US 20080126010 A1
A method a low cost and production-integrable technique for providing a signal diagram. The data signal is edge-detected and asynchronoulsy sampled (or alternatively a clock signal is latched). The data signal or a second signal is compared to a settable threshold voltage and sampled. The edge and comparison data are folded according to a swept timebase tofind a minimum jitter period. The crossing of the signal diagram edges is determined from a peak of a histogram of the folded edge data. A histogram of ratios of the sample values versus displacement from the position of the crossing location is generated for each threshold voltage. The technique is repeated over a range of settable threshold voltages. Then, the ratio counts are differentiated across the histograms with respect to threshold voltage, from which a signal diagram is populated.
1. A method of generating a signal diagram of a signal under measurement, comprising:
collecting first samples of indications of edge positions of a first signal at regular intervals having a sampling period differing from a multiple of a period of said first signal;
collecting second samples of values of said signal under measurement at said regular intervals;
determining a timebase period corresponding to a relationship between said sampling period and said period of said first signal from said first samples;
analyzing said second samples in accordance with said determined timebase to provide said signal diagram of said values of said signal under measurement across said period of said first signal; and
displaying said signal diagram.
2. The method of
3. The method of
4. The method of
5. The method of
6. The method of
7. The method of
8. The method of
comparing said signal under measurement to a variable threshold voltage to obtain said values of said signal under measurement;
setting said threshold voltage to each of a plurality of threshold voltage values during said collecting second samples, whereby said samples provide a cumulative distribution function of a value of said signal under measurement for each of said threshold voltage values;
populating bins of histograms, one histogram for each of said threshold voltage values, with a ratio of values of said said second samples to obtain a cumulative distribution histogram; and
computing a rate of change of counts in corresponding bins of said cumulative distribution histograms with respect to change in corresponding threshold voltage values to obtain said signal diagram.
9. The method of
computing an indication of an amount of effective jitter of said first signal for said estimated period; and
adjusting said timebase period until a timebase period yielding a minimum amount of said effective jitter is found.
10. The method of
This application is related to co-pending U.S. application Ser. No. 11/279,651 entitled “METHOD AND APPARATUS For Determining Jitter and Pulse Width from Clock Signal COMPARISONS” filed on Apr. 13, 2006; Ser. No. 11/______, Atty. Docket Number AUS920050913US1, entitled “METHOD OF Generating an Eye Diagram for Integrated Circuit Transmitted SIGNALS” and Ser. No. 11/______, Atty. Docket Number AUS920060199US1, entitled “METHOD AND APPARATUS FOR DETERMINING DATA SIGNAL JITTER VIA ASYNCHRONOUS SAMPLING” filed concurrently with this application, by the same Inventors and assigned to the same Assignee. The specifications of the above-referenced applications are incorporated herein by reference.
This invention was made with Government support under NBCH3039004, DARPA. THE GOVERNMENT HAS CERTAIN RIGHTS IN THIS INVENTION.
1. Technical Field
The present invention relates generally to digital interface circuits, and more particularly, to circuits and systems that evaluate data integrity and jitter via an eye diagram or other data diagram display.
2. Description of the Related Art
Eye diagrams are commonly used to display the integrity and jitter characteristics of a data signal. Typically, eye diagrams are used to view binary data, but have also been used in analog systems to view a modulation pattern or other analog signal variation and can also be applied to digital signals having more than two levels.
In general, an eye diagram is displayed by synchronizing the data channel of an oscilloscope with the data rate of the signal under observation. The result is that the display of the data window is made stationary via the synchronization of the timebase to the data rate, thereby permitting the observation of the amount and distribution of jitter at the data window edges, as well as observing the integrity of the data within the “eye”, which is the generally empty area between the transition regions and between the two signal levels. For example, a change in data within the eye such as a spike, indicates a mechanism other than jitter that may cause data errors. At the edges of the eye, the width of the diagram gives and indication of the amount of jitter, and if gaps are seen in the transition regions, chaotic and other meta-stable phenomena can be observed as specific lines within the jitter.
Data jitter and integrity determination is necessary to evaluate the performance of high-speed interface components and interfaces, as well as other circuits where jitter affect the bit error rate (BER) and eye diagrams provide a valuable tool for such determination. Other types of signal diagrams may also be employed in synchronization with a data signal. For example, a second signal or power supply voltage may be observed for noise, crosstalk or other variation that is synchronous with the data signal used to synchronize the timebase.
In laboratory environments, high-accuracy laboratory instruments may be used to provide eye diagrams via stable reference clocks. However, the challenge of probing a very high frequency data signal and/or high-impedance data signal is significant, as the effects of the probe must be accounted for in the measurements and probe characteristics can vary over time and the probe compensation model may not be accurate under actual measurement conditions. Further, significant circuit area can be consumed in the impedance-matched and isolated output pads that permit such precision measurements. Such equipment is expensive and it is typically unfeasible to incorporate the equivalent of such instrumentation within production circuits.
