Palo Alto, Calif.
Not long ago, power electronic circuits were relatively simple. They generally consisted of little more than discrete solid-state components such as diodes, thyristors, and power transistors. Nowadays, digital signal processors (DSPs) and microcontrollers play an increasingly important part in power management. As a result, it is not uncommon to see power electronic circuits incorporating numerous analog and digital signals with widely varying operating frequencies. This makes the design and troubleshooting of these mixed-signal circuits difficult.
Traditionally, voltmeters, logic analyzers, and oscilloscopes were the staples of testing and troubleshooting. With mixed signal designs becoming more prevalent, the average engineer's bench is increasingly stacked with one or more of these instruments. However, a mixed-signal oscilloscope, or MSO, can simplify the debugging of mixed-signal designs by combining the best of the analog and digital worlds into one package. It combines the high channel count and triggering power of a logic analyzer with the high resolution of an oscilloscope in a single instrument with a time-correlated display. This lets the developer view both analog and digital signals on the same screen and time axis.
Persistence of memory
An MSO makes it possible to simultaneously detect signals with enough time resolution to resolve analog signals in the dc to kilohertz range and digital signals in the megahertz range. When analyzing such mixed-signal systems, it is essential to have deep memory.
Memory depth is a measure of how many samples the scope can store. It is calculated by multiplying the length of the total capture time by the sample rate needed to re-produce the signal accurately.
Memory depth and sampling rate are intimately related. The memory depth needed depends on the overall time span that has to be measured, and the time resolution required. For instance, it may take 1 sec for a motor in a positioning system to accelerate from rest to some final velocity. The pulse-width-modulator (PWM) signal may be 10 kHz, meaning that in 1 sec there will be 10,000 pulses generated to accelerate the motor. To view all 10,000 pulses with adequate resolution, a good rule of thumb for sampling frequency is to use 10 times the highest frequency being captured. In this example, a sampling frequency of 100 kHz means that 100,000 data points will be captured. Consequently, at least 100 kbytes of memory must be available.
Keeping it real
At first glance, the multichannel MSO may look as though it operates like a typical digital scope. However, it has several qualities that eliminate the disadvantages associated with conventional digital scopes such as low memory capacity and slow response times. These include 2 Mbytes of memory per channel, a high-resolution display, and an assortment of triggering functions.
The 2 Mbytes/channel gives high-resolution sensing and permits fast searching of long signal sections. For instance, a conventional digital scope may require two separate measurements for capturing different components of a signal; a slow signal section and a fast event with sufficient resolution. To capture the slow signal, a slow sample rate has to be used. However, a faster sample rate is needed to capture the fast event with sufficient resolution. But measuring an event twice may produce different results and something of interest may be lost.
The MSO, on the other hand, only requires one measurement to capture all of the relevant information. With 2 Mbytes of memory per channel, the MSO lets users detect a signal section 10 msec long at a resolution of 5 nsec (10 msec/ 5 nsec = 2,000,000), and rapidly search for details.
A multiprocessor architecture also speeds things up by getting rid of data bottlenecks. Acquisition, waveform storage, and display functions are each controlled by their own microprocessor. A custom ASIC then does the mathematical calculations for storing and displaying the data, taking a large part of the processing burden off of the other processors. This ensures that data will move around quickly, providing an uninterrupted flow of data from the tip of the probe to the scope's display.
The multiprocessor architecture and deep memory combine to move large amounts of data in real time so that the waveform on the screen is a true representation of what is happening at that instant in time. It also lets the MSO respond without delay when modifying settings or searching through waveforms using the pan and zoom functions.
The display system is seamlessly integrated into the MSO analog-type user inter-face. This makes it easy to detect anomalies not related to triggering events.
One often-cited drawback to digital scopes is the poor signal integrity compared with analog oscilloscopes. This is due largely to the display of sub-harmonics, or aliasing, the lack of Z-axis information (greater intensity where waveform activity occurs most often), and stepwise approximations of real waveforms.
Over the last two decades, digital scope manufacturers have tried mightily to improve displays. A typical digital scope has a VGA display (640 480), which translates into 500 horizontal points. Currently, there are MSOs with 1,000 horizontal points that provide a higher resolution signal display using 32 levels of intensity. And due to the high screen refresh rate of up to 25 million vectors/sec on each channel, signal variations can be detected immediately.
The combination of deep memory and a high-definition display system reduces the risk of a small interference pulse or a glitch distortion influencing circuit function and remaining undetected. Complex video signals are now displayed to a standard previously experienced only with analog oscilloscopes. But now, they can be frozen and magnified with one rotation, to a resolution of one line.
Trigger, set, go
Versatile triggering functions for isolation and analysis of complex mixed-signal designs are also part of the mix. Particularly where mixed-signal designs are concerned, it is often possible to find the problem only by triggering on the anomaly and correlating it with other signals.
This is where conventional scopes reach their limits. But the MSO offers a wide assortment of triggering functions: glitch, pulse width, and video, among others. All channels can be used and combined for triggering. This makes it simple to isolate complex signals and to analyze interactions in mixed analog and digital circuits.
The MSO is also suitable for debugging microcontroller systems with serial I 2 C (Inter Integrated Circuit) communication. I 2 C trig-gering functions work by capturing only a portion of a signal from the entire serial data stream. Trig-gering on a specific event means the scope will begin to store data after that event. This makes it possible to check not only the inter-IC handshake, but also data transmission between components.
MSOs and power electronics
A modern brushless dc motor contains not only conventional motor components but often sensors, a DSP or microprocessor, and output transistors. A motor of this type has a rapid response and is more reliable than a conventional dc motor. However, in this kind of system, the complex interactions that occur make it much more difficult to locate the causes of faults.
The motor's speed of rotation is typically controlled by varying the duty cycle of the pulse-width-modulated (PWM) motor drive. The rotor position and speed are detected using three Hall-effect sensors, which are part of a closed-loop control system to keep speed constant even during load variations.
Suppose that the motor is being used in a positioning stage and that it takes one second to accelerate from standstill to full speed. During that time, 25,000 pulses of the 25-kHz PWM signal will have taken place.
Viewing the signals with enough resolution is the problem. If the analog signal is the only one of interest, 1,000 samples over a second is probably enough to get a sense of what is going on. But to capture the PWM signals, a different strategy is needed. In 1 sec, the PWM executes 25,000 pulses. Here, 1,000 samples will not be enough resolution. A much higher sampling rate is needed which, in turn, means more memory. So while a higher sampling rate gives more resolution, available memory is just as important.
The scope must have enough memory to capture the 25,000 cycles. The correct amount can be calculated by dividing the longest relevant time interval by the shortest one. In this example, the longest relevant time interval is 1 sec. The shortest time interval comes from the 25-kHz PWM frequency. This corresponds to a period of 40 µsec. However, for adequate resolution, using the ten times rule of thumb gives a highest frequency of 250 kHz, which gives a time interval of 4 sec. Therefore, 250 kBytes of memory is adequate in this case for signal representation with a sufficient level of detail (1 sec/ 4 sec.).
Because of a deep-memory capacity of up to 2 Mbytes per channel, the MSO can display and precisely analyze each individual pulse within a motor acceleration phase with 1-sec duration