An easy-to-fabricate source measure unit for real-time DC and time-varying characterization of multi-terminal semiconductor devices

Semiconductor devices have occupied an immeasurable position in scientific applications that provide electronic control to a system and reveal the intriguing properties of the semiconducting material. The properties of semiconducting materials are sensitive to surrounding physical and chemical variations, including light illumination, temperature, and chemical or gas agent exposure [1]. Such effect of physical and chemical variations on the semiconductor properties can be well understood in terms of its electronic transport characteristics. Generally, the electrical transport characterization is conducted in direct current (DC) mode by applying a bias voltage across the device and measuring its current flow. Such current-voltage (I–V) characteristics of devices are essential to draw a comprehensive understanding of its junction properties, interface defects, efficiencies, life-time, and other relevant device properties [2–4]. The bipolar or bidirectional I–V characterization technique where the bias voltage is swept from negative to a positive potential and vice-versa provides vital information regarding the nature of the measuring devices such as a resistor, diode, photodiode, and especially resistive memory device [2, 5, 6]. Several types of commercial source measure unit (SMU) are available to measure the DC I–V of various semiconductor devices with high precision. Such an SMU system can source and sink power in all four quadrants for characterizing memory devices, solar cells, or energy harvesting devices. These commercial SMU systems are highly efficient yet expensive and can only be afforded by research institutions or laboratories. Researchers from the interdisciplinary area need a customized and cost-effective solution to demonstrate the proof-of-concept trials for their grown samples. The required essential characterization by almost every lab working with organic/inorganic material is the I–V measurement. Such measurement can also be used to optimize the as-grown sample before fabricating functional devices with it. The microcontroller-based digital platforms can be employed to meet the demand for cost-effective and accurate characterizations. Generally, a microcontroller has an arithmetic logic unit (ALU), a memory, programmable input/output ports, serial communication protocols including inter-integrated circuit (I²C), and serial peripheral interface (SPI). Such a complete system on a chip is an excellent choice for control of peripheral systems and data acquisitions. However, most of the microcontroller system does not possess analog output port and therefore a commercial digital-to-analog converter (DAC) peripheral chip is required. Such a programmable DAC chip is extremely important for biasing the device samples with desired constant voltage with time delay. Later, the measured parameters such as current and voltage are acquired with an analog-to-digital converter (ADC) and numerically processed in ALU before transferring the data to the computer for generating desired plots. Being a digital platform, a microcontroller saves power consumption and reduces the size of instrumentation, which makes it cost-effective. Apart from instrument control, microcontrollers have been rarely employed for device characterizations [7–11], whereas few published reports only focus on the I–V characterization of photovoltaic systems [12–17]. The majority of these reported work is based on the unidirectional I–V characterizations, i.e., the bias is swept only in a positive voltage axis and cannot be employed for non-linear devices such as Zener diodes, photodetector, photovoltaic cell. Recently, Corazza et al demonstrated a bidirectional I–V characterization system with a microcontroller to measure small laboratory-scale photovoltaic devices [12]. However, all these reports only focus on the application of the designed SMU system, but the detailed optimization of DAC and ADC performance is still untouched. Furthermore, all the reported SMU systems are dedicated only to a particular type of device. Therefore, a versatile microcontroller-based SMU system is needed to measure various semiconductor device loads with considering cost-effectiveness, power consumption, and customizability of operation.

In this work, a simple Arduino-based source measure unit (ASMU) system for semiconductor device characterization has been described for the first time, which can source and sink power in all four quadrants. The designed ASMU consists of fully digital source and measure units, which reduces noise and increases the reliability of measurements. Two external DAC modules have been used in differential mode to sweep bias voltage from reverse to forward direction. The resultant bias voltage and device current are measured with two peripheral programmable gain amplifier (PGA) based ADC modules. The DAC and ADC modules are controlled by an Arduino Nano microcontroller. The microcontroller is interfaced with a computer via a universal serial bus (USB) port. The control and data acquisition programs are written in the LabVIEW environment for real-time visualization of the data. It is to be noted that the present work also focuses on the optimization of DAC and ADC performances before developing the main SMU program. Finally, the I–V characterizations are performed for both linear and non-linear devices, including resistors, p-n diodes, and photodiodes. The ASMU system is observed to be measuring devices from 10⁻³ A down to a low current of 10⁻⁶ A range. The acquisition program for current versus time (I-t) measurement is also tested for a photodetector. The ASMU system is tested for a commercial transistor, which shows its capability of measuring three-terminal devices. The entire system runs with a 5 V USB supply and thus consumes significantly less power, making it a cost-effective solution for device characterizations. However, it is to be noted that any small-scale, low-cost, single microcontroller-based SMU system should not be expected to deliver high precision data compared to any commercial SMU system. The reported work can be helpful to motivate researchers to build customized SMU in labs indigenously.

