Pulsed measurements are defined in Part 1, and common pulsed measurement challenges are discussed in Part 2.
By DAVID WYBAN, Keithley Instruments, a Tektronix Company, Solon, Ohio
Performing a DC measurement starts with applying the test signal (typically a DC voltage), then waiting long enough for all the transients in the DUT and the test system to settle out. The measurements themselves are typically performed using a sigma-delta or integrating-type analog-to-digital converter (ADC). The conversion takes place over one or more power line cycles to eliminate noise in the measurements due to ambient power line noise in the test environment. Multiple measurements are often averaged to increase accuracy. It can take 100ms or longer to acquire a single reading using DC measurement techniques.
In contrast, pulsed measurements are fast. The test signal is applied only briefly before the signal is returned to some base level. To fit measurements into these short windows, sigma-delta ADCs are run at sub-power-line interval integration times; sometimes, the even faster successive approximation register (SAR) type ADCs are used. Because of these high speeds, readings from pulsed measurements are noisier than readings returned by DC measurements. However, in on-wafer semiconductor testing, pulse testing techniques are essential to prevent device damage or destruction. Wafers have no heat sinking to pull away heat generated by current flow; if DC currents were used, the heat would increase rapidly until the device was destroyed. Pulse testing allows applying test signals for very short periods, avoiding this heat buildup and damage.
Why use pulsed measurements?
The most common reason for using pulsed measurements is to reduce joule heating (i.e., device self-heating). When a test signal is applied to a DUT, the device consumes power and turns it into heat, increasing the device’s temperature. The longer that power is applied, the hotter the device becomes, which affects its electrical characteristics. If a DUT’s temperature can’t be kept constant, it can’t be characterized accurately. However, with pulsed testing, power is only applied to the DUT briefly, minimizing self-heating. Duty cycles of 1 percent or less are recommended to reduce the average power dissipated by the device over time. Pulsed measurements are designed to minimize the power applied to the device so much that its internal temperature rise is nearly zero, so heating will have little or no effect on the measurements.
Because they minimize joule heating, pulsed measurements are widely used in nanotechnology research, such as when characterizing delicate materials and structures like CNT FETs, semiconductor nanowires, graphene-based devices, molecular- based electronics and MEMs structures. The heat produced with traditional DC measurement techniques could easily alter or destroy them.
To survive high levels of continuous DC power, devices like MOSFETs and IGBTs require packaging with a solid metal backing and even heat-sinking. However, during the early stages of device development, packaging these experimental devices would be much too costly and time consuming, so early testing is performed at the wafer level. Because pulsed testing minimizes the power applied to a device, it allows for complete characterization of these devices on the probe station, reducing the cost of test.
The reduction in joule heating that pulsed testing allows also simplifies the process of characterizing devices at varying temperatures. Semiconductor devices are typically so small that it is impossible
to measure their temperature directly with a probe. With pulsed measurements, however, the self- heating of the device can be made so insignificant that its internal temperature can be assumed to be equal to the surrounding ambient temperature. To characterize the device at a specific temperature, simply change the surrounding ambient temperature with a thermal chamber or temperature-controlled heat sink. Once the device has reached thermal equilibrium at the new ambient temperature, repeat the pulsed measurements to characterize the device at the new temperature.
Pulsed measurements are also useful for extending instruments’ operating boundaries. A growing number of power semiconductor devices are capable of operating at 100A or higher, but building an instrument capable of sourcing this much DC current would be prohibitive. However, when delivering pulse mode power, these high power outputs are only for very short intervals, which can be done by storing the required energy from a smaller power supply within capacitors and delivering it all in one short burst. This allows instruments like the Model 2651A High Power SourceMeter SMU instrument to combine sourcing up to 50A with precision current and voltage measurements.
