A new device known as the advanced data stream engine (ADSE) is certain to catch the attention of many in the test and measurement field. It's been developed for use in the DL9000 series of digital oscilloscopes, bringing with it functionality and performance claimed to be superior to those of conventional signal processing engines.
The ADSE, a CMOS IC based on a 0.13µm process, contains data memory on the same chip to reduce bottlenecks caused by the memory bus. This signal-processing block has been configured using a dedicated architecture aimed at achieving high waveform-display update rates.
In general, the DL9000 scopes contain basic performance features that include a 4-channel analogue input, a maximum frequency bandwidth of 1.5GHz, a maximum sampling rate of 10GS/s, and a maximum recording length of 6.25Mword. In these oscilloscopes, the signalprocessing block (Fig. 1) is dedicated to the generation of display data from the A/D-converted data and to performing calculations for various waveforms and parameters.
One important application for a digital oscilloscope is the eye-pattern observation of transmission channels, which is traditionally achieved via an accumulation function that overlays and displays a multitude of waveforms. An eye pattern is a figure created by overlaying many waveforms provided as various symbol patterns, and is quite useful in debugging communications systems (including memory and system buses). Let's take a closer look at these patterns.
Eye-pattern applications
When using eye patterns for debugging purposes, the observer "browses" observation points to establish the status of the device under test (DUT), or to change the DUT's operating conditions to see how the change affects the device. Therefore, it's desirable to inform the observer of signal changes as quickly as possible. For this reason, the observer sets accumulation time intervals to eliminate unnecessary viewing of earlier waveforms.
An accumulation operation with set time intervals will draw waveforms one at a time as bitmapped images, and then erase them concurrently as the given time elapses. This means that the brightness of earlier waveforms gradually decreases, making accumulation operations ideally suited to observing the order in which phenomena occur. As a result, this mode of operation isn't optimal for eye-pattern observation or other applications in which multiple waveforms must be treated with the same degree of weight.
History memory function
Another way to observe eye patterns on a digital oscilloscope is with a history memory function. The history memory function is automatically activated whenever the amount of memory allocated for each acquisition is less than the maximum amount of memory the unit has in each channel. The total capacity of the history memory varies depending on the allocated acquisition memory size. The smaller the acquisition memory, the larger the capacity of history memory available in the unit. With history memory, it's possible to overlay or search through waveforms accumulated in the memory after waveform acquisition is stopped.
Figure 2 shows where the waveform of a CPU bus is displayed, along with the clock and chip select signals, using the history memory function. In this figure, colour accumulation converts the rate of occurrence of points into a colour image to emphasise the manner in which waveforms are overlaid. From this image, we can understand that the transition of data begins in synchronisation with the clock. We can also see that although most data items share the same transition time, some data items have longer transition times. Hence, overlaying a multitude of accumulated waveforms offers a great deal of useful information.
However, generating this image requires the instrument to deal with enormous amounts of data at high processing speeds. In addition, it couldn't be used in conjunction with the history memory function in traditional oscilloscope architectures.
ADSE dynamic images
With the new ADSE signalprocessing engine, though, it's possible—dynamically and continuously—to generate images equivalent to the overlaid historical waveform image. This can occur concurrently with waveform acquisition.
The number of waveforms overlaid by utilising this method is the same as can be accessed by the history memory function. A maximum of 2000 can be handled when the acquisition memory length is set to 2.5ksample points per waveform. As long as this maximum number isn't exceeded, the oscilloscope is able to develop the specified number of waveforms into bitmaps each time it generates the images of these waveforms. The scope then counts the number of sample points overlapping with one another on a pixel-by-pixel basis, and converts this frequency distribution to colour or brightness to generate the corresponding waveform images.
Comparing the different modes of waveform accumulation, Figure 3a shows an image in which waveforms are erased according to the lapse of time. Figure 3b illustrates an image in which 500 waveforms are overlaid with the same degree of weight and the frequency of sample points is assigned to brightness. In this example, the number of data items comprising each waveform has been set to 12.5k points.
In both accumulation modes, the oscilloscope transfers the accumulated waveforms to the LCD as images, making it possible to acquire and display many waveforms without being constrained by the LCD's refresh cycle. Hence, the oscilloscope can continuously acquire and display waveforms at a maximum trigger rate of 25,000 per second when 2.5k memory length is selected, or 9000 per second when the number of sample points per waveform is set to 12.5k.
This trigger rate is the same as that for single-channel operation, even when all four channels are simultaneously put in operation. Assuming that the instrument is using 4-channel operation with 12.5k sample points per waveform, the oscilloscope as a whole can process 450Msample points per second.
Tracking fast triggers
By using the accumulation function, it's now possible to keep track of significantly fast triggers. In addition, there's a function that will keep track of trigger signals with a 400ns dead time. Consequently, users are able to observe phenomena in those instances where dead time between the triggers becomes an issue.
When this "N single" trigger mode is selected, the oscilloscope will only show waveforms after a designated number of acquisitions are completed. Each waveform subsequently can be accessed via the history memory function.
Often, stored waveforms in digital oscilloscopes are retrieved for use as reference waveforms to compare with incoming signals. Now, thanks to the improved signal-processing performance of the new oscilloscope, it's possible to apply the same principle to images consisting of multiple waveforms. In addition, eye patterns displayed in the past can be compared with those currently on display.
The ability to include multiple waveforms in the history memory as reference waveforms means that it's now possible to vertically reposition an image consisting of multiple waveforms in the same way as manipulating reference waveforms in the past. Furthermore, it's also possible to show an average data value out of the values of multiple waveform data.