF² of the random jitter is determined as follows.

On one hand, the impulse response h(t_edge/T_b−k, b(k)) is already known or can be determined, as it is the time derivative of the determined step response h_s(t/T_b−k,b(k)) evaluated at time t=t_edge. Moreover, the bit sequence b(k) is also known via the signal decoding procedure described above.

On the other hand, the time interval error TIE_RJ(k) is known from the separation of the random jitter and the horizontal periodic jitter described above (step S.3.6) and the autocorrelation function can be also calculated from this directly.

Thus, the only unknown quantity in equations (E.21) and (E.22) is the variance σ_ϵRJ² of the random jitter, which can thus be determined from these equations.

As shown in FIG. 3 as well as FIG. 9, the autocorrelation function can be determined for any jitter component.

In FIG. 9, the autocorrelation functions for the total jitter signal, the periodic jitter signal, the data dependent jitter signal as well as the random jitter are shown.

Generally, the respective result may be displayed on the display 22.

Power Spectral Density

The power spectral density R_TIE,TIE(f/f_b) of the time interval error is calculated based on the autocorrelation function by a Fourier series, which reads

$\begin{array}{cc}{R}_{\mathrm{TIE},\mathrm{TIE}}\left(f\text{/}f_b\right)=\sum _{m=-L_{\mathrm{ACF}}+1}^{+L_{\mathrm{ACF}}-1}r_{\mathrm{TIE},\mathrm{TIE}}\left(m\right)·e^{-j·2\pi ·f\text{/}f_b·m}.& \left(E\mathrm{.24}\right)\end{array}$

The analysis module 20 may be configured to selectively determine the power spectral density R_TIEJx,TIEJx (m) of any jitter component Jx.

Moreover, the measurement instrument 12 may be configured to selectively display the power spectral density R_TIEJxTIEJx (n) on the display 22.

As shown in FIG. 3 as well as FIG. 10, the power spectral density can be determined for any jitter component.

In FIG. 10, the power spectral densities for the total jitter signal, the periodic jitter signal, the data dependent jitter signal as well as the random jitter are shown.

Generally, the respective result may be displayed on the display 22.

Bit Error Rate

The analysis module 20 is configured to determine the bit error rate BER(t/T_b) that is caused by the time interval error TIE_Dj+Rj being associated with the deterministic jitter and the random jitter, i.e. with the total jitter (step S.3.8).

A bit error occurs if the time interval error TIE_DJ being associated with the deterministic jitter and the time interval error TIE_RJ being associated with the random jitter fulfill one of the following two conditions:

$\begin{array}{cc}\frac{t}{T_b}<{\mathrm{TIE}}_{\mathrm{DJ}}+{\mathrm{TIE}}_{\mathrm{RJ}},0\le \frac{t}{T_b}\le \frac{1}{2}\frac{t}{T_b}>1+{\mathrm{TIE}}_{\mathrm{DJ}}+{\mathrm{TIE}}_{\mathrm{RJ}},\frac{1}{2}<\frac{t}{T_b}<1& \left(E\mathrm{.25}\right)\end{array}$

Thus, based on the histogram of the time interval error TIE_DJ associated with deterministic jitter and based on the variance σ_RJ² of the time interval error TIE_RJ, the bit error rate BER(t/T_b) is determined as follows for times t/T_b<½:

$\begin{array}{cc}\mathrm{BER}\left(\frac{t}{T_b}\right)=P_{\mathrm{rise}}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{rise}}-1}P_{\mathrm{DJx},\mathrm{rise}}\left(i\right)·\stackrel{\infty }{\underset{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{rise}}\left(i\right)}{\int }}\frac{1}{\sqrt{2\pi }·{\sigma }_{\mathrm{RJ}}}·e^{\frac{-{\mathrm{RJ}}^2}{2{\sigma }_{\mathrm{RJ}}^2}}·\mathrm{dRJ}+P_{\mathrm{fall}}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{fall}}-1}P_{\mathrm{DJx},\mathrm{fall}}\left(i\right)·\stackrel{\infty }{\underset{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{fall}}\left(i\right)}{\int }}\frac{1}{\sqrt{2\pi }·{\sigma }_{\mathrm{RJ}}}·e^{\frac{-{\mathrm{RJ}}^2}{2{\sigma }_{\mathrm{RJ}}^2}}·\mathrm{dRJ}=\frac{P_{\mathrm{rise}}}{2}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{rise}}-1}P_{\mathrm{DJx},\mathrm{rise}}\left(i\right)·\mathrm{erfc}\left(\frac{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{rise}}\left(i\right)}{\sqrt{2}·{\sigma }_{\mathrm{RJ}}}\right)+\frac{P_{\mathrm{fall}}}{2}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{fall}}-1}P_{\mathrm{DJx},\mathrm{fall}}\left(i\right)·\mathrm{erfc}\left(\frac{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{fall}}\left(i\right)}{\sqrt{2}·{\sigma }_{\mathrm{RJ}}}\right).& \left(E\mathrm{.26}\right)\end{array}$

