Text preview for : www.thinksrs.com-IEEEcontrol_article_SIM.pdf part of Stanford Research Systems www.thinksrs.com-IEEEcontrol article SIM Stanford Research Systems www.thinksrs.com-IEEEcontrol_article_SIM.pdf
Back to : www.thinksrs.com-IEEEcont | Home
Developing a New Product Line The first control problem I really had to tackle myself was
CSM: Thank you for speaking with CSM. I understand while I was at NASA, where I was upgrading a home-built
you've had a diverse background beginning in physics and Fourier-transform infrared spectrometer. The problem
astronomy. Could you describe your career path leading was to move a somewhat massive linear stage at a uni-
up to your present position? form programmed velocity for a scan. We replaced a
Matt: I became interested in physics in high school in finicky optical interferometer with a linear encoder,
New Jersey, which led me to MIT, where my undergradu-
ate major was physics. Looking back, it was during this
time, surrounded by really good engineering students,
that I began to appreciate design work. I continued
studying physics at the University of California at Berke-
ley, working on high-energy particle physics experiments
at SLAC (Stanford Linear Accelerator Center, Menlo Park,
California) and Fermilab (Fermi National Accelerator
Laboratory, Batavia, Illinois). I cut my teeth in electron-
ics at Fermilab, and I've been heavily involved in elec-
tronics design ever since. After receiving my Ph.D. in
1992, I switched to astrophysics, joining a group based at
NASA's Goddard Space Flight Center (Greenbelt, Mary-
land) making measurements of the cosmic microwave
background radiation. Our group built instruments that
flew on high-altitude balloons, and the group had a tradi-
tion of building most of its own equipment.
After five years and four balloon campaigns, I was
interested in doing something different. In basic research,
a truly optimized experiment should fall apart shortly
after the final measurements are taken. If it doesn't, then Matt Kowitt and his son Quinn, age five. Matt is employed by
you might have spent too much time building a "gold- SRS in Sunnyvale, California, where he is project lead for
plated" apparatus, and you run the risk of your competi- SRS's new Small Instrumentation Modules product family,
tion beating you to publication! I was much more which includes the SIM960 Analog PID Controller.
interested in designing instruments that were built to
last, which led me to my current position at SRS. Our
research group at NASA had several of the company's
products, and we were always impressed with the techni-
cal support we received. After joining the R&D depart-
ment at SRS, I discovered that my path here was not at all
unique since a majority of the staff were former research
scientists in physics or physical chemistry.
CSM: Did any of the projects you worked on involve con-
trol technology, and, if so, what was your role in develop-
ing the designs and implementing the systems?
Matt: My first exposure to control systems was when
I was working at SLAC. Our experiment involved sending
short laser pulses out of a trailer, down a ventilation
shaft into the accelerator tunnel, and into the main
beam pipe, where the pulses would scatter off the high-
energy electron beam. Each mirror in the laser path was
instrumented with tilt actuators and optical sensors,
and a slow control system kept the beam centered all
the way down. A professional staff engineer from
Lawrence Berkeley Lab was responsible for building that The SIM960 Analog PID Controller is a broad-bandwidth, all-
system, but as a graduate student I was involved in analog controller with front-panel adjustable gains and built-
determining some of its performance requirements. in anti-windup and bumpless transfer features.
February 2005 IEEE Control Systems Magazine 13
along with a linear velocity transducer (LVT). My origi- find an approach that would work well on an analog con-
nal design ran entirely in software using an early version troller. Brian had a copy of Astrom & Hagglund's book
of LabVIEW on a 1990-vintage Macintosh computer, and PID Controllers: Theory, Design, and Tuning, (1995), which
the results were very poor! I didn't know any control the- has a very nice discussion of integrator windup. It was
ory at the time, and so there was a lot of trial and error. I here that he found a description of the conditional inte-
finally built an analog loop to control velocity using the gration scheme we eventually settled on. Basically, the
LVT and left the computer to supervise this loop with input to the integrator is switched off whenever the out-
slow setpoint updates to keep the velocity true to the put is in saturation and the sign of the error signal is "in
linear encoder. That was when I learned that a few well- the same direction" as the output saturation. This
placed op-amps can save weeks of software headaches. approach prevents the integrator from being pushed fur-
ther into saturation, and was straight-forward to imple-
CSM: SRS has introduced a unique product, namely, the ment with CMOS logic and some analog comparators.
SIM960 Analog PID Controller. How did this product origi- The bumpless transfer feature allows a user to switch
nate, and what is the market? from a manually controlled output level to the controller
Matt: The original focus area for the SIM product line output without a step discontinuity. The key to making
was instrumentation for researchers in low-temperature this work is the integrator stage, which has an undeter-
physics. A key element in that line was a temperature mined initial value in the control equation. We take advan-
controller. We were targeting a very demanding cus- tage of this free parameter to pre-charge the integrator to
tomer group, one where noise concerns are paramount. just the right value, so that the PID output starts at the
We wanted to avoid the prospect of quantization and manual output voltage when the switchover occurs. This
discretization artifacts upsetting a sensitive measure- way, a user can "manually fly" a plant to about where they
ment or damaging a sample. Working together with want it, and then switch over to control without suffering
Brian Mason, a physicist at SRS who is the design engi- a big "bump."
neer directly responsible for the SIM960, we decided on
an all-analog signal path. We also felt that computer CSM: Do you see future products for control applications?
control, as well as local (front-panel) control, was I was thinking, for example, of analog filters for noise
important in a modern instrument. The architecture reduction and antialiasing, which are needed for control
that emerged uses all analog op-amps in the signal path, experiments and prototyping, but few commercial prod-
with parameter adjustments implemented using resis- ucts are available.
tive D/A converters as programmable resistors. Matt: Funny you should ask about filters! Our newest
We soon realized that this instrument would also have release of products in this line includes a programmable
significant bandwidth and probably have interesting uses analog filter module, the SIM965. Briefly, this unit can be
outside of temperature control. Applications that imme- set to low-pass or high-pass as either a Bessel or Butter-
diately came to mind include vibration control and posi- worth filter, with the cutoff frequency programmable
tion servos for scanning probe microscopy. As it has from near 1 Hz to several hundred kHz, and the filter
turned out, a popular use for the SIM960 has been stabi- order selectable from 2-pole to 8-pole. Like the PID con-
lizing the wavelength of external cavity diode lasers. Our troller, the filter implementation is entirely analog, with
goal was to build an instrument that users could pick up continuous-time op-amp signal paths throughout. Addi-
and start using immediately without having to study a tional modules that should be of interest to the control
programming interface. We even put the transfer function community are a four-input summing amplifier (where
equation on the front-panel artwork, so users can quickly each input can be selected as inverted, noninverted, or
understand the exact form of the control law. off), a "vernier" scaling amplifier with continuously
adjustable gain and input offset, and an analog limiter
CSM: The SIM960 is advertised as having antiwindup and module with programmable upper and lower limits.
bumpless transfer features. Can you tell us a little about
how these features were designed and implemented? CSM: These control-related products from SRS should be
Matt: Well, the antiwindup feature involves preventing of interest to the control community for experiments and
the integral term from marching off into deep saturation prototyping. I'm sure we'll hear about successful applica-
whenever the loop saturates. We were sensitive to the tions in the future. Meanwhile, thanks for talking to IEEE
problems of windup, including delayed recovery from CSM, and we wish you and SRS the best of luck in your
saturation and potential limit oscillations, and wanted to traditional markets as well as in control technology.
14 IEEE Control Systems Magazine February 2005