Bench Talk

Imagine you’ve gone through dozens of possible DC-output supplies on the market and picked the one with the required output voltage, maximum current rating, size, form factor, and price. You put it in the system and start some tests and performance is inconsistent—sometimes everything runs fine for a long time, and sometimes the system crashes. Is it the software, noise in the circuitry, or other factors?

It’s a tough problem to track down, but eventually you notice a loose correlation between the motor starting and the system problems: There seems to be some sort of relationship. Putting the scope on the power rail, you see that the DC output of the supply—and thus the system supply rail—dips when the motor turns on. Sometimes the dip is slight, and sometimes it pulls the supply rail below the specified operating voltage for the electronic components. Then it becomes obvious: Your supply can deliver the needed output current, but it is not responding well when there is a transient demand for current.

The dynamic performance of a supply is just as critical, and perhaps more so, than its static performance. The reason is that a supply with poor response will cause all sorts of intermittent and soft failures and crashes, and yet may look fine when basic static measurements are made. As any engineer with debug experience knows, tracing intermittent problems is far more difficult than finding hard failures or identifying a component that does not meet its basic static specifications.

What are the key power-supply parameters? Of course, first comes nominal output voltage and tolerance, such as 3.3V ±5 percent or 3.3V ±0.2V. Also, what’s the temperature coefficient for that nominal rating? Is that voltage only at 25°C, or is it across the entire temperature range, such as -40°C to +85°C?

Next is the maximum output current rating, which is the maximum amount of current the supply can deliver at the designated output voltage. Again, tolerance is an issue, but also derating due to temperature is critical: A supply may only be able to deliver 80 percent or even less of the current at the maximum temperature rating.

Next comes line and load regulation. Line regulation tells you how much the supply output varies as the input voltage varies from nominal value, and applies to both AC/DC and DC/DC supplies. Load regulation indicates by what amount the DC output changes as the load on the supply varies; ideally, there should be zero change, but in most cases there will be a small change—and the supply output will still be within desired window.

But these are static specifications. The real challenge is ensuring consistent output when there is a sudden, large increase in output-current demand (Figure 1). Typically, the output “droops” and then recovers, but a lot of clock cycles and processing can be occurring during the droop period. Since the magnitude of the transient-induced droop is a function of variables—such as current output at the instant the load change began, the location in the AC-line input cycle that it occurred (for AC/DC supplies), temperature, and the magnitude of change in load—the droop is inconsistent and the resultant problems it causes may only appear occasionally, thus causing a classic hard-to-capture, intermittent crash.

Figure 1: Output droops when there is a large, sudden power demand.

Supply vendors provide tables and graphs showing the drop in output voltage that the supply presents as a function of load changes and showing the recovery time. But even a brief droop situation of milliseconds or microseconds below the minimum rail is usually unacceptable for most circuits, even if the load—such as a motor—can tolerate a short period of lower voltage.

Note that most supplies, due to their feedback loop which maintains the output, also have a tendency to overshoot the nominal output voltage as they recover from the transient load. This is generally not a problem for loads such as motors or lamps, but may be for lower-voltage or sensitive components. Power-supply designers have to find the “sweet spot” between a design with fast transient response but with minimal overshoot, as these two factors are in opposition. It’s the classic control-theory dilemma.

What is needed is known as a “stiff” supply, meaning one with enough reserve capacity and a fast closed-loop response so it prevents the output drop induced by the load transient from becoming so large that the supply output drops below the allowed threshold, but that also recovers quickly and smoothly from sudden load increases. Each supply is differently “tuned” for this dynamic situation because there are so many tradeoffs that must be factored into the design.  For this reason, some system engineers use one supply for the electronic circuitry and another for the load, with each optimized for the current-draw and transients of the differing situations.