A method for generating a test dataset for testing a receiver, a computer program product, and a test device, as well as the test dataset.

Abstract

This invention provides a method for generating a test dataset (TDS) for testing receivers, as well as test equipment. [Solution] A method for generating a TDS to adapt receiver 4 includes the steps of: applying the TDS to the receiver; capturing a receiver-side dataset (EDS) based on the TDS; evaluating the EDS to identify a plurality of misreceived symbols (FES) and the respective locations P of the FES in the EDS; forming a plurality of error-inducing datasets (FVS) by copying each section of the TDS having a predetermined length around each misreceived FES; evaluating the FVS to identify error-inducing locations (FVP) and / or error-inducing patterns (FVM); generating at least one improved error-inducing dataset (VFVS) based on the FVP and / or FVM; and modifying the TDS by embedding at least one VFVS within the TDS.

Description

[Technical Field]

[0001] The present invention relates to a method for generating a test dataset for testing a receiver, a computer program product, and a test device, as well as to the test dataset itself. [Background technology]

[0002] Since 2010, data rates in modern high-speed interfaces (HSIs) have doubled every 3-4 years. During this time, multiple high-speed interfaces, such as PCIe 6 and 7, IEEE 802.3ck, and dj.MIPI M-PHY G6, have achieved data rates exceeding 40 Gbit / s.

[0003] However, conventional methods for receiver adaptation and verification are not suitable for fully digital receivers, especially those with data rates exceeding 100 Gbit / s. Conventional adaptation and verification methods identify metrics such as eye diagrams, bathtub diagrams, or signal-to-noise ratio (SNR) by evaluating the output signal according to the receiver's equalization. These methods are not suitable for newer equalization schemes such as Maximum Likelihood Symbol Detection (MLSD) because discrete symbols are provided on the output side, and no signal is provided, as in Continuous Time Linear Equalization (CTLE) or Decision Feedback Equalization (DFE).

[0004] Testing high-speed interfaces such as PCIe 7 and IEEE 802.3dj is extremely difficult. On the one hand, the bandwidth of test equipment, such as oscilloscopes, signal analyzers, and vector network analyzers, does not increase as quickly as the transmission rate. Therefore, for example, there are few mass-produced oscilloscopes that provide sufficient bandwidth for calibrating PCIe-7 test signals at over 100 GHz. On the other hand, at high transmission rates, compensating for the impact of test equipment and units on the signal becomes increasingly difficult. These problems may necessitate replacing classic conformance testing with on-chip test / self-test functions, a situation already present for UCIe interfaces with very short transmission channels.

[0005] The error rate of modern receivers is 10 for practical length. 10 Test sequences shorter than a symbol cannot be measured with high confidence. There are two main reasons for this: 1. The probability of an error occurring at a particular symbol in the data stream depends on many symbols before and after it. One cause of this is reflection due to impedance errors in the transmission channel. To ensure that this method captures reflections with high reliability, the test data should include all symbol sequences that demonstrate the effect of reflections. Since the location of reflections is unknown in advance, the test data should include all conceivable partial sequences of a given length. 2. The multistep signal codes currently in use mean that each symbol has not only two possible values, as in the case of non-return-to-zero coding (NRZ) used previously, but also four values, as in the case of PAM4, and eight values, as in the case of PAM8. This means that a test dataset containing all symbol subsequences of a given length must be much longer than in the case of non-return-to-zero coding.

[0006] For example, in the case of PCIe 5-7, one of the reflection sources is the plug connector between the add-in card and the system board. The typical time of flight between the plug connector and the die on the add-in card is approximately t=0.7ns, so the reflection from the plug connector occurs with a delay of 2t=1.4ns. For PCIe 5, 6, and 7, this time is 45UI (unit interval), 45UI, and 90UI, respectively, which specifies the minimum length of the symbol sequence that will be used for testing. For non-zero return codes, the length of the test dataset that combines all binary sequences of 45UI is 2 45 ≈32·10 12 This is in symbolic units. Testing PCIe 5 RX using such a test dataset takes only 1000 seconds, which is less than 30 minutes. In PAM4, the length of the test dataset is 4 45 ≒10 27 This is in symbolic units. Testing PCIe 6 RX using this dataset would take millions of years. The situation is much worse with PCIe 7, where the length of the test dataset is 4 90 Because it is a symbolic expression.

[0007] Measuring the symbol error rate (SER) presents problems in the case of such high-speed interfaces. Receiver behavior is characterized by the long-term or "true symbol error rate," defined as the ratio of symbols incorrectly received by the receiver to the total number of symbols transmitted across all conceivable data sequences. The measured symbol error rate for a given test dataset is either higher than or equal to the true symbol error rate. Therefore, practical test datasets used contain only a small percentage of conceivable symbol sequences and cannot accurately assess the true symbol error rate. Since most loaded symbol sequences are (with a high probability) underestimated within the test dataset, the measured symbol error rate is likely to be lower than the true symbol error rate. [Overview of the project]

Problems to be Solved by the Invention

[0008] Therefore, an object of the present invention is to disclose a method capable of realizing improvement in this regard.

