Using a common digital storage oscilloscope to troubleshoot digital faults in embedded systems

Although today's servers and PCs increasingly favor advanced high-speed processors, low-end 20 or 30MHz processors still play a role in real life. Although these clunky microprocessors are old and have been around for many years, they still have a place in machinery, consumer electronics and automotive appliances.

What do these embedded processors and their applications have in common? In summary, there are several obvious characteristics: First, these processors have been fully verified by many products, people are very familiar with them, and their development can be widely supported and easy to design; second, compared with existing high-end solutions, their clock speed is relatively slow and the bus speed is also slow; third, the application system (from vending machines to avionics) must have high reliability; finally, the cost (including design, manufacturing and maintenance costs) must be as low as possible.

Another feature worth noting is that there is a steady trend toward increasing clock rates for these embedded devices and buses, not to keep up with the fastest servers, but toward "short-clock-cycle" devices, with clock rates 5 to 6 times faster than before. Compared to past processors, new devices have the same pins and functionality, but can do more work in a given time, executing more instruction cycles to complete more complex tasks without slowing down the entire system. This is especially beneficial to software developers, because time-consuming code optimization is no longer important, and new products will be brought to market faster and cheaper.

The bandwidth of ordinary oscilloscopes for basic digital inspection has doubled to 200MHz, and some very useful "high-end" measurement features such as advanced triggering, fast Fourier transform (FFT) analysis and color display have been added to low-end instruments. Today's designers can also use digital fault detection solutions when facing embedded processors in civilian products.

Bandwidth determines application

The processors produced not long ago have a hidden "performance" compared to the same devices produced ten years ago, that is, the signal edge transition speed is faster. From the CMOS process used to produce these products 15 years ago to the fast 5V process developed five years ago, the edge transition speed has increased by about 3 times. Many new designs use this fastest 5V process, and some even further reduce the voltage of the core part and use 5V only on the periphery. For the latter, faster clock rates can be achieved. This speed increase is a byproduct of the reduction in silicon feature size.

Faster edges are generally a good thing, reducing latency, setup time, and contention issues within the system, but shorter propagation delays (mostly due to the faster edge rates of CMOS) can also have a negative impact. As these delays get shorter, address decoding margins, which are usually dependent on delays between address line logic and bus control lines, run into more trouble. Therefore, designers need to know and understand these edge conditions, the increasingly narrow transients, and other pulse characteristics that can occur with high-speed transitions.

When selecting a DSO for a digital design with a 20MHz embedded processor, one might think that a 50MHz or 100MHz bandwidth instrument would be more than adequate for the job. While this is true for some basic troubleshooting, such as whether a signal is present or whether timing and synchronization are correct, other details may not be so obvious.

A DSO with a higher bandwidth can provide more insight into signal characteristics than a lower bandwidth instrument, because the oscilloscope rise time becomes one of the factors that determine the quality of the observed signal. The formula is as follows:
Measured rise time = √ (oscilloscope rise time)2 + (signal rise time)2

A pulse that appears "correct" when viewed at a low bandwidth may have an amplitude deviation on the leading edge, making it appear as two pulses; or a very narrow transient signal on the bus output may go completely unnoticed, causing instability at the input of a subsequent device. As shown in the above formula, a 200MHz DSO can capture details that a 100MHz instrument cannot see.

The benefits of a DSO's wide bandwidth extend beyond observing signal edges. Ground bounce, noise, crosstalk, and many other deviations are easier to see and harder to ignore when using a high-bandwidth instrument. The higher the bandwidth, the more accurate the signal reproduction. Figure 1 shows the different views of the same signal on a 60MHz and 200MHz bandwidth oscilloscope.

Detecting Timing Problems with Conditional Triggers

In digital storage oscilloscopes, trigger condition selection is an important but sometimes less known labor-saving tool that allows the DSO to trigger to meet the specified conditions. Like displaying waveforms, conditional triggering is a basic tool for embedded system detection. Many people use noise suppression (usually increasing trigger hysteresis) to limit short pulses and use various bandwidth limits to select the desired signal. [page]

One of the most common trigger characteristics

Pulse width triggering has recently migrated from high-end lab instruments to ordinary DSOs. This setting causes the oscilloscope to trigger when the input signal pulse width is:
Less than a specified time Greater
than a specified time Equal to a specified time (within nominal error)
Not equal to a specified time (within nominal error)
"Less than" pulse width triggering is one of the fastest ways to find suspicious transient pulses at the output of a bus or device. Intermittent problems occur when brief transient pulses caused by crosstalk or timing gate the output enable or chip select input of a device, causing the device to send data to the bus at the wrong time, resulting in unpredictable results. "Less than" triggering detects pulses shorter than the user-specified width, allowing the oscilloscope to capture all signals present at the probe input. This method not only captures the transient phenomenon itself, but also the results generated by the output enable and data bus.