Pulsed I-V vs. transient measurements
Pulsed measurements come in two forms, pulsed I-V and transient. Pulsed I-V (FIGURE 1) is a technique for gathering DC-like current vs. voltage curves using pulses rather than DC signals. In the pulsed I-V technique, the current and voltage is measured near the end of the flat top of the pulse, before the falling edge. In this technique, the shape of the pulse is extremely important because it determines the quality of the measurement. If the top of the pulse has not settled before this measurement is taken, the resulting reading will be noisy and or incorrect. Sigma-delta or integrating ADCs should be configured to perform their conversion over as much of this flat top as possible to maximize accuracy and reduce measurement noise.
Two techniques can improve the accuracy of pulsed I-V measurements. If the width of the pulse and measurement speed permit, multiple measurements made during the flat portion of the pulse can be averaged together to create a “spot mean” measurement. This technique is commonly employed with instruments that use high speed Summation Approximation Register (SAR) ADCs, which perform conversions quickly, often at rates of 1μs per sample or faster, thereby sacrificing resolution for speed. At these high speeds, many samples can be made during the flat portion of the pulse. Averaging as many samples as possible enhances the resolution of the measurements and reduces noise. Many instruments have averaging filters that can be used to produce a single reading. If even greater accuracy is required, the measurement can be repeated over several pulses and the readings averaged to get a single reading. To obtain valid results using this method, the individual pulsed measurements should be made in quick succession to avoid variations in the readings due to changes in temperature or humidity.
Transient pulsed measurements (FIGURE 2) are performed by sampling the signal at high speed to create a signal vs. time waveform. An oscilloscope is often used for these measurements but they can also be made with traditional DC instruments by running the ADCs at high speed. Some DC instruments even include high-speed SAR type ADCs for performing transient pulsed measurements. Transient measurements are useful for investigating device behaviors like self-heating and charge trapping.
The simplest pulse measurement instrumentation option is a pulse generator to source the pulse combined with an oscilloscope to measure the pulse (FIGURE 3). Voltage measurements can be made by connecting a probe from the scope directly to the DUT; current measurements can be made by connecting a current probe around one of the DUT test leads. If a current probe is unavailable, a precision shunt resistor can be placed in series with the device and the voltage across the shunt measured with a standard probe, then converted to current using a math function in the scope. This simple setup offers a variety of advantages. Pulse generators provide full control over pulse width, pulse period, rise time and fall time. They are capable of pulse widths as narrow as 10 nanoseconds and rise and fall times as short as 2-3 nanoseconds. Oscilloscopes are ideal for transient pulse measurements because of their ability to sample the signal at very high speeds.
Although a simple pulse generator/oscilloscope combination is good for fast transient pulse measurements, it’s not appropriate for all pulse measurement applications. A scope’s measurement resolution is relatively low (8–12 bits). Because scopes are designed to capture waveforms, they’re not well suited for making pulse I-V measurements. Although the built-in pulse measure functions can help with measuring the level of a pulse, this represents only a single point on the I-V curve. Generating a complete curve with this setup would be time consuming, requiring either manual data collection or a lot of programming. Pulse generators are typically limited to outputting 10-20V max with a current delivery capability of only a couple hundred milliamps, which would limit this setup to lower power devices and/or lower power tests. Test setup can also be complex. Getting the desired voltage at the device requires impedance matching with the pulse generator. If a shunt resistor is used to measure current, then the voltage drop across this resistor must be taken into account as well.
Curve tracers were all-in-one instruments designed specifically for I-V characterization of 2- and 3-terminal power semiconductor devices. They featured high current and high voltage supplies for stimulating the device and a configurable voltage/ current source for stimulating the device’s control terminal, a built-in test fixture for making connections, a scope like display for real-time feedback, and a knob for controlling the magnitude of the output. However, Source measure unit (SMU) instruments (FIGURE 4) have now largely taken up the functions they once performed.