For times ½<t/T_b<1, the bit error rate BER(t/T_b) is determined to be

$\begin{array}{cc}\mathrm{BER}\left(\frac{t}{T_b}\right)=P_{\mathrm{rise}}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{rise}}-1}P_{\mathrm{DJx},\mathrm{rise}}\left(i\right)·\stackrel{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{rise}}\left(i\right)-1}{\underset{-\infty }{\int }}\frac{1}{\sqrt{2\pi }·{\sigma }_{\mathrm{RJ}}}·e^{\frac{-{\mathrm{RJ}}^2}{2{\sigma }_{\mathrm{RJ}}^2}}·\mathrm{dRJ}+P_{\mathrm{fall}}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{fall}}-1}P_{\mathrm{DJx},\mathrm{fall}}\left(i\right)·\stackrel{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{fall}}\left(i\right)-1}{\underset{-\infty }{\int }}\frac{1}{\sqrt{2\pi }·{\sigma }_{\mathrm{RJ}}}·e^{\frac{-{\mathrm{RJ}}^2}{2{\sigma }_{\mathrm{RJ}}^2}}·\mathrm{dRJ}=P_{\mathrm{rise}}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{rise}}-1}P_{\mathrm{DJx},\mathrm{rise}}\left(i\right)·\left[1-\frac{1}{2}\mathrm{erfc}\left(\frac{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{rise}}\left(i\right)-1}{\sqrt{2}·{\sigma }_{\mathrm{RJ}}}\right)\right]+P_{\mathrm{fall}}·\sum _{i=0}^{N_{\mathrm{DJ},\mathrm{fall}}-1}P_{\mathrm{DJx},\mathrm{fall}}\left(i\right)·\left[1-\frac{1}{2}\mathrm{erfc}\left(\frac{\frac{t}{T_b}-{\mathrm{TIE}}_{\mathrm{DJ},\mathrm{fall}}\left(i\right)-1}{\sqrt{2}·{\sigma }_{\mathrm{RJ}}}\right)\right].& \left(E\mathrm{.27}\right)\end{array}$

Therein, P_rise and P_fall are the probabilities of a rising signal edge and of a falling signal edge, respectively. N_DJ,rise and N_DJ,fall are the numbers of histogram containers of the deterministic jitter for rising signal edges and for falling signal edges, respectively. Correspondingly, TIE_DJ,rise (i) and TIE_DJ,fall(i) are the histogram values of the deterministic jitter for rising signal edges and for falling signal edges, respectively.

Thus, the bit error rate BER(t/T_b) is calculated based on the histogram of the deterministic jitter and based on the variance of the random jitter rather than determined directly by measuring the number of bit errors occurring within a certain number of bits.

Generally spoken, the bit error rate BER(t/T_b) is determined based on the respective time interval error used for deriving at the corresponding histogram.

This way, the bit error rate can also be determined in regions that are not accessible via direct measurements or that simply take a too long time to measure, for example for bit error rates BER(t/T_b)<10⁻⁶.

In some embodiments, bit error rates smaller than 10⁻⁸, smaller than 10⁻¹⁰ or even smaller than 10⁻¹² can be determined employing the method described above.

In order to linearize the curves describing the bit error rate, a mathematical scale transformation Q(t/T_b) may be applied to the bit error rate, which is, at least for the case of P_rise+P_fall=0.5 given by:

Q(t/T_b)=√{square root over (2)}·erf⁻¹(1−2·BER(t/T_b))   (E.28)

Instead of employing the histogram of the complete deterministic jitter, a histogram corresponding to at least one of the components of the deterministic jitter may be employed. Put differently, one or more of the components of the deterministic jitter may be selectively suppressed and the corresponding change of the bit error rate may be determined. This is also shown in FIG. 3.

More precisely, one of or an arbitrary sum of the data dependent jitter, the other bounded uncorrelated jitter, the horizontal periodic jitter and the vertical periodic jitter may be included and the remaining components of the deterministic jitter may be suppressed.