The schematic diagram of the ASMU system is depicted in figure 1(a), where the main microcontroller employed is Arduino Nano. The Arduino Nano is equipped with an ATmega328P chip which is a high performance, low power, AVR 8-bit microcontroller operating at 5 V with a clock speed of 16 MHz [18].

Zoom In Zoom Out Reset image size

It has eight analog input channels, each with 10-bit ADC, thereby providing multiple analog voltage measurements with a resolution of 4.88 mV (i.e., 5 V/2⁸). When it comes to measuring 10⁻³ to 10⁻⁶ order of voltage and current, the in-built ADC in Arduino fails to perform. Therefore two external 16-bit ADS1115 (Texas Instruments) based ADC breakout modules are used, which are precise, low-power, and have highly stable voltage reference [19]. The prime feature of this ADS1115 is its PGA, which offers a full-scale analog input range from ±256 mV to ±6.144 V with a corresponding resolution of 7.8125 μV to 187.5 μV, respectively. It has four analog input channels that allow four single-ended or two differential input measurements. Since the Arduino Nano cannot source analog voltage, two peripheral MCP4725 (Microchip Technology) based DAC breakout modules are used. The MCP4725 is a single channel, 12-bit buffered voltage output DAC chip with highly precision rail-to-rail analog output voltage sweep (0 V to 5 V) and a typical source/sink current of ∼15 mA at 5 V supply [20]. The DAC resolution is reported to be 1.22 mV (i.e., 5 V/2¹²) over the full-scale range of 5 V supply.

As shown in figure 1(a), the two MCP4725 DAC based voltage sources, namely DAC 1 and DAC 2, are addressed (I²C) as '60' and '61'. The output bias voltages of DAC 1 and DAC 2 are designated as V_B1 and V_B2, respectively. Since the DAC's can deliver only positive (unipolar) voltage to the load, the combination of these two in differential mode can efficiently produce a full-scale bipolar voltage swing from −V_CC to +V_CC. The corresponding LabVIEW program for bias sweep of each DAC is written inside a while loop expressed in equations (1) and (2). The total differential output voltage can be expressed by equation (3).

$$\begin{eqnarray}&&{V}_{B1}=({V}_{Start}+i\,* \,{\rm{\Delta }}{V}_{B})\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{V}_{B2}=({V}_{End}-i\,* \,{\rm{\Delta }}{V}_{B})\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{V}_{T}={V}_{B1}-{V}_{B2}\end{eqnarray} \tag{ 3 }$$

Where V_T is the total differential output voltage, V_Start and V_End are the programmable output voltage limit values; i is the while loop iteration count; ΔV_B is the biasing voltage step size. Let us consider that the bias voltage is to be swept from −1 V to +1 V with a step of 0.1 V. With the increasing loop iteration i from 0 to 10 counts, the equation (1) shows forward voltage sweep of V_B1 from 0 V to +1 V (for ${V}_{Start}$ set to 0 V). At the same time, equation (2) shows the reverse voltage sweep from +1 V to 0 V with the same i increasing from 0 to 10 counts (for ${V}_{End}$ set to 1 V). Hence, according to equation (3), the differential voltage V_T will swing from −1 V to +1 V with increasing loop iteration. The differential bias voltage V_T is applied to the test load resistor (R_L), and the load current (I_L) is calculated by measuring the voltage drop across a small shunt resistor (R_S), which is connected in series with the load as shown in figure 1(a). The load current is mathematically calculated inside the working LabVIEW program, which is expressed by equation (4) as,

$$\begin{eqnarray}&&{I}_{L}=\displaystyle \frac{\left({V}_{RS}-{V}_{B2}\right)}{{R}_{s}}\end{eqnarray} \tag{ 4 }$$

Where (V_RS – V_B2) is the net voltage drop across the R_S shunt resistor.

