Analyzing and Troubleshooting Serial Data Buses

When looking at protocol information on a bus, an oscilloscope may not be the first instrument to come to mind�until now.

As most engineers know, digital oscilloscopes are an indispensable instrument for viewing and analyzing electrical signals�the hammer, if you will, in an electrical engineer�s toolbox. But it may not be well known that oscilloscopes, in addition to being ideal instruments for physical-layer waveform analysis, also can be used for protocol analysis of common serial data buses such as Inter-Integrated Circuit (I²C), Serial Peripheral Interface (SPI), and Controller Area Network (CAN).

Serial Data Bus Problems
Serial data bus problems come in many forms. However, most bus problems are classified as noise, signal quality, timing, or data errors. Signal noise on a data bus can lead to a number of problems and typically is more problematic on high-speed buses and densely populated circuit boards as signal voltage levels decrease.

Since these are all trends occurring in today�s electronics, noise is a common problem in bus designs. Common noise sources include crosstalk from other signals, EMI, and switching noise from a power supply.

Signal quality is a general term referring to the overall behavior of a signal relative to ideal conditions. Waveform quality problems come in many forms, but some of the most common are signal levels that are too high or too low, overshoot and undershoot, and rise/fall times that are too slow or too fast.

An example of a waveform quality problem is a data signal that has a rise time that is too slow. At the time of the clock pulse, when the data is to be read, if the data signal has not yet risen to a sufficient voltage level, the data is read incorrectly as a 0 instead of a 1.

Signal timing on a bus is another common source of errors and frustration. In the most basic sense, timing problems refer to events on one or more waveforms not occurring at the desired time. For example, a bus specification may refer to data packets occurring no more frequently than every 400 �s. If two data packets occur less than 400 �s apart, the processor may not have sufficient time to process the data. Other timing errors include setup-and-hold violations.

Data errors simply refer to incorrect data flowing on the bus and can be a result of a software problem, a damaged IC, or any number of other hardware problems. Even if all other errors are under control but data errors exist, then there will be problems.

A Variety of Instruments for Bus Analysis
In general, different types of test equipment are better suited for detecting and analyzing the different errors. Depending on what needs to be looked at, what types of equipment are available, and your level of comfort with the different types of equipment, some combination of oscilloscopes, logic analyzers, or protocol analyzers likely will be used to troubleshoot these bus problems.

While logic analyzers, dedicated bus-specific protocol analyzers, and digital oscilloscopes offer varying degrees of insight, no single piece of test equipment is perfect for identifying or analyzing all these different problems. A logic analyzer, while ideal for identifying complex timing issues, doesn�t offer much help in the way of seeing noise or waveform quality problems. Similarly, a dedicated protocol analyzer is perfect for identifying exactly what information is flowing on the bus and when it occurs, but it typically will not help identify a noise or waveform quality issue.

A basic digital oscilloscope, on the other hand, certainly is more versatile than a logic analyzer or protocol analyzer. While scopes are excellent tools for identifying noise, waveform quality problems, and even many timing problems, traditional scopes don�t provide insight into bus protocol. By adding serial data bus triggering and protocol analysis, some of today�s new oscilloscopes take a big step toward being the right tool for serial bus analysis.

Key Oscilloscope Specifications
Before going into more detail about the serial bus-specific triggering and analysis features found on some of today�s scopes, let�s review the basic scope requirements for accurately capturing serial data bus signals. The three main considerations when selecting the right scope for the job are bandwidth, sampling rate, and memory.

The analog bandwidth of an oscilloscope is almost universally defined as the frequency where the manufacturer guarantees that a signal component of that frequency will not be attenuated by more than 3 dB. Ideally, the frequency response of the scope will be nearly flat and minimally attenuate signals less than the bandwidth of the scope. Typically, the frequency response rolls off quickly above the specified bandwidth and highly attenuates higher frequency signal content.

The obvious question regarding bandwidth is �How much is enough?�. There is no exact answer, and engineering judgment must be used. The common rule of thumb says that a scope with a bandwidth three to five times the highest frequency component of the signal of interest will suffice, which means the amplitude error will be less than ~5%.

