• Image: Mary Anne Tupta, senior staff applications engineer, and Bob Green, senior market development manager, both of Keithley Instruments, discuss measurement system requirements with Dr. Konstantin Novoselov of the Condensed Matter Physics Lab at the University of Manchester. This photo was taken on the day Dr. Novoselov and Dr. Andre Geim were awarded the 2010 Nobel Prize in Physics “for groundbreaking experiments regarding the two-dimensional material graphene.”

How did you get into electronics/engineering and when did you start?

I received my BSEE from Cornell in 1971, which was back when a computer still took up most of a room. Just to make sure I really wanted to be an engineer, I chose a college that had a co-op program so I could be certain I’d enjoy the field. Fortunately, while at Cornell, I had an opportunity to be a co-op in the medical electronics division of a larger instrumentation manufacturer, which is what started me down the biomedical engineering track rather than something like aerospace.

After Cornell, a new MSEE program in biomedical engineering was starting at Washington University, and I was fortunate enough to get a Public Health scholarship. So I repaid the government by going to work as a clinical engineer at the Veterans Administration hospital in Cleveland. I was designing special healthcare areas, acting as a consultant to the medical staff on medical instrumentation, and helping build an organization to maintain all the new electronics. It was a really exciting time to be in biomedical electronics: the first versions of ultrasound equipment and CAT scanners were coming out, and better stress test systems were being developed. Microprocessing technology and advancing electronic technology were being incorporated into new medical instrumentation.

My next stop was managing the controls division of Pyronics, a company that designed burners and control systems for heat-treating furnaces and boiler plants, as well as dyeing and drying systems for textile mills. In 1983, I had my first opportunity to join Keithley, where I supported some scientific products, including the electrometers and picoammeters, as well being the product manager for the digital multimeter line. I was also taking some more general business courses at Case Western Reserve University, which gave me the courage I needed to start my own company in 1990, Cortest Measurement Systems. We did contract design and manufacturing of quite a range of electromechanical and electronic products, including tool presetters, gas detection instruments and systems, specialty pumps, and battery module test systems.

I re-joined Keithley in 1996, where I’m a senior market development manager, now in charge of many of the new power supplies products being developed. In this role, I’ve really come to enjoy interfacing with customers, going out and seeing their applications, as well as being a part of the new product development team.

Any lessons you’ve learned over the course of your career?

One of the things you learn to appreciate when you’re in business for yourself is the importance of immediacy. At a big company, it’s all too easy to take your paycheck for granted a little bit. But when you’re running your own business, you get a different perspective. If you don’t produce, you don’t get paid. And you have to pay your employees first. And the federal government. And everybody else. When you’re the boss, getting things done quickly is very important.

Any favorite projects over the years?

At Cortest, I lead a team that developed a multi-channel gas detection system that was used for oxygen monitoring to protect worker health/safety in underground switching stations. We sold a very similar system design to Cleveland’s Regional Transit Authority for monitoring their bus storage garages for methane and carbon monoxide after they converted to natural gas-powered buses. My graduate work with biomedical engineering dovetailed well with that.

As a product manager, I introduced some specialty, battery-simulating power supplies for the rapidly expanding cellular phone industry in the late 1990s. These products became the production power supplies for the industry’s market leaders. We had some great design wins.

What would you consider the highlights of your career?

Probably the most recent career highlight was meeting the winners of the 2010 Nobel Prize in Physics on the day the prize was announced. I was making a customer visit to Manchester University’s Condensed Matter Physics Lab with two Keithley colleagues. We were there to talk to Dr. Novoselov about one of his test systems used to study the field effect properties of graphene, a one-atom-thick layer of carbon atoms densely packed in a honeycomb crystal lattice, because the system includes several Keithley instruments, including the Model 2182A nanovoltmeter and two Model 2400 SourceMeter® instruments. We were there to understand his measurement requirements and offer advice. He had to excuse himself to take a call, which turned out to be from the Nobel Committee. Everyone in the lab was absolutely ecstatic, cheering and taking pictures. Then the BBC came to interview him and his collaborator, Dr. Geim. Dr. Novoselov actually apologized to us for the interruptions. I felt like I was part of science history that day; it just does not get any better than that.

What direction do you see your business heading in the next few years?

New models of consumer electronic devices such as smartphones and tablet computers are superseding old ones at an accelerating rate, each demanding new system configurations to characterize its components and test the finished product. Electronics manufacturers face constantly new measurement challenges. Test vendors need to address these challenges in a way that makes economic sense for their customers—few electronics manufacturers could afford to start configuring new test lines from scratch, even if they wanted to. Keithley’s strategy is to offer customers a range of scalable, flexible solutions that can adapt readily and cost-effectively to accommodate new device technologies and test applications.