The two different ADS1115 based ADC modules, namely ADC 1 and ADC 2, are addressed (I²C) as '48' and '49' by connecting the respective 'ADDR' port to 0 V and 5 V. To measure the differential voltage V_T and voltage across R_S, the A2 and A3 ports of both ADC's are programmed to work as differential inputs. All of the I²C serial clock (SCL) and serial data (SDA) are connected to pin A5 and A4 of Arduino Nano. To measure the ±V_T and (V_RS – V_B2) in mV and μV range, the PGA of ADC 1 and ADC 2 is set to ±6.144 V and ±256 mV, respectively. The configuration register settings of ADC 1 and ADC 2 are given in table 1. The differential input channel and PGA settings of each ADC are configured by 14:12 bits and 11:9 bits, respectively. All the DAC and ADC modules are powered with 5 V and 0 V from the microcontroller's on-board power supply connections as shown in figure 1(a). An electronic buzzer is added to the digital pin D4 of the microcontroller and programmed with a 2.7 kHz frequency tone to indicate the initialization of the acquisition. The I²C serial interface communication is maintained at a baud rate of 9600 bps. The technique and driver programs to establish I²C communication between ADS1115, MCP4725, and Arduino microcontroller, are well documented in many YouTube video graphic tutorials and online platforms [21–24].

The actual fabricated ASMU circuit system on Veroboard is shown in figure 1(b) within a dimension of 10 cm × 10 cm. For user flexibility, initially, female headers are soldered on the Veroboard, and all the components are plugged into them. The ASMU is powered and interfaced with a computer via Mini-B USB cable. The bias control with the data acquisition program is written in the LabVIEW environment. The front panel window showing a graphical user interface (GUI) for real-time I–V measurement is depicted in figure 1(c). The LabVIEW programming is carried out in a 64-bit operating system with 4 GB RAM and 1.7 GHz Intel i3-4005 processor. It may be noted that the data acquisition can also be performed by employing other GUI-based programming tools such as MATLAB, Python, Arduino IDE with PLX-DAQ [25, 26], and so on.

3.1. Optimization of ASMU operation

Before testing functional semiconductor devices with the fabricated ASMU system, it is essential to understand the DAC voltage source and ADC measuring system's performance, including their linearity, measurement fluctuation, optimum bias step, and measurement time. The voltage source's linearity is studied by varying one of the input programmed DAC voltage (V_DAC) and measuring the voltage across it by both ADC's. It is to be noted that V_DAC signifies the real numeric data of voltage fed to DAC while programming. Figure 2(a) shows the variation of actual measured DAC output voltage with increasing V_DAC. The DAC output voltages measured by both ADC 1 and ADC 2 are observed to be increasing linearly with the applied V_DAC values. The fitting of both ADC's data shows linear behavior with a slope value of 0.94. The offset error in DAC output voltage is obtained to be 1.2 mV. The deviation of slope from its ideal value (i.e., 1) can be attributed to the cumulative effect of inherent errors in ADC, DAC, and fluctuations in the reference power supply. Such error in the slope value can easily be compensated by employing a simple open-loop calibration technique by software in applications [27]. However, the difference between the two ADC's voltage measurement, i.e., (ADC 1–ADC 2) shows a small deviation which is calculated to be within −0.18 mV to 2 mV for 0 V–5 V bias range as shown in figure 2(a) suggesting an almost identical working of both ADC's. Similarly, the linearity test of fully bidirectional, DAC's differential output V_T is shown in figure 2(b). The inset represents a variation of differential output voltage ±V_T measurement with programmed ±V_DAC, which is observed to be linear.