For instance, the bit error rate BER(t/T_b) is determined based on the histogram related to data dependent jitter, the histogram related to data dependent jitter and periodic jitter or the histogram related to data dependent jitter and other bounded uncorrelated jitter.

Moreover, the horizontal and vertical components may be selectively taken into account. In fact, the precision or rather accuracy may be improved.

Analogously, only the variance of the vertical random jitter or of the horizontal random jitter may be employed instead of the variance of the complete random jitter such that the other one of the two random jitter components is suppressed and the effect of this suppression may be determined.

The respective histograms may be combined in any manner. Hence, the periodic jitter may be obtained by subtracting the data dependent jitter from the deterministic jitter.

Depending on which of the deterministic jitter components is included, the final result for the bit error rate BER(t/T_b) includes only the contributions of these deterministic jitter components.

Thus, the bit error rate BER(t/T_b) corresponding to certain jitter components can selectively be determined.

The determined bit error rate BER(t/T_b) may be displayed on the display 22 as shown in FIG. 11.

In FIG. 11, a measured bit error rate as well as a bit error rate estimated with methods known in the prior art are also shown.

In FIG. 12, the respective mathematical scale transformation is shown that may also be displayed on the display 22.

In some embodiments, a bit error rate BER(t/T_b) containing only certain deterministic jitter components may be displayed on the display 22, wherein a user may choose which of the deterministic jitter components are included. Moreover, the fraction of the complete bit error rate that is due to the individual jitter components may be determined and displayed on the display 22.

Note that if the individual histograms of two statistically independent components TIE₀ and TIE₁ of the time interval error TIE are known, the resulting collective histogram containing both components can be determined by a convolution of the two individual histograms:

$\begin{array}{cc}{f}_{{\mathrm{TIE}}_0+{\mathrm{TIE}}_1}\left({\mathrm{TIE}}_0+{\mathrm{TIE}}_1\right)=\sum _{\xi =-\infty }^{+\infty }f_{{\mathrm{TIE}}_0}\left(\xi \right)·f_{{\mathrm{TIE}}_1}\left({\mathrm{TIE}}_0+{\mathrm{TIE}}_1-\xi \right).& \left(E\mathrm{.29}\right)\end{array}$

As mentioned already, the deterministic jitter and the random jitter are statistically independent from each other.

Thus, the histogram of the time interval error related to total jitter may be determined by convolution of the histograms of the time interval errors related to deterministic jitter and random jitter.

Joint Random Jitter and Random Noise Analysis

The analysis module 20 is configured to separate the vertical random noise and the horizontal random jitter contained within the input signal.

More precisely, the analysis module is configured to perform a joint random jitter and random noise analysis of the input signal in order to separate and/or determine the vertical random noise and the horizontal random jitter.

First, the determined data dependent jitter signal x_DDJ(t/T_b), which is determined in step S.4.4, and the determined vertical periodic noise signal x_pN(v) (step S.3.2) are subtracted from the input signal, labelled in the following by x_TJ(t/T_b), thereby generating a perturbation signal n₀(t/T_b), which is determined to be

$\begin{array}{cc}{n}_0\left(t\text{/}T_b\right)=x_{\mathrm{TJ}}\left(t\text{/}T_b\right)-x_{\mathrm{DDJ}}\left(t\text{/}T_b\right)-x_{\mathrm{PN}\left(v\right)}\left(t\text{/}T_b\right)=-\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}\varepsilon \left(k\right)\text{/}T_b·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t\text{/}T_b-k,b\left(k\right)\right)+x_{\mathrm{RN}\left(v\right)}\left(t\text{/}T_b\right).& \left(E\mathrm{.30}\right)\end{array}$

The perturbation signal n₀(t/T_b) approximately only contains horizontal random jitter, vertical random noises x_RN(v)(t/T_b) and horizontal periodic jitter, wherein the temporal jitter

$\begin{array}{cc}\varepsilon \left(k\right)\text{/}T_b={\varepsilon }_{\mathrm{PJ}}\left(k\right)\text{/}T_b+{\varepsilon }_{\mathrm{RJ}}\left(k\right)\text{/}T_b=\sum _{i=0}^{N_{\mathrm{PJ}\left(h\right)}-1}a_i\text{/}T_b·\sin \left(2\pi ·{\vartheta }_i\text{/}f_b·k+{\varphi }_i\right)+{\varepsilon }_{\mathrm{RJ}}\left(k\right)\text{/}T_b& \left(E\mathrm{.31}\right)\end{array}$

approximately only contains horizontal random jitter and horizontal periodic jitter.