Means for Solving the Problems

[0009] The object of the present invention is achieved by a method for generating a test data set for testing, particularly adapting, a receiver, which comprises the following steps, namely, making a test data set be applied to the receiver; capturing a receiver-side data set based on the test data set; evaluating the receiver-side data set to identify a plurality of symbols received inaccurately and their respective positions in the receiver-side data set; forming a plurality of error-inducing data sets by copying respective sections having a predetermined length of the test data set around each inaccurately received symbol; evaluating the error-inducing data sets to identify a plurality of error-inducing positions and / or error-inducing patterns, and generating at least one improved error-inducing data set based on the error-inducing positions and / or error-inducing patterns; changing the test data set by embedding at least one improved error-inducing data set into the test data set; and the like.

[0010] The receiver may be a receiving individual component designed as another component, or a part of a chiplet, or a component designed for self-test (built-in test). [[ID=​The test dataset can be an initial test dataset that is modified and optimized for multiple test passes. In other words, testing can start with the initial test data, then continue after generating a modified and optimized test dataset.

[0012] A test dataset can contain, for example, 5,000 to 100,000 symbols. Symbols can be binary (logical 1 and logical 0) or multi-valued, such as tetra-valued symbols, as in the case of PAM4.

[0013] The test dataset and the receiver dataset provided by the receiver are compared with each other to identify the symbols that were received incorrectly and their respective locations within the test dataset.

[0014] Subsequently, sections of a predetermined length from the test dataset are copied around each location of the incorrectly received symbols. These copied sections form multiple error-inducing datasets.

[0015] The error-inducing dataset is then modified to generate at least one improved error-inducing dataset. For this purpose, one or more symbols are modified at a location different from the predetermined error-inducing location. In other words, the incorrectly transmitted symbols remain untouched at that location, and other sections of the error-inducing dataset are modified. These could be individual symbols or discovered error-inducing patterns.

[0016] Finally, one or more improved error-inducing datasets are embedded within the initial test dataset or the next generation of the test dataset at that point, thus modifying the test dataset. For another test pass, it can then be verified whether the error frequency increases as a result of the changes to the test dataset. An increase in error frequency is assessed as indicating a more critical and improved test dataset, and these changes are adopted for another test pass. Conversely, if an increase in error frequency is not demonstrated, the changes are not adopted, and the updated test pass is run using the previous generation of the test dataset in other forms of modification.

[0017] Using this method, distortions within the transmission link (e.g., uneven reflection locations) can be identified even if an equalized output signal is not available, for example, by using Maximum Likelihood Sequence Detection (MLSD).

[0018] By identifying error-inducing patterns, this method can calculate the symbol error rate more accurately than by counting errors on a predetermined test sequence. This is particularly important for multi-level signals such as PAM4 and above, where the test dataset may not be representative.

[0019] This method can be used for any signal modulation and is not limited to, for example, non-zero return codes or PAM4.

[0020] This method makes it possible to automatically identify weak points in the receiver based on the identified error-inducing locations and error-inducing patterns.

[0021] The iterative transmission of test datasets with small changes allows for the automatic generation of complex statistics, which show how symbols before and after an error affect the probability of the error.

[0022] According to one embodiment, the step of generating at least one improved error-inducing dataset by varying each error-inducing dataset at locations other than the location of the incorrectly received symbol is as follows: The steps include generating at least one evaluation dataset by copying the symbols of each of the multiple error-inducing datasets to the same location in each of the multiple error-inducing datasets, The steps include evaluating the evaluation dataset and identifying symbols for a new improved error-inducing dataset at that location, Includes.

[0023] In other words, each error-inducing dataset in a set of error-inducing datasets can be understood as a row in a matrix, and the evaluation dataset can be understood as a vector where each of the corresponding positions in the error-inducing datasets is filled with its respective symbol. Thus, the evaluation dataset combines multiple errors from the set of error-inducing datasets.

[0024] Subsequently, the evaluation dataset is evaluated to identify the symbols in the improved error-inducing dataset at that location. For this purpose, for example, the arithmetic mean of the symbol values ​​in the evaluation dataset can be identified. For instance, if the identified arithmetic mean falls within the upper limit of a possible numerical range, such as PAM4, this indicates that symbols with fairly high values ​​within the upper limit are causing the error. Conversely, if the identified arithmetic mean falls within the lower limit of a possible numerical range for the PAM4 signal, this indicates that symbols with fairly low values ​​within the lower limit are causing the error. Other methods are possible instead of identifying the arithmetic mean. For example, the geometric mean or root mean square can also be identified instead of the arithmetic mean. Furthermore, the median can also be identified. The method can therefore be greatly improved.

[0025] In another embodiment, the step of generating at least one improved error-inducing dataset by changing each error-inducing dataset at locations other than the location of the incorrectly received symbol is as follows: Steps to sort error-inducing datasets according to error frequency. Includes.