The "greater than" trigger helps to find some "stuck" data or other signals that have not returned to the default state after processing. It triggers the oscilloscope when the falling edge of the pulse does not occur at the specified time. For example, a data bus output signal switches to "1" in response to an output enable action, and then does not switch to a new state. This may be caused by a variety of reasons, such as the output enable signal itself is inaccurate, the three-state conversion time of the driven device is too long, or the next value of the data bus does not appear. The "greater than" trigger can find this error and reproduce the signals that will affect all the channels connected to the oscilloscope. After some inspection, it can be found out what caused the problem. The time range here is the same as other pulse width trigger settings, from tens of nanoseconds to several seconds, which provides sufficient time to ensure that the measurement is actually the "stuck" signal without delaying the signal.

"Equal" triggering provides an alternative to voltage threshold triggering when the trigger signal (such as output enable) is corrupted by transient signals or noise, causing the oscilloscope to trigger falsely. This situation can be illustrated with a basic embedded microprocessor. Most of these devices include an external bus that allows the processor to expand the internal memory or peripheral interface, and usually the bus allows the peripheral circuit to control the timing of data transfer to and from the processor. The processor first gives an address, then sends an address select, and the selected peripheral circuit finally sends a "received" signal to confirm that the processor's instruction has been received. The clock delay required for this process is known (usually user-specified), and the specific circuit varies.

Knowing this delay time is the key to distinguishing each peripheral circuit and checking the response activity of the test point. The method is very simple. Use the address strobe as a trigger and set the pulse width trigger time to be equal to the specific peripheral clock delay number. The rising edge of the address strobe starts the countdown. The trigger circuit waits for a preset time, and then the oscilloscope will trigger and detect the signal situation on the test point. By definition, this is the time when the peripheral is active on the bus, so the "equal" pulse width trigger allows the oscilloscope to take on the work of a logic analyzer to a certain extent.

Frequency measurement with built-in counter

Automatic frequency measurement has been a feature of DSOs almost since their inception, and is generally obtained by examining the first cycle of an acquired waveform. This is a useful tool for measuring one-off events, but it does not provide a high-precision average frequency value for a continuous waveform.

Another method of measuring frequency is that used by common frequency counters, which are widely available and usually inexpensive. This method can also be implemented in DSOs using the trigger signal as the source signal for the average frequency reading, which is a new feature of current low-end DSOs. Frequency counters are measured in a variety of different ways. The most common and simplest methods are fixed frequency counters, which count the number of input cycles in a fixed time (displaying the count); or fixed period counters, which count the number of times a cycle takes (displaying the inverse of the count). Both methods are very accurate when the number of counts is large, but the accuracy is poor when the number of counts is small. A similar method is to divide the measurement interval in half, count the time and the number of excitations in the first half, and terminate the measurement when the input transitions (with the same polarity as when the measurement started) once half of the points have been reached. This method does not achieve the highest accuracy in extreme cases, but is generally accurate to about half of the highest accuracy, providing a stable and easy-to-read frequency display (accurate to 6 digits) for valid trigger events. Since any event (within a reasonable range of magnitude) can essentially serve as a trigger event, the "reading" here is actually a general frequency counter.

When troubleshooting embedded systems, it is often necessary to check the frequencies of various local clock signals, including the master crystal. The oscilloscope trigger counter provides a fast internal solution that is more accurate than automatic frequency measurements based on the waveform, and eliminates the need for a separate instrument for frequency counting.

The counter also helps to find the source of crosstalk and noise. For example, if the counter finds that the frequency of a noise signal on the bus is 100kHz, then there may be a problem with the crosstalk or grounding of the switching power supply; similarly, if the frequency of the noise signal is 1/2 of the main clock, the problem may be the crosstalk from the adjacent bus. Because the signal source can actually be any trigger signal, the counter can measure the frequency of any conditional trigger event, not just a voltage trigger that occurs once per cycle. For example, by combining the counter with a pulse width trigger, the frequency of a specific pulse width occurring within a continuous pulse can also be determined.

Color waveform display

Color LCDs used to be available only in high-end lab instruments, but now they can be found in some common DSOs. Color adds an extra layer of information to the display, making it easier to test with this instrument than before.
A waveform is just a line on the screen, so what benefits can be gained by representing it in color? The main benefit is that color can be seen more clearly when observing multiple waveform lines, each line using a different color. This color coding method is also used on the front panel of the oscilloscope, for example, the yellow knob controls the yellow wave, which is connected through the yellow probe. It can also be extended all the way to the probe and even to the circuit under test, marking the test points with different colors. In addition, color is also useful when superimposing two waveforms to compare and distinguish them, and some colors are clearer under dimming lighting conditions.
Color oscilloscopes can improve productivity. Simply put, they are easy to use and can reduce many small human errors, which often take hours to solve.