Zoom In Zoom Out Reset image size

In order to obtain the resolution of bias, the V_T measurements are recorded by varying ±V_DAC with different voltage steps ΔV_B. The average value of slope (V_T/V_DAC) is plotted against ΔV_B, as shown in figure 2(b). The calculated average slope for every bias step is obtained to be 0.93, which shows that the actual output bias range is ±4.65 V (i.e., 5 V × 0.93). The fluctuations in the slope are observed to be decreasing with increasing bias step size. The optimum resolution of bias is found to be 5 mV, which is limited by the single DAC's output resolution of 1.22 mV [20]. In order to obtain the optimum rate of voltage sweep, V_T measurements are recorded by varying ±V_DAC with different sweep delays for a fixed ΔV_B of 5 mV. The average value of slope (V_T/V_DAC) is plotted against sweep delays in figure 2(c). The calculated average slope is obtained to be within 0.91 to 0.95, and fluctuations in the slope are observed to be decreasing with increasing sweep delays. Figure 2(c) shows that the fluctuations in average slopes are significantly high for sweep delay lower than 10 ms. Such fluctuations are attributed to the time taken by ADS1115 for data conversion, which is 7.8 ms by default [19]. Hence, any measurement below this time can lead to inefficient data conversion and noise. Therefore, a minimum delay of 10 ms is essential between voltage sourcing and measurement at each 5 mV sweep. Finally, the I–V characterization of different commercial resistors with the ASMU system is performed which is shown in figure 2(d). The load resistor R_L ranging from 10 Ω to ∼1 M Ω is tested with bias voltage sweeping from −V to +V, keeping R_S = 1 Ω, and the measured current is plotted in log scale. The I–V characteristic for all resistors is symmetric and closely matches its corresponding theoretical value as depicted by the black line plot for each resistor. The upper current limit of ASMU is recorded to be ∼14.6 mA, which is associated with the limit of DAC's source/sink current of ∼15 mA [20]. However, the minimum load current measured by ASMU with R_L = 98 kΩ and R_S = 1 Ω is recorded to be 7.8125 μA. This current value is limited by the resolution of ADC voltage measurement, which is 7.8125 μV [19]. The lower current limit can also be minimized up to 0.781 25 μA by using R_S = 10 Ω for higher load resistance of 0.97 MΩ, as shown in figure 2(d).

3.2. DC and AC characterization of p-n diode with ASMU

Following optimization of the ASMU system, the electrical characterizations of commercial semiconductor devices are carried out. A general-purpose rectifier diode 1N4007 is connected in series with an R_L of 1.1 kΩ (connection as per figure 1(a)) and I–V characterization is shown in figure 3(a) with the current in log scale.

Zoom In Zoom Out Reset image size

It is evident from such a plot that the forward current is much greater than the reverse current showing a p-n junction behavior. The reverse current is observed to be saturating at 7.8125 μA throughout the reverse bias, which is attributed to the lower current limit of ASMU, as shown in figure 2(d). To validate this experimental result, the diode current is also measured with commercially available systems, including digital multimeter (ET-2076A, Minipa) and 14-bit data acquisition (DAQ) system (NI USB-6009, National Instruments) and is compared with reported ASMU system in figure 3(a).

It is observed that the forward current of ASMU closely follows the measurement performed by ET-2076A and NI USB-6009. However, the reverse current obtained by the other two systems is observed to be different, which shows their respective measurement limit. The cut-in voltage of the diode is measured to be 0.42 V. The values of the ideality factor (η) and series resistance (R') of a diode can be calculated by considering both the thermionic emission (TE) model of charge transport and effect of series resistance in the current transport direction, which is expressed as [28],

$$\begin{eqnarray}&&\displaystyle \frac{dV}{d\,\mathrm{ln}\left(I\right)}=IR^{\prime} +\eta \left(\displaystyle \frac{kT}{q}\right)\end{eqnarray} \tag{ 5 }$$

Where V is the bias voltage, I is the diode current, T is the temperature, q is the electronic charge, and k is the Boltzmann constant. The equation (5) is plotted with diode current measured by ASMU in the inset of figure 3(a), and linear fitting to it provides value for R' = 1.05 kΩ, which matches well with the actual load resistance of 1.1 kΩ connected with 1N4007. The value for η = 1.27 is also consistent with that for commercial silicon-based diodes. To measure the waveform response of the rectifier diode, the ASMU system is programmed to generate different time-varying voltage signals, including sinusoidal, triangular, sawtooth, and square wave. The output current waveform response of 1N4007 with R_L = 100 Ω is shown in figure 3(b) with different input AC voltage waveforms at 1 Hz of signal frequency. It is evident from such a plot that the output current follows its input voltage signal for a positive half-cycle and shows reverse saturation current for a negative half-cycle, which depicts the diode's operation as a half-wave rectifier.