[0026] In this way, a list of error datasets is created, with the error-inducing dataset with the highest error frequency at the top and the error-inducing dataset with the lowest error frequency at the bottom. In this way, it is particularly easy to select the error-inducing dataset with the highest error frequency.

[0027] In another embodiment, the step of generating at least one improved error-inducing dataset by changing each error-inducing dataset at locations other than the location of the incorrectly received symbol is as follows: Steps to group error-inducing datasets Includes.

[0028] Therefore, groups of error-inducing datasets are formed. Group formation can be based on error frequency, that is, the group with the highest error frequency is formed, followed by a second group with a lower error frequency than the first group, then a third group with a lower error frequency than the second group, and so on. Group formation can also be based on other characteristics of the error-inducing dataset, such as the type and location of the incorrectly transmitted symbols, or on symbol sequences, such as error-inducing patterns within the error-inducing dataset. The method can therefore be further improved.

[0029] In another embodiment, the step of generating at least one improved error-inducing dataset by changing each error-inducing dataset at locations other than the location of the incorrectly received symbol is as follows: Steps to identify an improved error-inducing dataset by changing the location of error-inducing patterns within the error-inducing dataset. Includes.

[0030] Therefore, error-inducing patterns are removed from their initial point in the error-inducing dataset and inserted into the dataset at other points, thereby identifying an improved error-inducing dataset. The method can thus be further improved.

[0031] In another embodiment, the step of generating at least one improved error-inducing dataset by changing each error-inducing dataset at points other than the error-inducing location is as follows: Steps to balance the test dataset Includes.

[0032] Balancing the test dataset should be understood here as balancing or compensating for changes made elsewhere in the improved error-inducing dataset. Balancing or compensating can aim to distribute multiple symbols evenly. In the case of the PAM4 signal, this would mean that each of the four possible symbols would appear with the same frequency in the test dataset. Alternatively, balancing or compensating can aim to keep the energy content of the test dataset constant, i.e., constant over time across the entire test dataset.

[0033] Furthermore, the present invention includes a computer program product designed to perform such a method, a test dataset identified by such a method, and test equipment.

[0034] The present invention will be described in more detail below with reference to the attached schematic diagram. [Brief explanation of the drawing]

[0035] [Figure 1] This shows a schematic representation of the test equipment used to test the receiver. [Figure 2] This provides a detailed outline of how to test the receiver. [Figure 3] Here is a summary of another detail on how to test the receiver. [Figure 4] A schematic representation of the sequence for testing the receiver is shown. [Modes for carrying out the invention]

[0036] First, refer to Figure 1.

[0037] A test apparatus is shown that has a test device 2 for testing a receiver 4 having a high-speed interface.

[0038] In this exemplary embodiment, the test device 2 can be designed as a separate component, for example, a transmitter component, particularly a test unit component, or as part of a chiplet or a component designed for self-testing (built-in testing).

[0039] In this exemplary embodiment, receiver 4 can be designed as a separate component, for example, as a receiver component, or as part of a chiplet, or as a component designed for self-testing (built-in testing).

[0040] A chiplet contains multiple chips (DIEs), which are also different, and all are housed within a single housing (multi-die package). Chiplets can be combined monolithically and non-monolithically to form a complete system function.

[0041] Users include computing centers, high-performance computing (HPC), and artificial intelligence (AI). Chiplets are also increasingly being used in the entertainment electronics, telecommunications, and automotive markets.

[0042] During operation, test device 2 ensures that the initial test dataset TDS is applied to receiver 4, and receiver-side dataset EDS, based on the test dataset TDS provided by receiver 4, is input to test device 2.

[0043] Based on the evaluation of the initial test dataset TDS and the receiver-side dataset EDS, test device 2 modifies the test dataset TDS and applies it again to receiver 4. The process then continues through another pass, and the test dataset TDS is modified each time. In other words, an iterative process is performed, and the test dataset TDS is refined step by step multiple times.

[0044] For this purpose, for the tasks and / or functions described below, test apparatus 2 may include hardware and / or software components designed to correspond thereto. The hardware components may be computer components, such as a processor and memory in which a computer program product for performing this method, executed by the processor, is stored.

[0045] Therefore, the test device 2 of this exemplary embodiment is designed to generate an initial test dataset TDS, and then another test dataset TDS, so that these are applied to the receiver 4.

[0046] The test dataset TDS may contain, for example, 5,000 to 100,000 symbols. In this exemplary embodiment, a quaternary signal, such as a PAM4 signal, is used. Notwithstanding this exemplary embodiment, a binary signal or a higher-value signal, such as a PAM8 signal, can also be used.

[0047] Furthermore, the test equipment 2 of this exemplary embodiment is designed to capture a receiver-side dataset EDS based on an initial or subsequent test dataset TDS, evaluate the receiver-side dataset EDS, and identify multiple error-inducing locations FVP and / or error-inducing patterns FVM within the receiver-side dataset EDS.