However, the frequency component of interest usually is not the fundamental frequency of the signal. Let�s say your serial bus has a data transfer rate of 1 Mb/s. Assuming a continuous series of 0101�, the fundamental frequency of the signal is 500 kHz. But remember, this is a square wave with frequency content extending out indefinitely in the frequency domain.

The result of the scope�s limited bandwidth attenuating the higher frequency harmonics of the input signal is a rounding off of the true square wave. The waveform displayed by the scope appears to have slower rising and falling edges than those of the actual input signal.

Here is where the engineering judgment comes in. Finding a scope with sufficient bandwidth typically is not a problem for buses with slower data rates. However, for buses with data rates of hundreds of megabits per second and higher, it can be difficult and expensive to find a scope with enough bandwidth.

Also keep in mind that an oscilloscope�s bandwidth limits the amount of noise displayed on a signal. For example, let�s look again at the 1-Mb/s signal and assume that the signal has some noise. A scope with a bandwidth of 100 MHz usually will display less noise than a 1-GHz scope.

While it may be advantageous at times during the design stage to use a lower bandwidth scope to reduce the amount of displayed higher frequency noise and isolate the lower frequency main waveform, don�t fool yourself into believing the noise really isn�t there. If you test your product on a 100-MHz scope and ship it to your customer who looks at it using a 1-GHz scope, there may be significant amounts of unacceptable noise.

An oscilloscope�s sampling rate also affects how accurately an input waveform is captured. The sampling rate refers to how fast the analog-to-digital converter (ADC) samples the incoming signal. It is important to note that a scope�s maximum sampling rate does not tell the entire story. Since a scope�s sampling rate is related to its memory length (available memory = sampling rate � the recording time), memory length also must be taken into consideration when selecting a scope.

Let�s compare two scopes. Scope #1 has a maximum sampling rate of 5 GS/s and 10 kpoints of memory per channel. Scope #2 has a maximum sampling rate of 200 MS/s but 1 Mpoint of memory per channel. Let�s say you want to record 100 ms of bus data. Scope #1 only samples at 100 kS/s because of its limited memory. Scope #2, even though it has a lower maximum sampling rate, can sample at 10 MS/s or 100 times faster than Scope #1.

So what is a sufficient sampling rate? Similar to bandwidth, there is no exact answer but rather a rule of thumb. For the best results with binary signals, try to sample at least 10 times faster than the highest frequency signal component you are interested in.

When considering the necessary sampling rate for square waves, the 9th harmonic of the fundamental frequency usually can be used as the highest frequency component of interest. If the signal you are capturing does not have fast transitions, you probably can sample at a lower rate and still get a good representation of the signal. Three to five times the fundamental frequency of the signal should suffice in this case.

Basic Scope Features for Bus Analysis
In addition to sufficient scope bandwidth, sampling rate, and memory, all of which are necessary for accurate waveform capture, many scopes offer a number of basic features that can be used to analyze and troubleshoot serial data buses. One of these common features is the capability of the scope to calculate a number of waveform parameters. For example, the scope could calculate and display the rise or fall times of a waveform, overshoot and undershoot, the max and min values, or the rms value of the waveform.

Other common scope functions useful for bus analysis are zoom windows. Most scopes today have at least one zoom window; some scopes have more.

When a scope captures a long string of serial data, zoom windows are useful for looking at and studying subsets of the total data in greater detail. For example, a scope with two zoom windows could simultaneously view two different data packets or maybe isolate both a rising and a falling edge of a problem signal.

Cursors also are useful tools when investigating waveform behavior. Vertical cursors can look at timing relationships between events on a single waveform or between events on multiple waveforms. Horizontal cursors are helpful when looking at max and min values, overshoot, and other voltage-level-related items.

Some scopes also include general-purpose serial bus cursors, which save you from the cumbersome task of counting lots of 1s and 0s. By simply selecting the bit rate of the data, the voltage threshold level, and the start position, the serial cursor function displays the binary data in hex or binary format (Figure 1).