3.3. Photodetector characterization with ASMU

The optoelectronic characterizations of the semiconductor device can also be performed with the ASMU system by replacing the R_L with a photodetector in figure 1(a) and customizing the acquisition program. An infrared (IR) light-emitting diode (LED) is employed as a light source, and the emission power is calibrated using an optical power meter (PM100 D, Thorlabs). Figure 4(a) shows I–V characteristics for a germanium photodetector (PHYWE 41736.7E) under different IR illumination powers. It is observed that the photocurrent is constant for reverse bias voltages and increases with an increase in optical power which depicts its photoconductive mode of operation. The dependence of photocurrent on incident power can be fitted by using power law as,

$$\begin{eqnarray}&&{I}_{P}={P}^{\alpha }\end{eqnarray} \tag{ 6 }$$

Where I_P is the photocurrent, P is the optical power, and exponent α. The inset of figure 4(a) shows the variation of I_P as a function of light power, and the fitting gives α = 1.02, which signifies the degree of linearity and defect-free nature of the photodetector.

Zoom In Zoom Out Reset image size

The device is also observed to be working in the 4th quadrant showing its photovoltaic mode of operation with an open-circuit voltage and short-circuit current of 0.19 V and −0.87 mA under 5.93 mW of IR power, respectively. To measure the detector's photoswitching response, a customized LabVIEW program is used where the differential bias voltage is kept constant at −1 V and real-time current values are acquired. Figure 4(b) shows the photoswitching response of the germanium detector for different IR light pulse duration times with 5.93 mW of optical power. It is evident from such a plot that the I_P is switching from low to a high state (in the negative y-axis) when the IR pulse switches OFF and ON. For each IR pulse duration, the highly stable and repetitive ON and OFF state current is measured to be −0.86 mA and 7.8125 μA, respectively. Since the ADC's data conversion time limits the data acquisition time up to 7.8 ms, it is not possible to measure the photoswitching response of the device for IR pulse duration less than 50 ms. The rise and fall time for photodetection is found to be ∼17.6 ms with an IR pulse duration of 500 ms.

3.4. Three-terminal transistor characterization with ASMU

The most significant feature of the ASMU system is its capability of measuring three-terminal devices such as transistors. In this context, the ASMU system is programmed in such a way that the two DAC's work independently to bias both the base-emitter (B-E) and collector-base (C-B) junction simultaneously. The two ADC's are used to measure respective current and junction voltages. An NPN epitaxial silicon transistor (BD433, Fairchild Semiconductor Corporation) [29] is connected to ASMU in a common-emitter (CE) configuration. The circuit connection to measure the input characteristics with ASMU is shown in the inset of figure 5(a). The B-E junction is forward biased with V_B2, and the base current is measured for the different collector-to-emitter (V_CE) reverse bias voltage, which is sourced by V_B1 as shown in figure 5(a).

Zoom In Zoom Out Reset image size

In this circuit configuration, both the DAC's are used as a single-ended voltage source with an additional ground reference (0 V). A 100 Ω collector resistor (R_C) is added in series with the collector to limit its current and eliminate any loading effect on the DAC. To avoid the domination of external base resistance over the transistor's small input resistance, no resistor is added in series with the base.