[0048] For this purpose, the test device 2 of this exemplary embodiment is designed to compare the initial and subsequent test dataset TDS with their respective corresponding receiver-side dataset EDS to identify the inaccurately received symbols FES and their respective locations within the test dataset TDS.

[0049] Furthermore, the test instrument 2 of this exemplary embodiment is designed to find error-inducing patterns FVMs within a section of the test dataset TDS having a predetermined length, around the incorrectly transmitted symbols.

[0050] The test apparatus 2 of this exemplary embodiment is further designed to copy each section of the test dataset TDS having a predetermined length around each incorrectly received symbol FES to form multiple error-inducing datasets FVS.

[0051] In addition, the test apparatus 2 of this exemplary embodiment is designed to evaluate the error-inducing dataset FVS to identify multiple error-inducing locations FVP and / or error-inducing patterns (FVMs).

[0052] In addition, the test apparatus 2 of this exemplary embodiment is designed to generate multiple improved error-inducing datasets VFVS by varying each error-inducing dataset FVS based on the error-inducing location FVP or error-inducing pattern FVM, which will be described in more detail later.

[0053] Finally, the test apparatus 2 of this exemplary embodiment is designed to embed and modify the improved error-inducing dataset VFVS thus formed into each test dataset TDS.

[0054] Next, refer to Figure 2.

[0055] Four error-inducing datasets (FVS) are presented, which are identified, for example, by comparing the initial test dataset TDS with the corresponding receiving dataset EDS.

[0056] In this exemplary embodiment, each error-inducing dataset FVS has a length of 11 symbols S. In each case, the symbol FES with the parenthetical value 0 at the 10th point was incorrectly received. In other words, the error-inducing dataset FVS includes the nine symbols S before it and the one symbol S after it, in addition to the incorrectly received symbol FES at the 10th point.

[0057] Furthermore, in this exemplary embodiment, the error-inducing pattern FVM was identified by comparison in the first and third error-inducing datasets FVS. It extends from the fifth to the eighth position in the first error-inducing dataset FVS, while it extends from the fourth to the seventh position in the third error-inducing dataset FVS.

[0058] Note that each of the four error-inducing datasets (FVS) was detected as inaccurate multiple times and was grouped accordingly. In other words, the four error-inducing datasets (FVS) are included multiple times within the test dataset (TDS). Furthermore, the four error-inducing datasets (FVS) were separated according to their error frequency.

[0059] In this exemplary embodiment, 40% of the errors occurred in the first of the four error-inducing datasets FVS, 20% of the errors occurred in the second and third of the four error-inducing datasets FVS, and 10% of the errors occurred in the fourth of the four error-inducing datasets FVS. Another 10% of the errors occurred in other datasets (not shown). Thus, in this exemplary embodiment, 90% of the errors occurred in the four error-inducing datasets FVS.

[0060] Furthermore, in this exemplary embodiment, 60% of the errors occurred after the error-inducing pattern FVM.

[0061] A further criterion for evaluation is that a predetermined minimum number of errors were detected.

[0062] To evaluate the four error-inducing datasets FVS, evaluation datasets ADS are formed by the test instrument 2, specifically one evaluation dataset ADS for each symbol position. For this purpose, each symbol S of the error-inducing dataset FV is copied to the corresponding position P of the error-inducing dataset FVS, compiled, and the respective evaluation dataset ADS is formed.

[0063] In this exemplary embodiment, the four error-inducing datasets FVS can be understood as matrices, and each evaluation dataset ADS can be understood as a vector.

[0064] The evaluation dataset ADS thus formed is then evaluated. For this purpose, in this exemplary embodiment, the mean or arithmetic mean of the values ​​of symbol S in each evaluation dataset ADS is determined by the test instrument 2. In the case of evaluation dataset ADS, a value of 0.5 is obtained for the first position P, and a value of 2.5 is obtained for the third position P. A value of 2.5 is also obtained for the last position P.

[0065] Assuming that errors occur particularly frequently at extreme values, i.e., at the maximum value, which in this case is 3 in the case of PAM4, and at the minimum value, which in this case is zero, the first value of 0.5 represents the minimum value of zero, and the value of 2.5 represents the maximum value of 3.

[0066] Therefore, as a result, the four error-inducing datasets FVS will have the same structure, having the format 0x3xxxxxx(0)3, that is, having a symbol S with a minimum value of zero at the first position P and a symbol S with a maximum value of 3 at the third and last positions P.

[0067] Next, refer to Figure 3.

[0068] The study begins with four error-inducing datasets (FVS) from the first generation, and then shows how the four subsequent generations and the two subsequent error-inducing datasets (FVS) are formed.

