Digital storage oscilloscopes (DSOs) have become increasingly capable of a larger and more sophisticated set of measurements as vendors have added to their capabilities—typically at the request of end users. Many scope users now say that they only use about 10% of what their scope can do. They don’t want to read the whole manual, but they’re sure they could be more productive if they knew more about using the scope.
This article discusses several ways to improve the everyday use of a digital oscilloscope, all of which are applicable to a wide variety of use cases. The discussed techniques are easy to use and allow for better capture and viewing of signals (on all types of digital scopes) as well as improved measurements and analysis of circuit behavior.
Table Of Contents
- The Basics—What Does A Digital Scope Do?
- Using The Full Range Of The ADC
- Using Persistence Mode And Exclusion Trigger
- Maintain A High Sampling Rate
- Using Signal And Parameter Histograms
- Summary
- References
The Basics—What Does A Digital Scope Do?
All DSOs comprise similar building blocks. There is a front-end set of amplifiers/attenuators. Users can choose from among several amplifier/attenuator combinations in the front end of the scope by changing the volts/division knob’s setting. The amplifier is a crucial element in the scope’s bandwidth rating. The vendor guarantees that signal energy up to the rated bandwidth will pass through the amplifier (and the rest of the signal path) with an attenuation of less than 3 dB (about 30%).
After the amplifier stage, there is an analog-to-digital converter (ADC). The amplifier passes a continuous voltage-versus-time waveform to the ADC, which then samples the voltage level and stores a set of voltage measurements in the instrument’s acquisition memory. This happens “live” during signal acquisition. The sample clock controls the time between ADC samples.
Faster is better. A faster sampling rate captures more detail in the shape of the signal. Most scopes have an 8-bit ADC (though new, more accurate high-resolution oscilloscopes are now appearing), so each measurement of the voltage is made with 8-bit resolution. In terms of scope specifications, longer memory is better. A DSO with a longer memory can maintain a high sampling rate for a longer signal length than one that has a shorter memory.
Using The Full Range Of The ADC
Since most engineers’ first action is to capture a signal and view it, improvements in this process can have beneficial effects for just about every scope user. Let’s look at what is perhaps the most common mistake (Fig. 1). The overwhelming majority of scope users have four-channel scopes—and they often want to look at four simultaneous signals. To see each signal clearly, the user sets the volts/division knob and offset so each signal occupies one quarter of the vertical range of the grid.
1. In this screenshot, a scope displays four signals on a single grid. Each signal is only using 64 counts of its ADC, which amounts to 6-bit resolution. For example, the lower signal, channel 4, is using counts 0-63. Although each signal is only 2 V in amplitude, the scale is set to 1 V/division, which means the full scale of each ADC covers 8 V.
Though the display is nice and simple, the drawback of this technique is that each signal is only using one-fourth of the range of its ADC. The user gets 6-bit resolution (64 counts) instead of 8 bits (256 counts). Six-bit ADCs are very inexpensive. A few years ago, some scopes with 6-bit ADCs were offered for sale. They were not very popular because users discovered that such scopes were not very accurate. However, many scope users today are paying for 8-bit scopes and then using only 64 counts of the ADCs.
Now we will consider a better technique for capturing and viewing four signals (Fig. 2). Here, we display each signal on its own grid. The signal occupies a large part of the range of the ADC. If you did not look closely at the display, Figures 1 and 2 look very similar. However, the quality of the data is very different. Any type of measurement made using the “6-bit technique” will be much less accurate.
2. Here, we capture four signals and view them using four separate grids. Each grid corresponds to the full scale of the ADC. At 250 mV/division, each 2-V signal occupies all eight divisions of each grid. Thus, the user makes a more accurate capture of the signals using the full resolution available.
Of course, not all scopes offer quad-display grids. Most LeCroy scopes include the choice of single, dual, quad, or even octal display (in case the user wants to view that many signals and zoom/math traces). Some LeCroy scopes and many from other vendors have only a single or dual grid, though. There’s still a way to make sure you use the ADC’s full scale.
Set up each signal to use the full range of the grid. You can get a good clean view of each signal one at a time by toggling off the view of the other signals. If you want to view several signals simultaneously, you can turn on several traces. The screen may be a bit messy because all the signals are full scale, but you can see things such as relative timing of edges and signal shapes.