What challenges do you foresee in our industry?

Because of this constant technology churn, all electronics manufacturers are being pushed to do more with less and do it faster than ever before. Not that long ago, many of the test systems and much of the software that electronics manufacturers used were configured and maintained in-house by their own employees. Today, a lot of those employees simply aren’t there any longer as manufacturers have cut their test engineering staffs to the bone and beyond to minimize their costs. That means that lots of electronics OEMs are forced to rely more extensively on their vendors to acquire the full range of test solutions they need, for environments from R&D through production. They’re also counting on their vendors to help them control their cost of test, both at the time of initial test system purchase and over the long term. That’s been a big part of Keithley’s focus for a long time–helping customers choose solutions that will allow them to adapt as new test requirements emerge and showing them how to wring the maximum productivity from their existing test assets.

For characterizing many types of electronic devices, manufacturers need equipment with a wide dynamic range, that is, that’s capable of operating with a wide range of signal levels. For example, power MOSFETs are designed to have very low resistance and handle very large currents when turned on but are also designed to have very high resistance and allow nearly zero current to flow when turned off. In the on state, this current is commonly as high as tens of amps; in the off state, this current may be less than nano-amps. Power diodes and high brightness LEDs have similar dynamic range requirements as well for full characterization. On these kinds of devices, when a forward bias voltage below the threshold value is applied, device currents are very low. As voltage is swept from 0V to the threshold, device current goes from the sub-nano-amp range up to milli-amps. As the bias voltage reaches and then exceeds the threshold, test currents increase very rapidly, reaching as high as tens to hundreds of amps depending on the device.

Obviously, manufacturers of these devices (or of the products into which they are built) prefer test equipment capable of taking accurate measurements over a wide range because the use of such equipment reduces the number of pieces of test hardware required, as well as both system cost and complexity. One of the ways in which Keithley has been helping electronics manufacturers meet this challenge is with our line of Series 2600A System SourceMeter® instruments, which combine the most power with the widest range of signals available in a single instrument. The latest addition to the line, the Model 2651A High Power SourceMeter instrument, can deliver up to 200W of DC power and 2000W of pulsed power to a device. It can measure as much as 50A of current and is also capable of measuring with a maximum resolution of 1pA. The dual-channel Model 2636A has the industry’s widest dynamic range with the ability to measure signals from 10A all the way down to 1fA, offering 16 decades of current resolution.

Our SourceMeter solutions are in a class of instruments known as source-measurement units or SMUs, which are often incorporated into automated test systems because of their speed and flexibility. Essentially, SMUs are fast-response, read-back voltage and current sources with high accuracy measurement capabilities, all tightly integrated in a single enclosure. They are designed for circuit and device evaluation where a DC signal must be applied to a device under test and the response to that signal measured. They are capable of four-quadrant operation, acting as a positive or negative DC source or as a load. They also provide highly repeatable measurements, typically with 5-1/2- or 6-1/2-digit resolution. SMUs can typically be used to perform sweeps of both current and voltage that can be used to determine the I-V characteristics of a device under test.

Since they were first introduced more than two decades ago, SMUs have evolved into a class of multi-purpose instruments that electronics manufacturers turn to regularly for applications like semiconductor device fabrication, process development, and product research/design; production verification of electronic products such as portable wireless devices; production and development of new advanced materials for devices like solar cells and high-brightness LEDs, and many others. They’re available in a variety of form factors, including the more traditional instrument SMUs (the kind Keithley designs) and component SMUs (often card- or module-based instruments designed to be plugged into a backplane in a mainframe or chassis as part of a larger test system). Despite some claims to the contrary, traditional instrumentation remains a vital, growing part of the test and measurement industry. Although older communication interfaces like GPIB or RS-232 may become obsolete over time, instrument-based SMUs, used either alone or integrated with other SMUs in a system, typically provide the fastest, most accurate, most flexible solutions for the widest range of demanding applications. In contrast, component SMUs often must compromise their performance to offer a specific form factor.

It sounds like the SMU market is getting bigger and increasingly competitive.