[0069] The four error-inducing datasets FVS of the second generation were modified by test equipment 2 as follows to obtain the improved error-inducing dataset VFVS. According to the values ​​identified above, at the first position P of the first and fourth error-inducing datasets FVS of the first generation, the symbol S with a value of 1 is replaced with the symbol S with a value of 0. Furthermore, according to the values ​​identified above, at the third error-inducing dataset FVS of the first generation, the symbol S with a value of 2 at the third position P is replaced with the symbol S with a value of 3. Furthermore, at the second error-inducing dataset FVS of the first generation, the symbol S with a value of 0 at the seventh position P is replaced with the symbol S with a value of 2, reproducing the error-inducing pattern FVM, which has been previously identified in other error-inducing datasets FVS.

[0070] In addition, balance can be achieved using test equipment 2. For this purpose, for example, a symbol S with a value of 3 in the test dataset TDS at any third position P of the four error-inducing datasets FVS, or outside the four error-inducing datasets FVS, can be replaced with a symbol S with a value of 1. Furthermore, for example, a symbol S with a value of zero in the test dataset TDS at any first and / or fourth position P of the four error-inducing datasets FVS, or outside the four error-inducing datasets FVS, can be replaced with a symbol S with a value of 1.

[0071] The second generation of four error-inducing datasets, FVS or improved error-inducing dataset VFVS, are now used in yet another pass so that they are applied again to receiver 4. For this purpose, the improved error-inducing dataset VFVS is embedded within the test dataset TDS.

[0072] Evaluation of error frequency using test equipment 2 revealed that the error frequency could not increase in the first and third of the four error-inducing datasets FVS, but could increase in the second and fourth of the four error-inducing datasets FVS.

[0073] In the second of the four error-inducing datasets FVS, the symbol S with value 3 at the eighth position is replaced by the test device 2 with the symbol S with value 1, and in the fourth of the four error-inducing datasets FVS, the symbol S with value 3 at the seventh position P is replaced by the symbol S with value 2, thereby obtaining the error-inducing pattern FVM.

[0074] The remaining two error-inducing datasets FVS of the third generation are now used in a different pass by test instrument 2, as described above, so that they are applied again to receiver 4 using the modified test dataset TDS. Test instrument 2's evaluation of error frequencies shows that the error frequencies cannot increase again in the first and third of the four error-inducing datasets FVS, but they can increase in the second and fourth of the four error-inducing datasets FVS.

[0075] In the second of the four error-inducing datasets FVS, test instrument 2 replaces the symbol S with a value of zero at the ninth position P with the symbol S with a value of 2, and in the fourth of the four error-inducing datasets FVS, the symbol S with a value of zero at the sixth position P is replaced with the symbol S with a value of 1, thereby reproducing the error-inducing pattern FVM.

[0076] The remaining two error-inducing datasets FVS of the third generation are now used in a different pass by test instrument 2, as described above, so that they are applied again to receiver 4. Evaluation of error frequencies by test instrument 2 yields the result that the error frequency cannot increase in the second of the four error-inducing datasets FVS, but can increase in the fourth of the four error-inducing datasets FVS.

[0077] In the fourth of the four error-inducing datasets FVS, test instrument 2 replaces symbol S with a value of 2 at the last position with symbol S with a value of 3. This is because the evaluation dataset ADS at the last position P yields a value of 2.5, and as a result, as mentioned above, symbol S with a value of 3 is selected.

[0078] If this change does not increase the error frequency, then we can conclude that the final position P does not affect the error frequency and is not an error-inducing dataset (FVS). In another pass, the final change can be reversed by test instrument 2, and a different change can be performed instead.

[0079] As an extension of this exemplary embodiment, in the evaluation of the evaluation dataset ADS, each value of symbol S can be weighted by their respective error rates, and therefore, for example, by the aforementioned values ​​of 40%, 20%, and 10%.

[0080] Furthermore, as an evolution from this exemplary embodiment, in the evaluation of the evaluation dataset ADS, the test device 2 first checks, as a partial step, whether the difference between the transmitted value of symbol S in the test dataset TDS and the corresponding value of symbol S in the receiving dataset EDS at the respective error locations in the test dataset TDS and the receiving dataset EDS is greater than or less than zero.

[0081] If this difference is greater than zero, the corresponding value of symbol S in the evaluation dataset ADS at the error-inducing location FVP, which is different from the error location, remains unchanged. Conversely, if the difference is less than zero, as another partial step, symbol S in the evaluation dataset ADS is inverted by the test instrument 2.

[0082] In this exemplary embodiment, for inversion, the value W of symbol S at the error-inducing position FVP is subtracted from the maximum value Smax of the signal, i.e., Smax-W. In the case of the PAM4 signal, the maximum value is 3, and therefore 3-W.

[0083] For example, the following occurs in three error-inducing datasets FVS: 50% of the errors were 1130(0)3. 25% of the errors were 3100(3)0. 25% of errors occurred with code 0321(0)3. The symbols S that were incorrectly received next to the last position of these character sequences are zero and 3.

[0084] The error rates for each of these three error-inducing datasets (FVS) are 0.5 and 0.25, occurring twice each.

