The power supply's command processing time, output response and settling time and measurement speed factors into test throughput, and many power supplies can be surprisingly slow. The device chosen may cost less, but it may not be the most cost-effective choice. Carefully benchmarking the device's throughput attributes to weigh against your test time cost could prove to pay back handsomely when selecting your next system DC power supply.
System DC power supply output response speed
Many devices are tested at several different DC bias voltage levels to assure correct performance over their specified operating range. Making several bias voltage changes can add up to seconds and thus become a large portion of the test time. A few tasks have to be carried out when changing a power supply output voltage setting to a new value, as depicted in Figure 1. These steps all take a finite amount of time. Once a command is received by the power supply, the latter must process it: this is its command processing time. The power supply's output then responds and changes to the new setting. The time it takes to reach its final value, within a certain settling band, is its output response time. The difference between various system DC power supplies can be quite dramatic. Table 1 compares command processing and output response time representative of conventional system DC power supplies to one optimised for throughput (in this case an Agilent N6750A series DC power supply module, which belongs to the N6700 Modular Power System family shown in Figure 2). Having a fast output response can reduce the time to change to a new voltage setting by hundreds of milliseconds.
A down programmer impacts output response time
Changing voltage rapidly in both directions is critical to high throughput testing, so it is important to take note of down programming output response time. Many power supplies depend upon the actual loading of the DUT to bring the voltage down. Under light loading conditions, it can take a second for some power supplies without down programmers to reach their final value. A power supply optimised for high throughput incorporates an internal down programmer. This device is a load circuit that quickly discharges the power supply and DUT capacitances for fast down programming, independent of the DUT loading.
Response speed impact on test throughput
An automotive Electronic Control Unit (ECU) is tested at many voltages, as illustrated in Figure 3. Assume there is a 200-millisecond output response time reduction achieved by switching to a fast response power supply. For example, an ECU having 15 voltage changes during its 20-second test, a 3 second test time reduction or 15% improvement in throughput is realised. Such an improvement is easily justified and welcome in the automotive electronics industry.
System DC power supply measurement speed
DC bias current measurements are virtually always made during test as the latter readily identifies defective electronic devices. Oftentimes, several bias current measurements are taken on a device during test, cumulatively creating a large impact on throughput. Measurement accuracy and speed are a trade-off, complicating matters further. There are three steps that can make up measurement time: a measurement command is received and processed by the power supply, the power supply acquires the actual measurement, and the acquired value is returned back to the PC.
Use the built-in measurement system
It makes a lot of sense to rely on the system DC power supply's internal current read-back for carrying out measurements. Implementing external measurements using shunts and DMMs introduces voltage drop and presents accuracy and calibration issues. Relay multiplexers, when used, introduce additional error and switching time delays. Today's system DC power supplies specify current measurement performance ranging from basic to very high accuracy, suiting most applications. Agilent N6700 DC power supply modules come in different performance levels, for example, because users' needs can vary greatly. Most general-purpose system DC power supplies employ a conventional read-back circuit like the one depicted in Figure 4. For this conventional approach, the command-processing time is typically the dominant element, adding up to 100 milliseconds of measurement delay. The lapse of time for a fast ADC measurement acquisition and returning a single reading is small, usually no more than 2 milliseconds. The overall time is usually treated as a single benchmark. This basic approach provides reasonable balance of accuracy and throughput for many, but not all, situations. It can suffer from measurement repeatability error, or «jitter", for high crest factor current signals, such as the pulsed current drain from a digital mobile phone. Averaging several measurements helps, but it adds a lot more test time.
Programmable integration provides fast and accurate measurements
System DC power supplies optimised for throughput will have much less measurement command processing time. Several also feature programmable measurement integration time in place of a fixed ADC and low pass filter, shown in Figure 4, for enhanced performance. It can be set to a fast measurement acquisition time comparable to the conventional approach. Alternatively, programmable integration is used to cancel out periodic noise and AC components in the signal, providing a quantum improvement in measurement performance, at the expense of a little more time. A familiar example is integrating over a Power Line Cycle (1 PLC, 16.7 or 20 milliseconds) to cancel out AC line noise. Agilent has taken a somewhat different approach with many of its system power products, including the N6760A series DC power supply modules. They incorporate a programmable sampling period and digital signal processing to make both fast and accurate measurements. When the measurement acquisition integration time is programmable, it needs to be taken into consideration in the overall measurement time benchmark.
Measurement speed impact on test throughput
Table 2 compares measurement command processing and acquisition times representative of conventional system DC power supplies to those of a device optimised for throughput, in this case the Agilent N6760A series DC power supply module. Switching from a conventional system DC power supply to one optimised for measurement throughput can reduce the measurement time from around 100 milliseconds down to as little as just a few milliseconds. For testing devices like an ECU, several current drain measurements are usually taken. A timesaving of 0.5 to 1 second is easily achieved, providing a worthwhile gain in test throughput. Semiconductor device testing is even more demanding. With a test time of just a couple of seconds or less, there is simply no room for 100 millisecond-long measurements.
Summary
System DC power supplies are one of a few instruments that both source and measure, having a greater impact on test throughput than usually anticipated. Their sourcing and measurement speed attributes need to be carefully reviewed and benchmarked to assess impact on test throughput. Switching to a system DC power supply optimised for throughput can reduce test time by several seconds while providing fast and accurate measurements. As test time for many devices nowadays is often as short as just seconds, the resulting throughput improvement is easily justified and most welcome.