Absolutely. There are a lot of competing claims out there about the advantages the various form factors offer. For example, vendors of component SMUs often try to claim they offer the same wide dynamic range that some instrument SMUs can provide. However, the form factor of these devices limits them to a dynamic range several decades smaller than Keithley’s instrument-based SMUs. On the high end of the range, they are limited by how much power the chassis can provide, and most component SMUs will top out at 100mA. On the low end, the electromagnetic interference of all the circuitry designed into a small space with inadequate room for shielding creates too much electrical noise for any kind of low-level measurement to be practical; as a result, it is uncommon to see a component SMU with any current ranges lower than 10 micro-amps.

What are some of the other differences between instrument SMUs and component SMUs?

As I mentioned before, SMUs are often incorporated into automated test systems. No matter how fast an individual SMU may be, its merits are wasted if its performance slows to a crawl when it’s integrated into a system. Component SMUs are inherently less affected by this issue thanks to their high speed, low latency connection to the host system through a PCI or PCIe backplane (133MB/s for PCI. 250MB/s for PCIe x1). In contrast, instrument-based SMUs communicate with a host system through an external bus such as GPIB, which has only a fraction of a backplane’s speed (1.8MB/s standard). Keithley’s engineers recognized this when they designed the Series 2600A SourceMeter instruments and built them to operate autonomously from the host system by using a Test Script Processor (TSP®) and to communicate and synchronize with one another via the TSP-Link® bus, a high speed, low latency inter-unit communication interface.

Earlier instrument-based SMU designs required that each command be sent to them from the host one line at a time and, because all the instruments share the same bus, only one instrument can be addressed and talked to at a time. That means a lot of time has to be spent sending commands and data across the bus to one instrument at a time while the remaining instruments often sit idle. In contrast, TSP technology all but eliminates the time spent sending commands by allowing the instruments to run test scripts autonomously from the host system. Once a script has been loaded into a TSP-based SourceMeter instrument, an entire test sequence can be performed, with the host required to send only a single command that instructs the instrument to run the script.

TSP-Link technology removes the need for multiple SourceMeter instruments to be attached to and individually addressed by the same bandwidth-limited GPIB bus. With TSP-Link technology, only one SourceMeter instrument is connected to the GPIB bus; the remaining SourceMeter instruments are connected to the first in a “daisy-chain” configuration using inexpensive CAT5e crossover cables. By connecting the additional SourceMeter instruments to the first via TSP-Link technology, the SMUs of these instruments appear as additional SMU channels on the first SourceMeter instrument and can be accessed quickly by a script running on the first SourceMeter instrument.

Are there any other advantages to the system builder of this test configuration approach?

Unlike with component SMUs, channel expansion with TSP-Link technology is not limited to a handful of slots in a mainframe. TSP-Link technology’s mainframe-less expansion allows connecting up to 32 instruments together, making it possible to create a system with as many as 64 SMU channels. Also, because the SMUs are instrument based, the amount of power available to them is not limited by what a chassis can provide to them. Even in high power component SMU-based systems, some models provide a maximum power output of only 84W. By connecting 32 Model 2651A High Power SourceMeter instruments together via TSP-Link interface, it is possible to create a system capable of delivering 6.4kW of DC power.

TSP-Link technology can boost system throughput simply by reducing the amount of communication necessary between the instruments and the host. However, the real power of this technology lies in its ability to run multiple tests in parallel. In traditional SMU systems, whether they are component-based SMUs in a chassis or instrument-based SMUs on a GPIB bus, access to the bus is limited and the host must send commands to each SMU one at a time. Adding more SMUs to the system means increasing the number of devices the host must address and to which it must send commands. Because commands cannot be sent to more than one SMU at a time in these systems, all tests must be performed sequentially. In contrast, in a system connected via a TSP-Link interface, instruments in the network can be divided into groups, with each group having its own Test Script Processor capable of running a script in parallel with every other group in the system. Groups may contain a single SourceMeter instrument or multiple SourceMeter instruments and are typically sized according to the number of SMU channels required to test a device.

For example, if the device under test is a MOSFET with four terminals (Gate, Drain, Source, and Base) and the test is on wafer and requires one SMU per pin, then the group could be made from two dual-channel SourceMeter instruments, such as the Model 2636A Dual-channel System SourceMeter instrument. Once groups have been defined and each group has been given a script to run, the host can instruct all groups to begin execution in parallel with a single command. With the scripts for each group already in memory, the host can repeat the test simply by sending the command again.

Now, for example, if a test sequence on an on-wafer device takes a total of one second to complete, at that rate it would be possible to test 60 sites per minute. If another group is added to the TSP-Link network, the test would still take one second to complete. However, the addition of a second group makes it possible to test two parts in parallel so throughput is doubled to 120 sites per minute. TSP-Link technology simplifies increasing system throughput just by adding more groups of instruments to the network.