[0085] Furthermore, for the first and third of the three error-inducing datasets FVS, a reversal occurs because zero was received as the incorrectly received symbol S. Therefore, the correct symbol S must represent a value W greater than zero, and the difference must be less than zero. Conversely, for the second of the three error-inducing datasets FVS, the difference must be greater than zero because 3 was received as the symbol.

[0086] Then, the average profile value based on the first evaluation dataset ADS at the first location is calculated as follows: = 0.5·(3-1) + 0.25·3 + 0.25·(3-0) = 1 + 0.75 + 0.75 = 2.5.

[0087] Similarly, different average profile values ​​based on different evaluation datasets (ADS) based on different locations are 1.25, 0.25, 2, 3, and zero.

[0088] Generally, the average profile value can be calculated as follows:

number

number

[0089] To adapt the symbol S of the error - inducing data set FVS to obtain the improved error - inducing data set VFVS, threshold comparison is performed using a lower threshold and an upper threshold, and each value can be changed only when it exceeds the lower threshold or the upper threshold. For example, in the case of PAM4, the lower threshold can be 0.5 and the upper threshold can be 2.5.

[0090] Furthermore, the upper threshold oSW and the lower threshold uSW can be adapted gradually, that is, during each pass. For this purpose, the number of non - random errors ANF regarding the total number of errors GF is weighted using a predetermined weight value, for example, using a weight value GW = 0.5, and then subtracted from the lower limit uOffset of 1.25 and added to the upper limit oOffset of 1.75: uSW = uOffset - GW·ANF / GF oSW = oOffset + GW·ANF / GF.

[0091] More generally expressed, this parameter characterizes the randomness of the errors, and this method can adapt itself to the channel state. In a transmission channel with very high noise, the ratio of the number of non - random errors ANF to the total number of errors GF can be made close to zero, while in a transmission channel with very high inter - symbol interference (ISI), the ratio of the number of non - random errors ANF to the total number of errors GF is close to 1.

[0092] For example, oSW=2.25 can be applied to the upper threshold, and uSW=0.75 can be applied to the lower threshold.

[0093] The average profile from the above example includes the following values: 2.5 1.25 0.25 2 3 0

[0094] Position P was selected for this update based on a comparison with the upper threshold oSW and the lower threshold uSW, and is shown in bold (the value "3" is not in bold because it is located at the error-inducing position FVP).

[0095] The symbol S in the improved error-inducing dataset VFVS represents a negative sign error (FZ). k =-1) and positive sign errors (FZ k The mean profile that maximizes the probability of =1) is obtained: 3 * 0 * (3)0 and 0 * 3 * (0)3.

[0096] The symbol S for each error-inducing dataset FVS is updated as follows to correspond to the mean profile value: - 1130(0)3 → -0130(0)3 (At the first position P, symbol 1 is replaced by symbol 0, which means this sequence is (FZ k (This is because it has -1, i.e., the inverted 0 is 3.) - 3100(3)0 -> No change - 0321(0)3→0331(0)3 (At the third position P, symbol 2 is replaced by symbol 3, which means this sequence is (FZ k It has =1), that is, the inverted 3 is 0).

[0097] In this exemplary embodiment, only one change is made in each error-inducing dataset FVS during each pass, thus obtaining an improved error-inducing dataset VFVS for embedding into the modified test dataset TDS.

[0098] Now, refer to Figure 4.

[0099] In the first step S100, the receiver 4 is provided with an initial test dataset TDS having a predetermined length by the test equipment 2.

[0100] In the next step, S200, test equipment 2 ensures that the initial test dataset TDS, or the next generation of the initial test dataset TDS, i.e., another test dataset TDS, is applied to receiver 4.

[0101] In the next step, S300, test device 2 captures the receiver-side dataset EDS based on its respective current test dataset TDS.

[0102] In the next step, S400, test equipment 2 evaluates the receiver-side dataset EDS to identify multiple error-inducing locations FVP.

[0103] In the next step S500, the test device 2 forms multiple error-inducing datasets FVS by copying each section of the test dataset TDS having a predetermined length around each incorrectly received symbol FES.

[0104] In the next step S600, the test instrument 2 evaluates the error-inducing dataset FVS to identify multiple error-inducing location FVPs and / or error-inducing pattern FVMs, and generates at least one improved error-inducing dataset VFVS based on the error-inducing location FVPs and / or error-inducing pattern FVMs.

[0105] In another step S610, the test instrument 2 sorts the error-inducing dataset FVS according to error frequency for this purpose.

[0106] In the next step, S620, test equipment 2 groups the error-inducing dataset FVS for this purpose.

[0107] In the next step S630, the test instrument 2 generates at least one evaluation dataset ADS by copying the respective symbols S of the multiple error-inducing datasets FVS at the same position P of each of the multiple error-inducing datasets FVS.

[0108] In the next step, S640, the test instrument 2 identifies symbol S for the improved error-inducing dataset VSVF at position P by evaluating the evaluation dataset ADS.