It is observed that the B-E junction behaves like a p-n junction with a cut-in voltage of 0.67 V, which is close to the ideal value for a silicon-based transistor (0.7 V). The increase in V_CE depletes the base region resulting in an increase in the cut-in voltage and decrease in the base current (I_B) as shown in figure 5(a). The transistor's input resistance can be calculated with equation (7), which is found to be 7.39 Ω for V_CE = 0 V.

$$\begin{eqnarray}&&{R}_{in}={\left.\displaystyle \frac{{\rm{\Delta }}{V}_{BE}}{{\rm{\Delta }}{I}_{B}}\right|}_{{V}_{CE=const.}}\end{eqnarray} \tag{ 7 }$$

$$\begin{eqnarray}&&{R}_{out}={\left.\frac{{\rm{\Delta }}{V}_{CE}}{{\rm{\Delta }}{I}_{C}}\right|}_{{I}_{B=const.}}\end{eqnarray} \tag{ 8 }$$

$$\begin{eqnarray}&&\beta ={\left.\displaystyle \frac{{\rm{\Delta }}{I}_{C}}{{\rm{\Delta }}{I}_{B}}\right|}_{{V}_{CE=const.}}\end{eqnarray} \tag{ 9 }$$

Where, ${R}_{in}$ and ${R}_{out}$ are the input and output resistance, I_C is collector current, and β is the current gain of the transistor in CE mode. Similarly, the transistor's output characteristics are carried out by following the circuit connection as depicted in the inset of figure 5(b). The I_C is measured as a function of V_CE by varying I_B, as shown in figure 5(b). For V_CE ≥ 0.33 V, the transistor is observed to be working in the active region, where output I_C becomes a function of input I_B. The transistor's output resistance is calculated using equation (8) and is found to be 27 kΩ for I_B = 19.4 μA. Using the equation (9), the transistor's current gain is calculated to be 146.1 at V_CE = 1 V, which is very close to the typical gain of 140 mentioned in the transistor datasheet [29].

Although the developed ASMU system with two 12-bit DAC and 16-bit ADC is a versatile characterization tool for multi-terminal semiconductor devices, it has few technical limitations, including the moderate measurement resolution, low speed of response, and the inability of current sourcing. However, such limitations of measurement can be addressed by employing a high-bit, high-speed, precision ADC chip [30]. The reported ASMU system can also be upgraded to a dual-source ASMU by utilizing a high-bit DAC chip with programmable current and voltage-output feature [31].

In summary, a simple technique for the fabrication of a microcontroller-based source measure unit system is described for characterizing multi-terminal semiconductor devices. The ASMU system consists of an Arduino Nano microcontroller, two 12-bit DAC, and 16-bit ADC for bidirectional voltage biasing and measurement in differential mode. The system is powered with the computer's +5 V USB port. The ASMU is programmed with the LabVIEW environment for real-time data acquisition. The system is capable of sourcing and sinking power in all four quadrants. The differential voltage source's linearity is found to be 0.93, which shows that the output voltage closely follows its corresponding programmed value. The differential voltage biasing range is observed to be ±4.65 V with an optimum resolution of 5 mV. The current measurement range is found to be ±14.6 mA with a resolution of 7.8125 μA. A minimum time delay of 10 ms is found to be essential between voltage sourcing and measurement. The ASMU system's performance for the 1N4007 diode is highly comparable with other commercial measuring systems, including ET-2076A and NI USB-6009. The LabVIEW program is customized to generate different time-varying low-frequency AC voltage signals, including sinusoidal, triangular, sawtooth, square wave, and the diode's corresponding output wave response is measured. Similarly, the LabVIEW program is also customized to measure the real-time photoswitching response of IR photodiode at a fixed voltage bias. The most extraordinary feature of the ASMU system is its ability to measure a three-terminal transistor device. The input resistance, output resistance, and current gain for BD433 transistor are found to be 7.39 Ω, 27 kΩ, and 146.1, respectively. The experimental results of devices obtained with ASMU are observed to follow their respective technical datasheets. Therefore, all these results suggest that the fabricated ASMU system can be employed for characterizing semiconductor devices with maintaining adequate precision, cost-effectiveness, and low-power consumption.

The author is grateful to Prof. Anderson S L Gomes from the Department of Physics, UFPE, for providing essential laboratory facilities for this research. The author would like to thank the PNPD Program, CAPES/UFPE, for a post-doctoral fellowship.