Triggering Wide-Bandwidth Sampling Oscilloscopes

Correct Triggering Is Essential For Accurate Waveform Displays

Virtually every test scenario for the design, development, and manufacturing of high-speed digital communications systems and components includes some method to observe the transmitter waveforms. A common tool for this is a wide-bandwidth sam­pling oscilloscope.

An oscilloscope displays the amplitude of a signal as a function of time. An important issue in setting up an ‘amplitude versus time’ display is to determine what the time axis of the oscilloscope represents. A fundamental requirement to acquire a signal using a wide-bandwidth sampling oscilloscope, such as the 86100 Infiniium Digital Communica­tions Analyzer (DCA), is to provide a signal that is used to establish a timing reference. This signal is commonly known as a trigger. The time locations of all displayed amplitude information are determined by their position in time relative to the triggering signal. To correctly display and interpret a waveform, it is critical to understand the different trig­gering options and how they affect measurements.

The most common way to view high-speed digital communications signals is in the eye diagram format. The eye diagram provides an easy way to quickly check the overall quali­ty of the signal. In one view, the many bit combinations of data patterns are overlaid in a composite waveform. Because the eye diagram represents a comprehensive view of the signal, it is easy to visually obtain the overall characteristics of the signal such as noise, jitter, distortion and signal strength.

Another possible way to display a digital communications signal is in the pattern display. In this display, the bits of the signal are shown in a sequential format. This is useful to see in great detail the shape and trajectories of the signal, which are often difficult to observe in the composite eye diagram. However, due to a limited time span of the oscil­loscope it is difficult to show a large number of bits with high resolution.

Whether an eye diagram or a pattern waveform is displayed is dependent upon the type of signal that is used to trigger the DCA. To understand why this is true requires a basic understanding of how a widebandwidth sampling oscilloscope operates.

Digital oscilloscope architectures

Two types of oscilloscopes can be used to view very high-speed signals. One is the wide-bandwidth sampling oscilloscope (sometimes referred to as an equivalent time scope) such as the DCA mentioned above, or a real-time sampling oscilloscope (some­times referred to as a single-shot scope). In the simplest analysis, a real-time oscillo­scope can be thought of as an ultra fast analog-to-digital converter. The sampling rate (as of 2005) can be as fast as 40 Gigasamples/second. The bandwidth of this type of oscilloscope is determined in large measure by the sampling rate and is approximately 13 GHz for a 40 Gsa/s architecture. At these sampling rates/bandwidths, signals as fast as 5 Gb/s can be accurately viewed. Higher speed signals can be observed, but with degraded accuracy. A wide-bandwidth sampling (equivalent-time) oscilloscope like the DCA can have a bandwidth in excess of 80 GHz. However, the sampling rate is under 1 Megasamples/second. How can an instrument with a sampling rate several orders of magnitude below that of the real-time oscilloscope have a bandwidth several times high­er? The answer lies in how the instruments are triggered.

As mentioned, the real-time oscilloscope samples data as fast as 40 GSa/s. Thus a large stream of data points are captured in one contiguous record. Something is needed to determine what portion of the data stream gets displayed. This is determined by the trig­ger. Generally, the trigger event is a feature of the data signal itself. The trigger can be as simple as when the input signal level crosses a certain amplitude level to as sophisticat­ed as a complex series of logic events in the data stream. See Figure 3.

With its relatively low sampling rate, the DCA will acquire only one sample point for any triggering event. Also, the sampling architecture does not allow the triggering event to be some feature of the incoming signal. A trigger separate from the signal being ob­served is required to synchronize the sampling process. First, the DCA must be armed and waiting for the trigger event to occur. The armed DCA is triggered when the trigger signal crosses a defined voltage threshold. The trigger establishes the time reference and initiates the first sampled point. The sampling takes place a fixed delay time after the trigger event. The delay value must be controlled with extremely tight tolerance (under 100 fs in high-resolution configurations). This delay term is at least 24 ns. (It can be larger if the DCA timebase is configured to observe the signal at a delay value larger than the minimum instrument setting.) The DCA must rearm before another trigger signal can be accepted. The rearm time is approximately 25 μs after a sample is taken. This is a critical concept to understand and has a significant impact on how signals are recon­structed. Once the DCA has rearmed, a second sample is taken when the next trigger event occurs. Like the first sample, the second sample takes place after the delay time has elapsed. However, to avoid sample two occurring at the identical time (relative to the trigger event) as sample one, a small incremental delay is added to the large delay term. The incremental or sequential delay term is dictated by the time span of the DCA and the number of points (record length) defined to reconstruct a waveform. For example, if the DCA time span is configured to be 1 ns wide (100 ps/division and a 10 division display) and 1001 points are going to be used to construct one waveform, the time between points, which is also the sequential delay term, will be 1 ps. A complete waveform record will be constructed and displayed when 1001 rearm and trigger cycles have taken place.