[0109] In the next step, S650, test equipment 2 changes the position of the error-inducing pattern FVM within the error-inducing dataset FVS to identify the improved error-inducing dataset VFVS.

[0110] In the next step, S660, test device 2 balances the test dataset.

[0111] In another step S700, test equipment 2 modifies the test dataset TDS by embedding at least one of several improved error-inducing datasets VFVS into the test dataset TDS.

[0112] Based on the identified error-inducing location FVP and error-inducing pattern FVM, the receiver 4's ability to recover data from the input signal can be characterized. This characterization allows for, for example, the adaptation of clock data recovery settings (CDR). For example, the following options are also possible: 1. Adapt receiver 4 to minimize the impact of error-inducing location FVP. This can be done, for example, by applying FTDFE tap (Floating Tap Decision Feedback Equalization) at error-inducing location FVP, by adapting the Maximum Likelihood Sequence Detection parameter (MLSD), or by changing the equilibrium state between the transmitter and feedforward equalization (FFE). 2. Minimize the impact of error-inducing pattern FVM by optimizing clock data recovery (CDR). Error-inducing pattern FVM, in which certain symbol transitions are underestimated or overestimated, likely indicates a problem with clock data recovery. 3. Optimize Maximum Likelihood Sequence Detection (MLSD) to minimize the impact of error-inducing pattern FVM. Error-inducing pattern FVM may indicate suboptimal Maximum Likelihood Sequence Detection settings. 4. Calculate the true symbol error rate and its usage as performance features. In particular, the symbol error rate is calculated by assigning numbers from 1 to N to all combinations of error initiation locations FVP and / or error initiation patterns FVM, Σ i=1..N FR i ·p i It can be calculated as follows, in the formula, FR i is the error rate for the i-th combination of the error inducing location FVP and / or error inducing pattern FVM, and p i This is the probability of this combination.

[0113] For an exemplary calculation of the symbol error rate, the error rate for the following combinations of error initiation location FVP and / or error initiation pattern FVM is assumed to be determined as follows (irrelevant symbols are replaced with asterisks): 0 * 3 ** 1212(0) = 0.007 1 * 3 ** 1212(0) = 0.003 0 * 2 ** 1212(0) = 0.001 1 * 2 ** 1212(0) = 0.0006 0 * 3 * 1212 * (0) = 0.003 1 * 3 * 1212 * (0) = 0.001 0 * 2 * 1212 * (0) = 0.0007 1 * 2 * 1212 * (0) = 0.0001 0 * 3 ****** (0) = 0.000004 1 * 3 ****** (0) = 0.00002 0 * 2 ****** (0) = 0.00001

[0114] The error rates for other combinations should be negligible. The first eight combinations have eight specified PAM4 symbols, and their relative frequency in the data stream is 1 / (4 7 ) The last three combinations have three fixed PAM4 symbols, and their relative frequencies are 1 / (4 3 )

[0115] The results for the true symbol error rate are as follows: =0.001 * (7+3+3+1+1+0.7+0.06+0.3)4 7 +0.00001 * (4+2+1) / 4 3 =(0.0164) / 16384+(0.00007) / 64 =2.1069 * 10 -6 .

[0116] This example is simplified: if the probability of error depends only on eight preceding symbols S, then the symbol error rate itself is the same as a short test dataset (4 9 It can be estimated with high confidence from (=262144 symbols). The above method works the same way when the error initiation location FVP and / or error initiation pattern FVM are distributed across hundreds of symbols S, which is generally true in real transmission channels when the symbol error rate measured from a finite test dataset is likely an unreliable estimate of the true symbol error rate.

[0117] By using this method, distortions within the transmission channel (e.g., uneven reflection locations) can be identified even when an equalized output signal is unavailable, for example, when Maximum Likelihood Sequence Detection (MLSD) is used.

[0118] By identifying error-inducing patterns in FVM, this method can calculate the symbol error rate more accurately than by counting errors over a given test sequence. This is particularly important when the test dataset is likely not representative of multi-level signals, such as PAM4 and beyond.

[0119] This method can be used with any signal modulation and is not limited to, for example, non-zero return codes or PAM4.

[0120] This method enables the automatic identification of weak points in receiver 4 based on the identified error-inducing location FVP and / or error-inducing pattern FVM.