On-chip sampling of a data signal is typically either performed using an internal global clock or an external sample clock with an internal delay line. Use of the internal global clock is limited in that clock jitter and data jitter cannot be separated. When an external sample clock is used, the delay accuracy, and thus the accuracy of the presumed positions of the sampled data, is difficult to control.
It is therefore desirable to provide a method and apparatus for generating a signal diagram that is low cost, can be at least partially integrated in a production circuit with no probing error.
The above-stated objectives of providing a low-cost apparatus and method for generating a signal diagram is provided in a method and apparatus.
The method analyzes raw data provided by the apparatus, and the performance of the method may be distributed between a test circuit and a workstation computer or may be completely embedded in a test instrument or production integrated circuit.
The method samples edges of a signal provided from an edge detector using a sampling clock signal that differs at least slightly in frequency from rational multiples of the data rate of interest. The method also collects the values of the signal or a second signal via comparison with a settable voltage threshold. Alternatively, a clock signal can be sampled rather than edge-detected and the samples used to provide the edge position indications.
The sampled edge position data is folded into a unit interval according to a guess of the relationship between the periods/frequencies of the sampling clock and the data rate. The guess is adjusted either over a range, or until an indication that the guess is correct is obtained, such as minimum jitter in the folded edge position data.
The period obtained above is used to generate a timebase to fold the edge position data and the signal value samples. The frequency or period of neither the sampling clock signal or edge data signal needs to be known in order to further analyze the data, only the relationship between the two, as determined by the method described above.
The values of the signal value samples are then folded according to period determined above and the entire process is repeated over a range of settable threshold voltage, in order to generate histograms of sampled data values for time slots within the period. The histograms correspond to the cumulative distribution function (cdf) for each threshold voltage, which are then differentiated to obtain the probability density function (pdf) for each threshold voltage. The pdfs can then be used to populate the signal diagram, which may be an eye diagram of the data signal that was edge-detected, or may be a signal diagram of another signal that was sampled along with the edges of the data signal.
Drift between the frequency of the sampling clock signal and the data signal providing the edge data can be removed from the analysis results by using a linear or other shift model to progress the guessed period across the sample set, and the exact frequency drift can be identified from the histogram having the minimum widths of peaks in the probability density function.
The sampling circuit can be included in a production die and the data collected by a workstation computer via direct probing or a boundary-scan data interface. Alternatively, a processing circuit or algorithm may be present or loaded into a production integrated circuit that includes a processor or a dedicated digital circuit use to conduct the above-described analysis. The analysis method may be embodied in a computer-program product containing stored program instructions for carrying out the steps of the method.
The foregoing and other objectives, features, and advantages of the invention will be apparent from the following, more particular, description of the preferred embodiment of the invention, as illustrated in the accompanying drawings.
Details of the invention and the uses thereof will be understood by a person of ordinary skill in the art, when reading the following detailed description of illustrative embodiment in conjunction with the accompanying drawings. Further objectives and advantages presented by the invention will be apparent in light of the following description and drawings, wherein like reference numerals indicate like components, and:
The present invention concerns the generation and display of signal diagrams, and in particular, the eye diagram of interface data signals. In order to produce accurate results, the capture of measurement data is generally performed by a sampling circuit within the interface (or other data-receiving circuit) rather than a separate test probe. However, the techniques of the present invention can be applied to and within test equipment, as well.
Primarily, the present invention provides new methods for generating and displaying a signal diagram obtained by sampling a signal under measurement with an asynchronous sampling clock simultaneously with sampling edge position of either the signal under measurement of or another signal for which synchronous variation in the signal under measurement is to be observed. If the reference source is a clock signal, then the clock signal can be sampled with a latch to obtain the edge positions, otherwise an edge detector is used. Rather than synchronizing the sampling clock with the signal under measurement or other reference source, as is done within a digital oscilloscope, the present invention resolves the sampled data to identify a relationship between the period of the signal for which edges that were sampled relative to the sampling clock. The method then analyzes the sample values of the signal under measurement according to the determined relationship between edge position samples and the sampling clock to generate the signal diagram.
Therefore, the sampling clock used to sample the signal under measurement and the signal for which edges are sample (and integer multiples of its frequency) must differ in frequency at least slightly and a sufficient number of samples must be collected so that jitter can be distinguished from variation due to slight frequency differences. Neither the period/frequency of the data signal under measurement (and optional other edge position data source) nor the sampling clock period/frequency need be known in order to generate the signal diagram. If the edge position data source and the signal under measurement are the same, then the generated signal diagram is an eye diagram of the signal under measurement.