Keep in mind that you do not have to view signals to characterize them using the parameter measurements of the oscilloscope. Parameters are calculated using the data in the memory, not by using the pixels on the screen.
Using Persistence Mode And Exclusion Trigger
Most digital scopes offer a persistence display mode for the capture and display of signals. With persistence display selected, the oscilloscope will trigger, place the trace of the signal on the display, and then trigger again and add another trace to the display.
The traces “persist” on the display for a time constant that can be infinite (all traces stay on the screen until a “clear sweeps” command is given), or there can be a shorter time constant after which the older acquired signals fade out. Persistence mode can be a good way to acquire and view large amounts of data.
Typically, we use persistence mode to look for rare intermittent events. It can be easy to turn on the persistence accumulation of data and then look at this visual mass of data in search of unusually shaped traces. Typical use of the persistence display requires a stable trigger point and a reasonably repetitive signal. Otherwise, the display just becomes a jumble of traces. Nevertheless, if you have a test signal that should normally be repetitive, and there are occasional glitches or other problems, then the persistence display might reveal them.
Whether or not you can successfully use persistence mode to see an indication of a signal fault, typically the next step would be to set up a trigger that allows the user to more clearly capture the intermittent fault and troubleshoot it. An intermediate step can often be productive. Some scopes allow the user to set up a trigger condition to prevent the scope from triggering on normally shaped signals and to trigger only on abnormal ones.
For example, if you’re troubleshooting an intermittent problem on a clock, you may have no idea what sort of defect is occurring in the signal, but you know the timing of the period and pulse width of properly shaped clock signals. There is no useful troubleshooting information for the user when the scope repeatedly triggers and displays the normal shape.
What’s necessary is to see the abnormal shape. In such a case, the trigger can be set to prevent the scope from triggering on the (very frequent) occurrence of the normal signal and thereby enhance the probability of being able to capture and view the abnormalities. This is called “exclusion” triggering. The scope is excluded from triggering on the normally shaped signal.
We show an example of a persistence display used in combination with exclusion triggering (Fig. 3). In this example, the user knows that the correct clock pulses are 325 µs wide. The scope is set to capture and display only those pulses that deviate from 325 µs by more than a “delta” of 15 µs.
3. By using exclusion triggering, the scope is prevented from triggering on the normal signal shape. The trigger is set to capture only pulses with widths different by at least 15 µs from the typical 325 µs.
This technique can be very important when troubleshooting intermittent signals. Every scope has some dead time after each trigger. During that time, the data is stored and the scope is re-armed for the next event. Even with a very fast triggering rate, a scope has limited “live” time for signal capture.
Suppose an oscilloscope can trigger 400,000 times per second on a timebase of 40 ns/division. That means the scope will capture 400 ns of signal (since there are 10 divisions on the screen for each trigger) 400,000 times per second. The total “live” time of the scope per second is 400 ns X 400,000 = 0.16 seconds. This is excellent performance but it means the scope is “live” 16% of the time and “dead” 84% of the time.
If you’re looking for an intermittent glitch that occurs many times per second, you’re likely to see it quickly. But if you’re hunting for an intermittent glitch that occurs rarely, there’s an 84% chance you’ll miss the important signal because the scope was “dead” during the re-arm time from capturing the normal signal.
It’s more effective in such cases to keep the trigger circuit armed and looking for the intermittent event. Triggering often (on the common signal) is not as important as triggering on the (rare) signal that you want. Exclusion triggers can be set to look for uncommon signal period, pulse width, interval between edges, and various other known signal qualities—even when the nature of the anomalous signal behavior is unknown.
Maintain A High Sampling Rate
One of the most important aspects of making accurate measurements with a digital scope is ensuring the instrument maintains a high sampling rate. A high sampling rate enhances the accuracy of most measurements. Conversely, a decreased sampling rate can adversely affect accuracy.
In the worst case, some signal components can be “aliased.” This manifests itself as corruption of the true signal shape by the addition of “fake” signal components resulting from undersampling of real signal components.
Nyquist’s theorem tells us that the sampling rate of a digital scope must be at least twice the speed of the highest-frequency content in the signal. Any signal energy at frequencies higher than the Nyquist rate will be undersampled and aliased.