[0121] By iteratively transmitting a test dataset TDS with small changes, complex statistics can be automatically generated, showing how the symbols S before and after an error affect the probability of the error. [Explanation of symbols]

[0122] 2 Test equipment 4 Receiver ADS Evaluation Dataset EDS Receiver Dataset FES Incorrectly received symbol FVM Error Inducing Patterns FVP Error Trigger Location FVS Error-Inducing Dataset P position S symbol TDS Test Dataset VFVS Improved Error-Inducing Dataset S100 Step S200 Step S300 Step S400 Step S500 Step S600 Step S610 Step S620 Step S630 Step S640 Step S650 Step S700 Step

Claims

1. A method for generating a test dataset (TDS) for testing a receiver (4), particularly for adaptation, comprising the following steps: (S200) The step of applying the test data set (TDS) to the receiver (4), (S300) A step of capturing a receiver-side data set (EDS) based on the test data set (TDS), (S400) Evaluating the receiver-side dataset (EDS) to identify a plurality of incorrectly received symbols (FES) and their respective positions (P) within the receiver-side dataset (EDS), (S500) The steps of forming a plurality of error-inducing datasets (FVS) by copying each section of the test dataset (TDS) having a predetermined length around each of the inaccurately received symbols (FES), (S600) The steps of: (S600) Evaluating the error-inducing dataset (FVS) to identify a plurality of error-inducing locations (FVP) and / or error-inducing patterns (FVM), and generating at least one improved error-inducing dataset (VFVS) based on the error-inducing locations (FVP) and / or error-inducing patterns (FVM); (S700) The step of modifying the test dataset (TDS) by embedding at least one improved error-inducing dataset (VFVS) into the test dataset (TDS), A method that includes this.

2. The step (S600) of generating at least one improved error-inducing dataset (VFVS) by changing each of the error-inducing datasets (FVS) at positions (P) other than the position (P) of the incorrectly received symbol (FES) is as follows: (S630) A step of generating at least one evaluation dataset (ADS) by copying the respective symbols (S) of the plurality of error-inducing datasets (FVS) at the respective identified positions (P) of the plurality of error-inducing datasets (FVS), (S640) A step of evaluating the evaluation dataset (ADS) to identify the symbols (S) of the new improved error-inducing dataset (VFVS) at the position (P), The method according to claim 1, including the method described in claim 1.

3. The step (S600) of generating at least one improved error-inducing dataset (VFVS) by changing each of the error-inducing datasets (FVS) at positions (P) other than the position (P) of the incorrectly received symbol (FES) is as follows: (S610) Step of sorting the error-inducing dataset (FVS) according to the error frequency. The method according to claim 1 or 2, including the method described in claim 1 or 2.

4. The step (S600) of generating at least one improved error-inducing dataset (VFVS) by changing each of the error-inducing datasets (FVS) at positions (P) other than the position (P) of the incorrectly received symbol (FES) is as follows: (S620) Step of grouping the error-inducing dataset (FVS) The method according to any one of claims 1 to 3, including

5. The step (S600) of generating at least one improved error-inducing dataset (VFVS) by changing each of the error-inducing datasets (FVS) at positions (P) other than the position (P) of the incorrectly received symbol (FES) is as follows: (S650) A step to identify an improved error-inducing dataset (VFVS) by changing the position of the error-inducing pattern (FVM) in the error-inducing dataset (FVS). The method according to any one of claims 1 to 4, including

6. The step (S600) of generating at least one improved error-inducing dataset (VFVS) by changing each of the error-inducing datasets (FVS) at positions (P) other than the position (P) of the incorrectly received symbol (FES) is as follows: (S660) Step of balancing the test dataset (TDS) The method according to any one of claims 1 to 5, including

7. A computer program product designed to perform the method described in any one of claims 1 to 6.

8. A test dataset (TDS) specified by the method described in any one of claims 1 to 6.

9. A test device (2) for testing a receiver (4), which generates a test dataset (TDS) for particular adaptation, wherein the test dataset (TDS) is applied to the receiver (4), captures a receiver-side dataset (EDS) based on the test dataset (TDS), evaluates the receiver-side dataset (EDS) to identify a plurality of misreceived symbols (FES) and their respective positions (P) in the receiver-side dataset (EDS), and around each of the misreceived symbols (FES), the test data... Test equipment (2) is designed to modify the test dataset (TDS) by forming multiple error-inducing datasets (FVS) by copying each section of the TDS having a predetermined length, identifying multiple error-inducing locations (FVPs) by evaluating the error-inducing datasets (FVS), generating at least one improved error-inducing dataset (VFVS) based on the error-inducing locations (FVPs), and embedding at least one improved error-inducing dataset (VFVS) into the test dataset (TDS).

10. The test apparatus (2) according to claim 9, which is designed to generate at least one evaluation dataset (ADS) by copying the respective symbols (S) of the plurality of error-inducing datasets (FVS) at the same location (P) of each of the plurality of error-inducing datasets (FVS), and to evaluate the evaluation dataset (ADS) to identify the symbols (S) of the improved error-inducing dataset (VFVS) at the location (P).

11. The test apparatus (2) according to claim 9 or 10, which is designed to sort the error-inducing dataset (FVS) according to error frequency.

12. Test apparatus (2) according to claim 9, 10, or 11, which is designed to group the error-inducing dataset (FVS).

13. Test apparatus (2) according to any one of claims 9 to 12, which is designed to identify an improved error-inducing dataset (VFVS) by changing the position of the error-inducing pattern (FVM) in the error-inducing dataset (FVS).

14. A test apparatus (2) according to any one of claims 9 to 13, which is designed to balance the aforementioned test dataset (TDS).

Images