With reference now to the figures, and in particular with reference to
The circuit required for data collection is illustrated separately as a measurement circuit 11 and includes a data edge detector 20, a sampling clock 15 and a storage 16 for samples of edges of interface 12 Data signal collected from the output of an edge detector 20 at edges of sampling clock 15. Alternatively, if the edge position reference source is a clock signal such as the Clock signal shown optionally provided from interface 12, then a latch 21 can be used to provide the edge position data that is stored in storage 16. Storage 16 also collects samples of the outputs of a signal value detector 22 at the same time that edge position samples are collected from data edge detector 20 (or alternatively from latch 21). Signal value detector, for the purposes of generating eye diagrams, receives the Data signal of interface 12, but may alternatively be provided with input from optional Other signals for which characteristics in synchronization with the Data signal may be observed (e.g., a power supply rail internal to electrical unit 10B might be observed for noise synchronous with Data signal).
A processor 18 and memory 19 are optionally included for performing methods according to the present invention, or the raw sample data from storage 16 may be clocked out by a test system via boundary latches 17 or otherwise read from unit 10B via interface 13B, e.g., by electrical unit 10A as Sample Data over interface 12. Also, if processor 18 and memory 19 are included, and the sampled data is processed locally, interface 13B and/or boundary latches 17 may be read to retrieve the results of the analysis performed by a method according to the present invention.
Some degree of tuning of sampling clock 15 must generally be provided, so that the sampling clock frequency will not land on an exact multiple of the data rate of the Data signal (or period of the Clock signal), thus generating a zero beat frequency that will yield only a signal dependent on the presence or absence of data signal edges in the folded data and not revealing the jitter, which will be used to determine the correct relationship between the data rate of the Data signal and the sampling clock. Sampling clock 15 is therefore shown coupled to boundary latches in order to provide to least a single bit adjustment that can change the clock frequency of sampling clock 15 if needed to avoid sampling too close to a multiple of the data rate of the Data signal. Sampling clock 15 may also be optionally provided from an external source such as a test system.
Referring now to
A workstation computer 38, having a processor 36 coupled to a memory 37, for executing program instructions from memory 37, wherein the program instructions include program instructions for executing one or more methods in accordance with an embodiment of the present invention, is coupled to wafer tester 30, whereby the sampled clock data (or analysis results from processor 18 and memory 19 of
Workstation computer 38 is also coupled to a graphical display 39 for displaying program output such as the signal diagram computed by embodiments of the present invention, as well as graphical data such as the graphs depicted in
While the system of
Referring now to
Referring now to
Referring now to
If the edge reference source is not a clock signal, a reasonably large number of edge samples is required due the edge detector output forming a very small portion of the overall data period. Therefore, the sample values may be encoded in storage via a run-length or other compression code, and also the number of edge samples collected may be larger than that of the data value samples. For example, the threshold voltage may be ramped or programmed to different values during the collection of data and edge values in one measurement so that the number of edge samples is larger than each individual set of per-threshold data samples. In order to conserve storage space, edge location values such as those stored in storage 16 of
Referring now to
Next, the timebase is adjusted according to the relative position between each sample and the eye crossing location, and the data values are placed in histogram bins according to the timebase (step 55). The ratio of logical “1” counts to logical “0” counts is computed for each histogram bin, forming a histogram representing the cumulative distribution function of the data signal values for each position within the unit interval (step 56). Until all comparator threshold voltages (VT) are scanned (decision 57), steps 50-56 are repeated for each threshold voltage. Finally, the signal diagram is generated by differentiating the cdf at each bin position with respect to the change in threshold voltage between the histograms, i.e., computing a rate of change of the count ratios in corresponding bins with respect to the change in threshold voltage between histograms (step 58). The result of the differentiation yields a histogram of the probability density across the period of the data signal, i.e., the pdf of the signal values. Next the eye diagram is displayed from the pdf histogram (step 59). If the method is used to generate a signal diagram as described above, rather than an eye diagram, the steps taken are the same; however, the histogram and derivative will not be the cdf and pdf of the data values, but rather the cdf and pdf of the signal under measurement.
Referring now to
The above-described steps from decision 62 to step 67 are repeated for all increments of TG up to T/2 (decision 68). Also, until all desired drift corrections are applied (decision 69), new drift corrections are applied (step 70) and step 61 through decision 68 are repeated. After the edge position data values have been analyzed over all timebase periods and drift corrections, the TG period for which the minimum jitter is present (step 71) is taken as the correct clock period and the edge data histogram for the period is also used in step 54 of
Referring now to
Referring now to
The very low jitter values at periods 82 and 83 are due to the DC value yielded from the sample set being read at effective identical positions within the waveform of the clock under measurement, i.e. coherent sampling, which is rejected by the above-described method in step 62 of
Referring now to
The description provided above constitutes a description of the preferred embodiments of the invention, but the invention is not limited to the particular implementations shown or described. Those skilled in the art, having seen the above description and accompanying drawings, will understand that changes in form, structure and other details, as well as the order of operation of any operative steps may be varied without departing from the spirit and scope of the invention.