As already mentioned, more than a single bit period may be taken into account.

The next step performed by the analysis module 20 is to determine the horizontal periodic jitter components.

For this purpose, a time variant equalizer filter ĥ_e(k,t/T_b) is applied to the perturbation signal n₀(t/T_b). The time variant equalizer filter ĥ_e(k,t/T_b) is determined from a time variant equalizer filter {tilde over (h)}_e(k,t/T_b) that is defined by:

{tilde over (h)} _e(k,t/T_b)=[b(−k)−b(−k+1)]·h(k−t/T_b,b(k)).   (E.32)

More precisely, the time variant equalizer filter is determined by minimizing the following cost functional K, for example by applying a least mean squares approach:

$\begin{array}{cc}K=\sum _{f\text{/}f_b}{\frac{1}{{\tilde{H}}_e\left(f\text{/}f_b\right)}-\sum _k{\hat{h}}_e\left(k\right)·e^{-j2\pi ·f\text{/}f_b·k}}^2& \left(E\mathrm{.33}\right)\end{array}$

Therein, {tilde over (H)}_e(f/f_b) is the Fourier transform of the time variant equalizer filter {tilde over (h)}_e(k,t/T_b). Of course, this analysis could also be performed in time domain instead of the frequency domain as in equation (E.33).

The resulting time variant equalizer filter is then applied to the perturbation signal n₀(t/T_b) such that a filtered perturbation signal is obtained, which is determined to be

$\begin{array}{cc}{\tilde{n}}_0\left(k,t\text{/}T_b\right)={\varepsilon }_{\mathrm{PJ}}\left(k\right)\text{/}T_b+{\tilde{\varepsilon }}_{\mathrm{RJ}}\left(k\right)\text{/}T_b=\sum _k{\hat{h}}_e\left(k,t\text{/}T_b\right)·n_0\left(t\text{/}T_b-k\right)=\sum _{i=0}^{N_{\mathrm{PJ}\left(h\right)}-1}a_i\text{/}T_b·\sin \left(2\pi ·{\vartheta }_i\text{/}f_b·k+{\varphi }_i\right)+{\tilde{\varepsilon }}_{\mathrm{RJ}\left(h\right),\mathrm{RN}\left(v\right)}\left(k\right)\text{/}T_b,& \left(E\mathrm{.34}\right)\end{array}$

Now, the frequencies ϑ_i and the phases φ_i are roughly estimated at first and then the amplitudes {circumflex over (α)}_t, the frequencies {circumflex over (ϑ)}_i and the phases {circumflex over (φ)}_i are determined jointly. For this purpose, the following cost functional

$\begin{array}{cc}K=\sum _{t/T_b}{\left[n_0\left(t/T_b\right)+\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}\sum _{i=0}^{N_{{\mathrm{PJ}\left(h\right)}^{-1}}}\frac{{\hat{a}}_i}{T_b}·\sin \left(2\pi ·{\hat{\vartheta }}_i/f_b·k+{\hat{\varphi }}_i\right)·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t/T_b-k,b\left(k\right)\right)\right]}^2& \left(E\mathrm{.35}\right)\end{array}$

is minimized analogously to the joint parameter analysis method outlined above that corresponds to step S.3.2. shown in FIG. 3.

If there is no duty cycle distortion or if the duty cycle distortion present in the input signal is much smaller than the horizontal periodic jitter, a time invariant equalizer filter {tilde over (h)}_e(k,t/T_b)=h(k−t/T_b) may be used for determining the time invariant equalizer filter ĥ_e(k,t/T_b).

In this case, the filtered perturbation signal

$\begin{array}{cc}\begin{array}{c}{\tilde{n}}_0\left(k,t/T_b\right)=\left[b\left(k\right)-b\left(k-1\right)\right]·{\varepsilon }_{\mathrm{PJ}}\left(k\right)/T_b+{\tilde{\varepsilon }}_{\mathrm{RJ}}\left(k\right)/T_b\\ =\sum _k{\hat{h}}_e\left(k,t/T_b\right)·n_0\left(t/T_b-k\right)\\ =\left[b\left(k\right)-b\left(k-1\right)\right]·\sum _{i=0}^{N_{\mathrm{PJ}\left(h\right)}-1}a_i/T_b·\\ \sin \left(2\pi ·{\vartheta }_i/f_b·k+{\varphi }_i\right)+{\tilde{\varepsilon }}_{\mathrm{RJ}\left(h\right),\mathrm{RN}\left(v\right)}\left(k\right)/T_b,\end{array}& \left(E\mathrm{.36}\right)\end{array}$

still comprises a modulation [b(k)−b(k−1)] that is due to the bit sequence, which is however known and is removed before roughly estimating the frequencies ϑ_i and the phases φ_i.

