Improve Semiconductor Test Performance by 45X with the NI PXI-4110 Programmable Power Supply

The process of characterizing a device, whether it involves a simple discrete component like a diode or something more complex like an analog-to-digital converter, involves sequentially stepping through a series of values and recording the response. In the case of the diode, the stimulus value is typically voltage and the measured value is typically current. In order to fully understand the operation of the diode, as many different values of voltage are applied to the diode as possible. In a production environment, the speed at which voltage outputs can be varied and current measurements can be taken determines the total output capacity of the plant. This paper will compare the performance of a diode characterization system using a traditional, GPIB-controlled power supply to the NI PXI-4110 programmable power supply.

Diode Basics

The basic concept of a diode is to allow current to flow in one direction but block current flow in the other direction. This is made possible by the bond between N-type and P-type silicon found in a diode. While both N-type and P-type silicon are conductors, when bonded together and oriented properly, they will (in theory) not conduct any electricity. When the terminals are reversed, they should allow current to flow freely. This, of course, does not happen perfectly in reality. The graph below shows a more realistic depiction of current flow versus voltage polarity:


When the diode is "reverse-biased", it should block all current. In reality, a small amount does flow, perhaps on the order of 10 uA. If the applied voltage is large enough, however, the junction completely breaks down and freely allows current flow. When the diode is "forward-biased", there is a threshold (usually around 0.7 V) that must be achieved before it operates ideally, allowing free current flow. This is often referred to as the "turn-on" voltage of the diode.

Diode Test

As was earlier mentioned, the process of testing a diode involves applying a series of voltage values to the device under test and measuring the current response. The test operator simply needs to sweep across more voltage values to achieve greater detail on the function of the diode. The following is example test code for a voltage sweep in LabVIEW:


The process is as follows:

1. Initialize the device
2. Configure the current range the device is used in - for devices with a single current range, this step isn't necessary
3. Configure the current limit to be used - this sets the maximum current the device should source
4. Enable the output
5. Set a voltage to be output
6. Read back the voltage and current on the output channel
7. Repeat for as many steps as needed - remember that more steps provides more resolution
8. Disable the output and close the device

Typical diode tests should include sweeping voltage from 0 V to at least the transition voltage region - again, near 0.7 V. More complete sweeps might also include some characterization in the negative region to verify reverse-biased operation. The number of points to be measured in these ranges will vary depending on the accuracy requirements of the test.

Comparison of Sweep Times - Agilent E3631A versus NI PXI-4110

Because the number of points swept determines the resolution, it is valuable to have a power supply that sweeps through a series of points very quickly. To compare sweep times (and, therefore, test time per device), a GPIB-based power supply (Agilent E3631A) can be compared to the NI PXI-4110. For this test, the diode will be swept from -1.6 V to +1.6 V to characterize operation in both the forward and reversed bias regions. The properties are as follows:

• Minimum voltage = -1.6 V
• Maximum voltage = +1.6 V
• Number of steps = 40
• Step size (based on number of steps) = 80 mV< /li >

In order to minimize any differences in code that could make the test otherwise unfair, the IVI-DCPower class driver will be used to program both devices. The concept of Interchangeable Virtual Instrument (IVI) programming was developed so that pieces of code could be written in a hardware-agnostic fashion - the same piece of code could work on devices made from multiple vendors. Both the Agilent E3631A and NI PXI-4110 power supplies are IVI-compliant, so using the IVI-DCPower class driver allows us to communicate with both devices and keep all settings, code inefficiencies, or other operating system delays consistent between devices. This helps ensure a fair comparison.

Also, because neither device has a single bipolar output (meaning it can output both positive and negative voltages), this test will involve using one channel for the negative values and another channel for the positive values. The settings for each device will then be as follows:

Agilent E3631A Negative Values Channel 3 (0 to -25 V operation) Start value = -1.6 V End value = 0 V Number of steps = 20< /li > Positive Values Channel 2 (0 to +25 V operation) Start value = 0 V End value = 1.6 V Number of steps = 20< /li >

NI PXI-4110 Negative Values Channel 2 (0 to -20 V operation) Start value = -1.6 V End value = 0 V Number of steps = 20< /li > Positive Values Channel 1 (0 to +20 V operation) Start value = 0 V End value = 1.6 V Number of steps = 20< /li >

A switch matrix can be placed between the power supply and diode to facilitate quick transfer from the negative channel to the positive channel in either test. Because the matrix will be a constant for both devices, it is considered negligible in the comparison.

