Data processing: measuring – calibrating – or testing – Measurement system – Measured signal processing
Reexamination Certificate
2003-01-31
2004-10-12
Assouad, Patrick (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system
Measured signal processing
C702S069000, C702S070000
Reexamination Certificate
active
06804633
ABSTRACT:
BACKGROUND OF THE INVENTION
Eye diagrams are a conventional format for representing parametric information about signals, and especially digital signals. Various prior art eye diagram testers are known, but we shall call the technique described in “METHOD AND APPARATUS FOR PERFORMING EYE DIAGRAM MEASUREMENTS” and “COMPOSITE EYE DIAGRAMS” (which is the preferred method for use herein), an Eye Diagram Analyzer, or EDA for short.
A modern eye diagram for a digital signal is not so much a trace formed continuously in the time domain (ala an analog oscilloscope), as it is an “eye” shape composed of closely spaced points (displayed dots, or illuminated pixels) representing many individual measurement samples (which are (time, voltage) pairs) taken upon separate instances of a signal occurring on a channel of interest, and which were then stored in a memory. Each measurement sample contributes to a displayed pixel. (In a simple case the correspondence is one to one, but the actual case might be many to one or one to many.) The eye shape appears continuous because the collection of pixels is rather dense, and because the signal is sampled at a large number of nearly adjacent locations. Unlike a true continuous technique, however, there may be detached dots that are separated from the main body of the eye shape. (Such detachment is an artifact of sampling.) A further difference with the continuous analog technique is that rare or infrequently occurring events, once sampled, do not appear faint in the display or disappear with the persistence of the CRT's phosphor. This latter difference is often quite an advantage, since it is often the case that such otherwise “hard to see” features of the trace are very much of interest.
In an eye diagram, an example of which is shown in
FIG. 1
, the vertical axis is voltage and the horizontal axis represents the differences in time (i.e., various offsets) between some reference event and the locations for the measurement samples. The reference event: is generally an edge of a clock signal in the system under test; represents directly or through some fixed delay the expected point in time when the value of an applied data signal would be captured by some receiving circuit in an SUT (System Under Test); and, is derived from an application of the SUT's clock to the Eye Diagram Analyzer. The time axis will generally have enough length to depict one complete eye-shape (cycle of an SUT signal) centered about the reference, with sometimes perhaps several additional eyes (cycles) before and after.
To one not familiar with eye diagrams
FIG. 1
looks like a poorly synchronized or unreliably triggered trace for a periodic digital signal. Let us dwell here briefly to dispel any mystery about why an eye diagram is the way it is. To begin with, it is assumed that a (non-return-to-zero) work signal being measured is synchronous with something, typically a system clock. Let us further suppose that work signal transitions are expected to occur on falling edges of the system clock, so that the work signal is expected to be stable for sampling on the rising edge of the system clock. If we always sampled the work signal at the same time (say, at the rising edge of the system clock), then the results would be repeated instances of one point on the work signal. If the work signal always had the same value (a steady one or a steady zero), then those repeated instances would be the same (time, voltage) values, and would contribute multiple instances of the same single point to the eye diagram. But keep these two things in mind: One, we do not always sample at the exact same time, but vary it on either side of the reference, by up to, or even more than, a clock period. Two, the work signal is typically not stuck at some steady value. It is sometimes a one, sometimes a zero, and sometimes it does not transition.
So, if the work signal were stuck at one or zero, then the varying sample times would produce adjacent points forming straight lines at either the voltage for one or for zero. And if the work signal had regular transitions of one zero one zero one zero . . . then the trace would resemble a square wave (as seen on a 'scope). But most work signals are not that regular: they are sometimes high, sometimes low, and sometimes they stay at the same value from clock cycle to clock cycle. So the eye diagram contains the superposition of all the individual oscillographic trace segments of the two straight lines (one for always high, one for always low) and the transitions from high to low and from low to high. For a properly working signal the central region of each eye is empty, since the good signal never transitions except when it is supposed to.
To continue, then, different (X, Y) regions within a (sample) space containing an eye diagram represent different combinations of time and voltage. Assume that the eye diagram is composed of a number of pixels, and temporarily assume that the resolution is such that each different (X, Y) pixel position can represent a different combination of time and voltage (and vice versa), which combinations of time and voltage we shall term “measurement points.” What the preferred Eye Diagram Analyzer measures is the number of times, out of a counted number of clock cycles, that the signal on the channel being monitored passed through a selected measurement point. Then another measurement point is selected, and the process repeated until there are enough measurement points for all the pixels needed for the display. Points along the visible eye diagram trace describe something about those (time, voltage) combinations that were observed to actually occur in the data signal under test. The value of a (time, voltage) combination is represented by its location, but the color or intensity of the measured result can be determined in a way that assists in appreciating the meaning of the measured data. The range over which the measurement points are varied is called a “sample space” and is defined during a measurement set-up operation. And in reality, we define the sample space and the resolution for neighboring measurement points first, start the measurement and then let the analyzer figure out later how to ascribe values to the pixels of the display. The “display” is, of course, an arbitrary graphic output device such as a printer or an X Window of some as yet unknown size in a window manager (e.g., X
11
) for a computer operating system. (A one-to-one correspondence between display pixels and measurement points is not required. It will be appreciated that it is conventional for display systems, such as X Windows, to figure out how to ascribe values to the pixels for an image when the correspondence between the display's pixel locations and the measurements that are the original image description is not one-to-one.)
Thus it is that a modern eye diagram trace itself is thus not a single time domain waveform (think: ‘single valued function’), but is instead equivalent to an accumulation of many such instances; it can present multiple voltage (Y axis) values for a given time value (X axis). So, for example, the upper left-hand region of an eye (see
FIG. 1
) might represent the combination of an adequate logical one at an adequately early time relative to the SUT's clock signal, and an eye diagram whose trace passes robustly through that region indicates to us that a signal of interest is generally achieving a proper onset of voltage at a proper time. Furthermore, we note that there are also other regions, say, near the center of an eye, that are not ordinarily transited by the trace, and which if that were indeed to happen, would presumably be an indication of trouble. Thickening of the traces is indicative of jitter, a rounding of a corner is indicative of slow transitions, and so on. An eye diagram by itself cannot reveal in the time domain which isolated instance (cycle) of the data signal caused such an exception, as other types of measurements might, but it does provide timely and valid information about signal integrit
Agilent Technologie,s Inc.
Assouad Patrick
Miller Edward L.
LandOfFree
Fast eye diagram analyzer uses nearest neighbor transverse... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Fast eye diagram analyzer uses nearest neighbor transverse..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Fast eye diagram analyzer uses nearest neighbor transverse... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3291326