SMU instruments combine the source capabilities of a precision power supply with the measurement capabilities of a high accuracy DMM. Although originally designed for making extremely accurate DC measurements, SMU instruments have been enhanced to include pulse measurement capabilities as well. These instruments can source much higher currents in pulse mode than in DC mode. For example, the Keithley Model 2602B SourceMeter SMU instrument can output up to 3A DC and up to 10A pulsed. For applications that require even high currents, the Model 2651A SourceMeter SMU instrument can output up 20A DC or 50A pulsed. If two Model 2651As are configured in parallel, pulse current outputs up to 100A are possible.
SMU instruments can source both voltage and current with high accuracy thanks to an active feedback loop that monitors the output and adjusts it as necessary to achieve the programmed output value. They can even sense voltage remotely, directly at the DUT, using a second set of test leads, ensuring the correct voltage at the device. These instruments measure with high precision as well, with dual 28-bit delta-sigma or integrating-type ADCs. Using these ADCs along with their flexible sourcing engines, SMUs can perform very accurate pulse I-V measurement sweeps to characterize devices. Some, including the Model 2651A, also include two SAR-type ADCs that can sample at 1 mega-sample per second with 18-bit resolution, making them excellent for transient pulse measurements as well.
In addition, some SMU instruments offer excellent low current capability, with ranges as low as 100pA with 100aA resolution. Their wide dynamic range makes SMU instruments an excellent choice for both ON- and OFF-state device characterization. Also, because they combine sourcing and measurement in a single instrument, SMU instruments reduce the number of instruments involved, which not only simplifies triggering and programming but reduces the overall cost of test.
Although SMU instruments are often used for pulse measurements, they don’t operate in the same way as a typical pulse generator. For example, an SMU instrument’s rise and fall times cannot be controlled by the user; they depend on the instrument’s gain and bandwidth of the feedback loop. Because these loops are designed to generate little or no overshoot when stepping the source, the minimum width of the pulses they produce are not as short as those possible from a pulse generator. However, an SMU instrument can produce pulse widths as short as 50–100μs, which minimizes device self-heating.
The terminology used to describe a pulse when using SMU instruments differs slightly from that used with pulse generators. Rather than referring to the output levels in the pulse as amplitude and base or the high level and the low level, with SMU instruments, the high level is referred to as the pulse level and the low level as the bias level. The term bias level originates from the SMU’s roots in DC testing where one terminal of a device might be biased with a fixed level. Pulse width is still used with SMU instruments, but its definition is slightly different. Given that rise and fall times cannot be set directly and vary with the range in use and the load connected to the output, pulse width can’t be accurately defined by Full Width at Half Maximum (FWHM). (refer to the sidebar for more information on FWHM). Instead, for most SMU instruments, pulse width is defined as the time from the start of the rising edge to the start of the falling edge, points chosen because they are under the user’s control.
In other words, the user can set the pulse width by setting the time between when the source is told to go to the pulse level and then told to go back to the bias level.
Pulse measure units (PMUs) combine the capabilities of a pulse generator and a high-resolution oscilloscope, which are sometimes implemented as card-based solutions designed to plug into a test mainframe. Keithley’s Model 4225-PMU, designed for use with the Model 4200 Semiconductor Charac- terization System (FIGURE 5), is one example. It has two independent channels capable of sourcing up to 40V at up to 800mA. Like a standard pulse generator, users can define all parameters of the pulse shape. Pulse widths as narrow as 60ns and rise and fall times as short as 20ns make it well suited for characterizing devices with fast transients. A Segment Arb mode allows outputting multi-level pulse waveforms in separately defined segments, with separate voltage levels and durations for each. Each PMU channel is capable of measuring both current and voltage using two 14-bit 200MS/s ADCs per channel for a total of four ADCs per card. Additionally, all four ADCs are capable of sampling together synchronously at full speed. By combining a pulse generator with scope- like measurement capability in one instrument, a PMU can not only make high-resolution transient pulse measurements but also perform pulse I-V measurement sweeps easily using a spot mean method for enhanced resolution.
EGBERT WOELK, PH.D., is director of marketing at Dow Electronic Materials, North Andover, MA. ROGER LOO, PH.D., is a principal scientist at imec, Leuven, Belgium.