New Pipeline Processing Measures Every Instance Providing Unique Views of Signal History
Scope designs have always focused on accurate portrayals of signal activity. But regardless of trigger rates, acquisition dead times called "gaps" occur between each trigger creating time slices of missing signal information.
The problem was getting larger as many signals evolved to higher complexities over longer periods of time. Views that contain gaps masked problems that need to be seen or measured. Up until now, the deeper the memory, the greater the processing load on a scope's CPU which further increased gap times between new acquisitions. Then everything changed.
The newest solution is to use a single trigger with a very long pipeline memory that is coupled to an extremely fast processing architecture. The combination provides a new and unique view of any signal's true history.
( Previously, a scope with a one million sample point (1 Mpts) acquisition memory was considered deep while scopes with only 50 kpoints were considered to have short memory.
Ironically, yesterday's "short" memory scope that sampled at 1 GS/s (one sample per nanosecond) into 50 kpoints of memory captured exactly the same amount of signal length as a 20 GS/s (20 samples per ns) that has 1 Mpt. Both of them are short in memory as they can only capture a short period of time (50 usec) at their maximum sample rates. )
The recent advance of acquisition memory length pushes the bar up to 100 million sample points (100 Mpts) in a single acquisition with 1 or 2 channels. This nonlinear increase of oscilloscope acquisition memory also produces a new operational model. Now each sweep is more than 126 miles of contiguous scope screens. At 20 GS/s, each sweep is up to 5 ms with no gaps. (Note: Most recent advances have 3 manufacturer's tied at 40 GS/s and 1 has just announced 80 GS/s in < 6 months.)
Consider the following equation:
Each 10.4" diagonal scope screen displays 8" of waveform data horizontally. Viewing a 3.125 Gb/s differential XAUI signal, 15 bits will be shown when the time-base is 500 ps/div or 5 ns across the display. At 20 GS/s, this time-base setting corresponds to 100 samples being displayed.
With 100 Mpts acquired instead, the scope memory will now contain 15,625,000 consecutive XAUI bits. With the time-base settings used in Figure 1, the 100 Mpt acquisition would occupy one million scope screens, or 126 miles of display distance. This 100 MSample record captures 5 ms of data at 20 GS/s. This allows measurement of periodic signal components down to 200 Hz while still maintaining the ability see spectral components up to 10 GHz. A long continuous acquisition is the best way to simultaneously observe both low frequency components (jitter due to power supply noise or low frequency modulation) or high frequency content (cycle-cycle jitter, inter-symbol interference) of a complex signal.
New Operational Model for Long Memory Scopes It is not practical to manually scroll through one million scope screens. Long memory scopes with "All-Instance Measurements" (AIM) can supply the complete statistical picture for any measurements selected, and can hold the AIM data for additional processing operations.
Tracks are time-correlated trends of parameter measurements. This makes finding signal aberrations fast and easy. The data acquisition memory of channel one, the yellow trace, contains all the data samples acquired. The track (blue) trace provides additional insight into signal behavior that cannot be determined from the yellow trace alone.
Another advantage of long acquisitions is the elimination of trigger jitter from timing measurements. In every long acquisition, all the samples have the same time displacement due to trigger jitter. Any relative time measurement (period, frequency, delay, width, etc) automatically removes this common displacement thereby eliminating trigger jitter from the measurement.
Track vs. Persistence Using a short acquisition memory, a persistence display generates a visual trace history as shown in Figure 3. This method is limited by a number of factors. A persistence display cannot show the signal aberrations that appear in Figure 2. The order of occurrence in the underlying modulation in Figure 2 is invisible to persistence. The track view provides information not available from simple viewing methods like persistence.
Statistically Significant The long memory trace below contains 40 million sample points of a 500 MHz clock. Notice in Figure 4 there are almost one million rise-time and period measurements collected from a single trigger. Histograms for each measurement reveal additional insight about the signal's true dynamics. Note that zooming still allows traditional cycle-by-cycle viewing, showing one millionth of the original record, with the 50 ps resolution per point clearly seen.
Serial Data Analysis with Long Memory Here is another example of a long memory record. This eye pattern display is generated from a 3.5 Gb/s SERDES chip. The traditional approach to collecting this data was to use a sampling scope in persistence mode. The screen image below (Figure 5) is unique in that the data was taken with just one long memory acquisition. This single block of data was then partitioned into unit intervals and overlaid to create this eye pattern. Unlike a standard persistence method, this display is a result of exactly one acquisition and therefore contains no trigger jitter.
Because the entire record is stored in memory, all of the data is available to perform BER analysis, ISI, random and deterministic jitter, and identification of which exact bits were the source of each mask failure. Additionally, this process is over 1000 times faster than traditional scopes.
Long scope memory provides the following benefits:
With the complexity of today's signals, long scope memory scope records reveal information not available from any other method. This new capability is a breakthrough in rapid diagnostics of issues associated with complex waveforms.