A graph of the test results (which is the same for both devices) is shown below:

Given the comparison test setup described previously, the results are as follows:

As the results show, the PXI-4110 power supply was approximately 45 times faster than the Agilent GPIB-based power supply. With around 600 ms required for each voltage step on the Agilent device, it is easy to see the trade-off that would have to be made when calculating the total number of points to sweep compared to the test time required. The PXI-4110, however, performs the entire 40-point sweep in less time (533 ms) than it takes one point to be measured on the GPIB-controlled Agilent E3631A.

With the PXI-4110, it would be very simple to double or triple the number of points swept across the diode and, in effect, increase the resolution of the test on the diode while maintaining short test times. With further testing on the PXI-4110, a sweep of values with 10X the resolution required only 4.6 seconds with step sizes at 8 mV. The PXI-4110 has the ability to step through voltages as low as 400 µV on the 20 V channels; for the -1.6 V to +1.6 V sweep, that would be just over 8000 voltage steps. Total test time for that ultimate level of resolution would be just 92 seconds - only 4 times as long as the Agilent GPIB power supply with 200 times the resolution!

As the comparison noted, both devices were programmed using the same IVI-DCPower class driver. Because of this, little or no code modification is required when switching between devices, making a hardware upgrade from the Agilent E3631A or other GPIB-based power supplies very simple, yielding decreased test time or increased resolution with the PXI-4110.

Reasons for the Speed Improvement using the PXI-4110

The comparison test involved not only setting a voltage value but also making two measurements - voltage and current - for each setpoint. Consequently, the transfer speeds of the bus and the measurement architecture used on each device plays a large role in the time used per step.

The fact that the PXI-4110 is built around the PXI bus significantly helps optimize programming and measurement speeds. Sending program parameters and retrieving data is greatly facilitated by the 132 MB/s PXI bus speeds. With three channels that each require voltage/current programming and measurement parameters plus status information (compliance limit, warnings, errors, temperature, etc), the amount of data that needs to be moved in both directions can challenge traditional bus solutions. PXI can move this data in microsecond timeframes compared to several milliseconds or 10s of milliseconds required with traditional instrument bus architectures (GPIB or RS232). Thus, the software and data path overheads are practically negligible for the PXI-4110.

The PXI-4110 measurement architecture is also notable for its speed advantage over traditional measurement approaches. Integrating analog-to-digital (ADC) architectures are traditionally used in power supply measurements. These ADCs have advantages for noise but don't give the user much flexibility to optimize speed, especially "under the hood" in dynamic stimulus-response devices such as precision power supplies or SMUs. With multichannel power supplies, the slower ADC creates significant overhead for acquiring the multiple parameters required to represent the status of the output.

The architecture used in the PXI-4110 is based on similar measurement engines used in National Instruments high-speed data acquisition systems. The ADCs are 200 kS/s, 16-bit high-bandwidth converters - one for the nonisolated channel and another for the two isolated channels. The ADCs are used for both measurement read back as well as output regulation. The net loop speed of the measurement is in the 3 kS/s range. In other words, every 300 µs, the measurement engine returns six measurements - voltage and current output for each of the three channels (as well as the output regulation loop data). This is fast enough to watch the settling time of all the channels simultaneously (rise times in the millisecond range) and is faster than required for any stimulus-response step waveforms required by the user.

Conclusion

In a high-volume manufacturing environment, a semiconductor characterization test on a device like a diode is a trade-off of test resolution and test time. There is a huge difference between a device based on the PXI platform and modern measurement techniques compared to a GPIB-based device with slower ADCs. Using the NI PXI-4110 can provide up to 45 times the total test throughput compared to traditional bus solutions, saving both time and money for the manufacturing facility. Also, because the PXI-4110 is built on the IVI driver standard, it is possible to upgrade existing equipment to the PXI-4110 with little modification to the software, providing an "instant upgrade" to the system test speed.
Related Links:
NI Programmable Power Supplies and Precision Sources
NI Modular Instrumentation
Measurement Fundamentals

Reader Comments | Submit a comment »