In general, the sampling rate of a scope may change with the number of channels in use and the time/division setting. At long signal-capture times, the sampling rate is reduced so the acquisition memory can cover the elapsed signal time. At lower sampling rates, a high-bandwidth front-end amplifier is a disadvantage because it will send lots of high-frequency content in the signal to the ADC, where those frequencies will be aliased.
Three scopes in your lab might all have the same specified bandwidth and sampling rate. But when capturing longer signals, the sampling rates could be different by a large factor if the oscilloscope memory lengths are different. If you’re capturing signals of long duration, such as those from power devices or serial data streams, make sure you’re aware of your actual sampling rate, which may be very different from the maximum sampling rate quoted on the instrument’s datasheet.
Also, be sure to enter the “horizontal” menu and select the maximum amount of acquisition memory. Some oscilloscopes additionally allow the user to specify whether all four channels are in use or only two (or one). If only two are in use, some scopes can take the memory from the unused channels and add it to the active channels, doubling the amount of time it captures at a high sampling rate.
Using Signal And Parameter Histograms
Many digital scopes allow the user to select a horizontal or vertical “slice” of the data accumulated in a persistence display and draw a histogram (or bar chart) that shows the distribution of the data inside the selected area. A horizontal slice shows the distribution of the time of arrival of signal edges (Fig. 4). A vertical slice shows the distribution of vertical noise on a signal. In Figure 4, the distribution is nearly flat.
A scope user should think very carefully when seeing such a distribution. Usually, one would expect the edge to arrive at a certain time with some amount of jitter caused by random noise. This would typically provide a Gaussian-shaped histogram. In Figure 4, something very different is happening. Some other process is affecting the arrival time of the edge. Perhaps a race condition is causing the edge arrival time to be equally likely at all times within a certain bound.
The most common measurements made with digital oscilloscopes are those that characterize key signal parameters. Often, an engineer must ensure that these signal characteristics fall within specifications or troubleshoot the cause when the parameters have too much variation from the desired value. It can be very useful to examine the distribution of values for a parameter to understand the underlying mechanism that is causing the parameter to vary.
A screen capture shows the persistence display of a signal in the top trace and the histogram of parameter 1 in the lower trace (Fig. 5). The parameter measures the delay between the arrival time of the edge in the center of the screen and a previous edge, which occurred about 2 µs earlier. A histogram distribution contains more information and can give more insight than simply looking at the “statistics” (average value, maximum, minimum, and standard deviation) of a parameter.
In Figure 5, it’s likely that the time between the trigger and the edge on the screen has two competing processes, causing the bimodal distribution. But if users only look at the parameter statistics without seeing the histogram, they would not have this important insight. There are many possible histogram shapes. Each one provides a different type of insight into the underlying cause of variation in key signal parameters.
The most powerful scopes can measure the signal characteristics of every period (or rise time or pulse width, and so on) in a captured signal—millions if necessary—and update histograms of up to eight parameters simultaneously. This requires a processing engine that can crunch large arrays of numbers quickly.
In general, newer scopes with more powerful processors will have an advantage in this type of measurement. But some scopes, even though they’re new, will only measure one parameter value per trigger (rather than millions) and thereby accumulate information much more slowly. Additionally, many oscilloscopes can only display the distribution of timing parameters. The best scopes will also show the distribution of RMS voltage, peak-to-peak voltage, and other amplitude-related parameters.
Summary
There are many tools in the modern digital oscilloscope. Essentially, once you capture the wave shape of a signal in detail, everything else is math. Yet you can make important improvements, even in simply achieving the most accurate view of a signal (by making sure you use the full range of the ADC and all of the acquisition memory).
Other tools can improve your confidence when troubleshooting intermittent faults. Among them is exclusion triggering, which improves the odds that the scope trigger will be “live” when a rare signal abnormality occurs. The best scopes make it very easy to quickly obtain highly accurate signal views and/or measurements. Getting a view or measurement of a signal quickly and easily is not helpful if the view or measurement does not contain the information you need—or even worse if it gives you misleading information.
Using the types of improved techniques and tools discussed in this article can help you make measurements that are more accurate, get the full value of what you bought in your scope, and help you to get better products to market faster. It may also help you sleep a bit better at night, with a higher level of confidence in the results of characterizing the performance of a circuit.