With the determined amplitudes {circumflex over (α)}_i, the determined frequencies {circumflex over (ϑ)}_i and the determined phases {circumflex over (φ)}_i, the horizontal periodic jitter signal is now reconstructed to be

$\begin{array}{cc}{\hat{x}}_{\mathrm{PJ}\left(h\right)}\left(t/T_b\right)=-\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}\sum _{i=0}^{N_{{\mathrm{PJ}\left(h\right)}^{-1}}}{\hat{a}}_i/T_b·\sin \left(2\pi ·{\hat{\vartheta }}_i/f_b·k+{\hat{\varphi }}_i\right)·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t/T_b-k,b\left(k\right)\right).& \left(E\mathrm{.37}\right)\end{array}$

Based on the reconstructed horizontal periodic jitter signal, a random perturbation signal n₁(t/T_b) is determined by subtracting the reconstructed horizontal periodic jitter signal shown in equation (E.37) from the perturbation signal. The determined random perturbation signal reads

$\begin{array}{cc}{n}_1\left(t/T_b\right)=n\left(t/T_b\right)-{\hat{x}}_{\mathrm{PJ}\left(h\right)}\left(t/T_b\right)\approx -\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}{\varepsilon }_{\mathrm{RJ}}\left(k\right)/T_b·\left[b\left(k\right)-b\left(k-1\right)\right]·h\left(t/T_b-k,b\left(k\right)\right)+x_{RN\left(v\right)}\left(t/T_b\right)& \left(E\mathrm{.38}\right)\end{array}$

and contains approximately only horizontal random jitter represented by the first term in the second line of equation (E.38) and vertical random noise represented by the second term in the second line of equation (E.38).

Generally speaking, the analysis module 20 now applies a statistical method to the signal of equation (E.38) at two different times in order to determine two statistical moments that are associated with the horizontal random jitter and with the vertical random noise, respectively.

More specifically, the analysis module 20 determines the variances σ_RF(h)² and σ_RN(v)² that are associated with the horizontal random jitter and with the vertical random noise, respectively, based on equation (E.38). Note that both the horizontal random jitter and the vertical random noise are normal-distributed. Further, they are statistically independent from each other.

According to a first variant, the conditional expected value of n₁²(t/T_b) for a particular realization (u,v) of the memory range is used and is determined to be

$\begin{array}{cc}{E\left\{n_1^2\left(t/T_b\right)\right\}}_{\left(uv\right)}\approx \frac{{\sigma }_{{ϵ}_{\mathrm{RJ}}}^2}{T_b^2}·\sum _{i=0}^{N-1}P_{\left(u_i,v_i\right)}·{\hspace{1em}\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}{\left[b\left(k\right)-b\left(k-1\right)\right]}^2·h^2\left(t/T_b-k,b\left(k\right)\right)}_{\left(u_i,v_i\right)}+{\sigma }_{\mathrm{RN}\left(v\right)}^2.& \left(E\mathrm{.39}\right)\end{array}$

Therein, P_(ui,vi) is the probability of the permutation (u_i, v_i). N is the number of permutations that are taken into account. Thus, the accuracy and/or the computational time can be adapted by varying N. In some embodiments, a user may choose the number N.

According to a second variant, all possible permutations (u_i, v_i) are taken into account in equation (E.39), such that an unconditional expected value of n₁²(t/T_b) is obtained that reads

$\begin{array}{cc}E\left\{n_1^2\left(t/T_b\right)\right\}\approx \frac{{\sigma }_{{\varepsilon }_{RJ}}^2}{T_b^2}·\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}P_r·2^2·h_r^2\left(t/T_b-k\right)+P_f·2^2·h_f^2\left(t/T_b-k\right)+{\sigma }_{\mathrm{RN}\left(v\right)}^2,& \left(E\mathrm{.40}\right)\end{array}$

Therein, P_r and P_f are the probabilities for a rising signal flank and for a falling signal flank, respectively.

If there is no duty cycle distortion present or if the duty cycle distortion is very small, equation (E.391 simplifies to

$\begin{array}{cc}E\left\{n_1^2\left(t/T_b\right)\right\}\approx \frac{{\sigma }_{{ϵ}_{\mathrm{RJ}}}^2}{T_b^2}·\left[2E\left\{b^2\left(k\right)\right\}-2E\left\{b\left(k\right)·b\left(k-1\right)\right\}\right]·\sum _{k=-N_{\mathrm{pre}}}^{N_{\mathrm{post}}}h^2\left(t/T_b-k\right)+{\sigma }_{RN\left(v\right)}^2.& \left(E\mathrm{.41}\right)\end{array}$

The analysis module 20 is configured to determine the variances σ_RJ(h)² and σ_RN(v)² based on at least one of equations (E.39) to (E.41). More precisely, the respective equation is evaluated for at least two different times t/T_b. For example, the signal edge time t₀/T_b=0 and the time t₁/T_b=½ may be chosen.

As everything except for the two variances is known in equations (E.39) to (E.41), the variances σ_RJ(h)² and σ_RN(v)² are then determined from the resulting at least two equations. It is to be noted that the variances σ_RJ(h)² and σ_RN(v)² correspond to the respective standard deviations.

In order to enhance accuracy, the equations can be evaluated at more than two times and fitted to match the measurement points in an optimal fashion, for example by applying a least mean squares approach.

Alternatively or additionally, only the variance σ_RN(v)² may be determined from the equations above and the variance σ_RJ(v)² may be determined from the following relation

$\begin{array}{cc}{\sigma }_{\mathrm{RJ}\left(v\right)}^2/T_b^2={\sigma }_{RN\left(v\right)}^2·\sum _iP_i·{\left[\frac{1}{dx_{{\mathrm{DDJ}}_i}\left(t_{\mathrm{edge}}/T_b\right)/\left(dt/T_b\right)}\right]}^2& \left(E\mathrm{.42}\right)\end{array}$

As the horizontal random jitter and the vertical random jitter are statistically independent, the variance σ_RJ(h)² is then determined to be

σ_RJ(h)²/T_b²=σ_RJ²/T_b²−σ_RJ(v)²/T_b²   (E.43)

Therein, P_i is the probability that a signal edge with slope dx_DDJi(t_edge/T_b)/(dt/T_b) arises. The numerical complexity of this method can be reduced by only taking into account substantially different slopes to contribute to the sum of equation (E.42).

Separation of Random Jitter and Other Bounded Uncorrelated Jitter

The analysis module 20 is further configured to determine a probability density f_x0(x₀) of a Gaussian random variable, for instance the random jitter, and a probability density f_x1(x₁) of a non-Gaussian bounded random variable, for instance the other bounded uncorrelated jitter.

For instance, the separation of the random jitter and other bounded uncorrelated jitter may be done by modelling the random jitter x₀ with a standard deviation σ_RJ whereas the other bounded uncorrelated jitter x₁ is random having the probability density f_x1(x₁).

The probability distribution may read as follows

$F_x\left(x\right)=P_0·\left[1-\frac{1}{2}\mathrm{erfc}\left(\frac{x-{\mu }_0}{\sqrt{2}·\sigma }\right)\right]+P_1·\left[1-\frac{1}{2}\mathrm{erfc}\left(\frac{x-{\mu }_1}{\sqrt{2}·\sigma }\right)\right].$

A mathematical scale transformation Q_x(x) as already described may be applied so that

$Q_x\left(x\right)={\mathrm{erfc}}^{-1}\left(2-2·\left(P_0+P_1\right)+P_0·\mathrm{erfc}\left(\frac{x-{\mu }_0}{\sqrt{2}·\sigma }\right)+P_1·\mathrm{erfc}\left(\frac{x-{\mu }_1}{\sqrt{2}·\sigma }\right)\right)$

is obtained, wherein the line obtained by the mathematical scale transformation may correspond to

${{Q_x\left(x\right)}_{\mathrm{left}}\approx \frac{x-{\mu }_0}{\sqrt{2}·\sigma },\mathrm{and}Q_x\left(x\right)}_{right}\approx \frac{x-{\mu }_1}{\sqrt{2}·\sigma }$

for respective ends of the mathematical scale transformation.

The standard deviation σ and the parameters μ₀, μ₁ may be determined.

The standard deviation σ may also be determined differently, for instance as already described above.

In the input signal, the random jitter and the other bounded uncorrelated jitter are superposed. Therefore, a collective probability density f_x(x) is given by a convolution of the individual probability densities with x=x₀+x₁, i.e.

$\begin{array}{cc}{f}_x\left(x\right)={\int }_{-\infty }^{+\infty }f_{x_0}\left(\xi \right)·f_{x_1}\left(x-\xi \right)·d\xi & \left(E\mathrm{.43}\right)\end{array}$

Transformed into frequency domain, the convolution of equation (E.43) becomes a mere product. The Fourier transform f_x0(f/f_a), i.e. the spectrum of the random jitter probability density f_x0(x₀), is normal distributed and reads:

$\begin{array}{cc}{F}_{x_0}\left(f/f_a\right)=e^{-2{\pi }^2·\ddot{{\sigma }^2}·{\left(\frac{f}{f_a}\right)}^2}.& \left(E\mathrm{.44}\right)\end{array}$

This property is employed in the separation of the random jitter component and the other bounded uncorrelated jitter component.

For example, the spectrum is determined based on measurements of the input signal and by matching the function of equation (E.44) to the measurement data.

In FIG. 13, an overview is shown wherein the probability densities of the random jitter, the other bounded uncorrelated jitter as well as a superposition of both are illustrated.

Alternatively or additionally, the variance σ_RJ² of the random jitter may already be known from one of the steps described above.

The probability density of the random jitter component is then determined to be

$\begin{array}{cc}{\hat{f}}_{x_0}\left(x_0\right)=\frac{1}{\sqrt{2\pi }·\hat{\sigma }}·e^{-\frac{x_0^2}{2{\hat{\sigma }}^2}}& \left(E\mathrm{.45}\right)\end{array}$

Based on the result of equation (E.45), the probability density f_x1(x₁) of the other bounded uncorrelated jitter component is then determined by a deconvolution of equation (E.43).

This is achieved by minimizing the following cost functional K, for example via a least mean squares approach:

$K=\sum _{x=x_{\min }}^{x_{\max }}{\left[f_x\left(x\right)-\sum _{\xi =x_{1,\min }}^{x_{1,\max }}{\hat{f}}_{x_1}\left(\xi \right)·{\hat{f}}_{x_0}\left(x-\xi \right)\right]}^2.$

Thus, the histogram of the other bounded uncorrelated jitter component can be determined.

Accordingly, histograms of all jitter components may be determined as already mentioned and shown in FIG. 3.

Certain embodiments disclosed herein utilize circuitry (e.g., one or more circuits) in order to implement protocols, methodologies or technologies disclosed herein, operably couple two or more components, generate information, process information, analyze information, generate signals, encode/decode signals, convert signals, transmit and/or receive signals, control other devices, etc. Circuitry of any type can be used.

In an embodiment, circuitry includes, among other things, one or more computing devices such as a processor (e.g., a microprocessor), a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a system on a chip (SoC), or the like, or any combinations thereof, and can include discrete digital or analog circuit elements or electronics, or combinations thereof. In an embodiment, circuitry includes hardware circuit implementations (e.g., implementations in analog circuitry, implementations in digital circuitry, and the like, and combinations thereof).

In an embodiment, circuitry includes combinations of circuits and computer program products having software or firmware instructions stored on one or more computer readable memories that work together to cause a device to perform one or more protocols, methodologies or technologies described herein. In an embodiment, circuitry includes circuits, such as, for example, microprocessors or portions of microprocessor, that require software, firmware, and the like for operation. In an embodiment, circuitry includes an implementation comprising one or more processors or portions thereof and accompanying software, firmware, hardware, and the like.

The present application may reference quantities and numbers. Unless specifically stated, such quantities and numbers are not to be considered restrictive, but exemplary of the possible quantities or numbers associated with the present application. Also in this regard, the present application may use the term “plurality” to reference a quantity or number. In this regard, the term “plurality” is meant to be any number that is more than one, for example, two, three, four, five, etc. The terms “about,” “approximately,” “near,” etc., mean plus or minus 5% of the stated value. For the purposes of the present disclosure, the phrase “at least one of A and B” is equivalent to “A and/or B” or vice versa, namely “A” alone, “B” alone or “A and B.”. Similarly, the phrase “at least one of A, B, and C,” for example, means (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C), including all further possible permutations when greater than three elements are listed.

The principles, representative embodiments, and modes of operation of the present disclosure have been described in the foregoing description. However, aspects of the present disclosure which are intended to be protected are not to be construed as limited to the particular embodiments disclosed. Further, the embodiments described herein are to be regarded as illustrative rather than restrictive. It will be appreciated that variations and changes may be made by others, and equivalents employed, without departing from the spirit of the present disclosure. Accordingly, it is expressly intended that all such variations, changes, and equivalents fall within the spirit and scope of the present disclosure, as claimed.

Claims

1. A jitter determination method for determining at least one random jitter component of an input signal, wherein the input signal is generated by a signal source, comprising: receiving said input signal;determining a time interval error associated with the random jitter component;determining at least one statistical moment of said time interval error based on the determined time interval error, wherein the order of said statistical moment is two or larger;at least one of determining an impulse response based on the input signal and receiving said impulse response, the impulse response being associated with at least said signal source; anddetermining the standard deviation of the random jitter component based on at least one of the determined statistical moment and the determined impulse response.

2. The jitter determination method of claim 1, wherein a linearized mathematical relation is used to determine said statistical moment, wherein the mathematical relation correlates the time interval error with a random jitter component of said input signal and to a slope of a data dependent jitter signal.

3. The jitter determination method of claim 1, wherein the at least one statistical moment comprises at least one of an autocorrelation function of said time interval error and a power spectral density.

4. The jitter determination method of claim 3, wherein at least one conditional autocorrelation function is determined in order to determine said autocorrelation function.

5. The jitter determination method of claim 4, wherein the at least one conditional autocorrelation function is determined by calculating an expected value of a product of a function representing the random jitter component with the same function shifted by a shifting parameter.

6. The jitter determination method of claim 4, wherein at least two conditional autocorrelation functions are summed with appropriate joint probability factors being the coefficients of the conditional autocorrelation functions in order to determine said autocorrelation function.

7. The jitter determination method of claim 6, wherein the joint probability factors are at least one of preset and calculated based on possible permutations of a symbol sequence contained within the input signal.

8. The jitter determination method of claim 1, wherein at least one of a data dependent jitter signal and a slope of said data dependent jitter signal is determined in order to determine the time interval error.

9. The jitter determination method of claim 8, wherein the at least one of the data dependent jitter signal and the slope of said data dependent jitter signal is evaluated at a clock time associated with a clock signal or at an edge time associated with a signal edge.

10. The jitter determination method of claim 1, wherein said input signal is PAM-n coded, wherein n is an integer bigger than 1.

11. The jitter determination method of claim 1, wherein the input signal is decoded, thereby generating a decoded input signal.

12. A measurement instrument for determining at least one random jitter component of an input signal, comprising at least one input channel and an analysis circuit being connected to the at least one input channel, the measurement instrument being configured to receive an input signal via said input channel and to forward the input signal to the analysis circuit,the analysis circuit being configured to determine a time interval error associated with the random jitter component,the analysis circuit being configured to determine at least one statistical moment of said time interval error based on the determined time interval error, wherein the order of said statistical moment is two or larger,the analysis circuit being configured to at least one of determine an impulse response based on the input signal and receive said impulse response, the impulse response being associated with at least said signal source, andthe analysis circuit being configured to determine the standard deviation of the random jitter component based on at least one of the determined statistical moment and the determined impulse response.

13. The measurement instrument of claim 12, wherein the at least one statistical moment comprises at least one of an autocorrelation function of said time interval error and a power spectral density.

14. The measurement instrument of claim 13, wherein the analysis circuit is configured to determine at least one conditional autocorrelation function in order to determine said autocorrelation function.

15. The measurement instrument of claim 14, wherein the analysis circuit is configured to determine the at least one conditional autocorrelation function by calculating an expected value of a product of a function representing the random jitter component with the same function shifted by a shifting parameter.

16. The measurement instrument of claim 12, wherein the analysis circuit is configured to sum at least two conditional autocorrelation functions with appropriate joint probability factors being the coefficients of the conditional autocorrelation functions in order to determine said autocorrelation function.

17. The measurement instrument of claim 12, wherein the analysis circuit is configured to calculate the joint probability factors based on possible permutations of a symbol sequence contained within the input signal.

18. The measurement instrument of claim 17, wherein the measurement instrument comprises an interface via which a user may preset the joint probability factors.

19. The measurement instrument of claim 17, wherein the analysis circuit is configured to determine at least one of a data dependent jitter signal and a slope of said data dependent jitter signal in order to determine the time interval error.

20. The measurement instrument of claim 19, wherein the analysis circuit is configured to evaluate the at least one of the data dependent jitter signal and the slope of said data dependent jitter signal a clock time associated with a clock signal or at an edge time associated with a signal edge.

21. The measurement instrument of claim 12, wherein the measurement instrument is established as at least one of an oscilloscope, a spectrum analyzer and a vector network analyzer.

22. The measurement instrument of claim 12, wherein the analysis circuit is configured to decode said input signal such that a decoded input signal is generated.