Reconfigurable Instrumentation Technologies,
Architectures and Trends
Rok Ursic
President and Founder
Instrumentation Technologies
Srebrnicev trg 4a, SI-5250 Solkan, Slovenia
rok@I-tech.si, www.I-tech.si
Abstract. Reconfigurability is liberating radio-based instrumentation devices from chronic
dependency on hard-wired characteristics of the radio front end. Today the evolution toward
practical "reconfigurable" instrumentation is accelerating through a combination of approaches.
This evolution is challenging analog, digital and software designers and the associated product
development process. The goal of this paper is to give a short overview of the latest
technologies, architectures and trends in this field.
INTRODUCTION
The focus of this paper is on instrumentation systems that process radio frequency
(RF) signals. The family includes beam position monitors, beam current monitors,
tune systems, low-level RF systems, transverse and longitudinal feedbacks and
similar. This is not a comprehensive overview or a tutorial on reconfigurability. It is
a vision that builds on my experience.
Motivation
The idea is extremely simple but ambitious: to build a system that can "grow" with
an accelerator and support new requirements or applications without changing
hardware. To realize this idea, we need reconfigurable systems.
The idea is not new and has been around in the telecommunication sector for a
while under the guise of software defined radio. Through the 1970s and 1980s, radio
systems migrated from analog to digital in almost every respect from system control to
source and channel coding to hardware technology. In the early 1990s, the software
radio revolution began to extend these horizons by liberating the radio-based services
from chronic dependency on hard-wired characteristic of the radio [1]. These kind of
systems are standardized in manufacturing, which translates into lower cost-per-unit,
but customized in application, which translates into flexibility and future proof
solution.
CP648, Beam Instrumentation Workshop 2002: Tenth Workshop, edited by G. A. Smith and T. Russo
© 2002 American Institute of Physics 0-7354-0103-9/02/$19.00
179
Space for Creativity and Innovation
System architecture is the creative ground for beam instrumentation designers.
Building blocks are, on the other hand, sophisticated integrated circuits. The beam
instrumentation designer has no influence on design, cost and performance trends in
this field. Developments are governed by fierce competition in the economy of scale
markets like telecommunications, radar, ultrasound and similar. Even though the
instrumentation designer cannot influence developments in this field, it can certainly
benefit from them.
TECHNOLOGIES
In 1965, Gordon Moore, one of the founders of Intel Corporation, predicted that the
number of transistors that could be constructed on a unit area of silicon would double
every 18 months [2]. Known as "Moore's Law", this rule-of-thumb has formed the
basis for predictions on such diverse electronic phenomena as the capacity of memory
devices, the capabilities of 3D graphics accelerators, and the performance of
microprocessors and DSPs. A second aspect of Moore's law is his prediction that the
doubling of transistors would be achieved for the same price. This means that if we
continue to use a constant number of transistors, then the price-per-transistor will
halve every 18 moths as new device technologies become available, which translates
into lower product costs to the end user.
Furthermore, a lesser-known section of the famous Moore's paper deals with the
linear electronics. He stated: "Integration would not change linear systems as radically
as digital systems". In other words, the cost/performance gap between analog and
digital world will widen with time motivating design engineers to implement more and
more functions in digital domain.
Time
FIGURE 1. One of the consequences of Moore's Law is that cost/performance gap between analog
and digital electronics is widening with time.
180
Analog to Digital Converter Integrated Circuits
The wideband ADC is one of the fundamental components of the reconfigurable
instrumentation. It is the key building block that connects analog and digital domains.
The ADC is a hybrid, mixed signal device with analog and digital sections. The
cost trend for ADC is not following either of the two curves shown in the graph in
figure 1. Analysis of this technology is outside the scope of this paper. Walden [3]
studied the relationship between ADC performance and technology parameters.
TABLE 1. State-of-the-art commercially available ADC (2002).
Model No.
MAX1420
CLC5958
AD6645
AD9226
AD9244
AD9433
ADS2807
Manufacturer
Maxim
National Semiconductor
Analog Devices
Analog Devices
Analog Devices
Analog Devices
Texas Instruments
Number of
bits
12
14
14
12
14
12
12
Sampling
Frequency
[MHz]
60
65
105
65
65
125
50
3dB
Bandwidth
[MHz]
400
210
250
750
750
750
270
Digital Processing Integrated Circuits
There are three established families of digital signal processing integrated circuits:
application specific integrated circuits (ASIC), field programmable gate arrays
(FPGA) and general-purpose digital signal processors (DSP). A new breed of
devices/technology, re-configurable computers, is gaining momentum in the wireless
market. Despite the fact that this technology will have to prove its performance and
market sustainability, we included them in table 2 due to their lucrative potential.
The cost/performance trend for digital signal processing devices approximately
follow the Moore's law for digital integrated electronics. Figure 2 compares these
technologies from hardware adaptability, programmability and performance point of
view. Before choosing a particular technology or combination of technologies for a
specific design, however, system designer must also consider other characteristics like
interconnect capacity, size-power tradeoffs, degree of required parallelism, etc.
Discussion on those topics is outside the scope of this paper and can be found in [1].
TABLE 2. A coarse comparison of different signal processing technologies._________
Reconfigurable
__________________ASIC_______DSP______FPGA_____computers
Hardware adaptability
+
+
Programmability
+
+
Performance
+
0
0
+
181
ARCHITECTURES
Conventional instrumentation designs implement algorithms in a hard-wired analog
circuitry. They are optimized to perform specific functions. Reconfigurable designs,
on the other hand, take advantage of system architectures that combine analog, digital
and software domains. As one can see from figure 2, which shows the evolution of
super-heterodyne receiver, more and more functions are implemented in digital. This
requires new radio system architecture. It also requires a different approach to radio
system design. Design and development team must be able to integrate broad range of
engineering and management disciplines in order to design develop and lunch a
successful reconfigurable instrumentation device.
Yesterday: Digital basebands
Controller
Today: Digital IF
Controller
Tomorrow: Digital RF
Controller
FIGURE 2. The evolution of superheterodyne receiver. The boundary between analog and digital is
moving towards the "antenna".
The Interface Question
An important aspect of any architecture is how the system interfaces with external
world. In case of beam instrumentation devices this is relatively straightforward as
long as the interfaces are analog. This design practice, which was the norm for years
in this field, has a significant advantage; it offers a clean interface between an
instrumentation device and a control system. However, it has a significant
disadvantage; it hinders development of more sophisticated re-configurable systems.
Analog interface does not allow straightforward integration of digital signal
processing hardware. If we accept the idea that instrumentation people are responsible
for the performance of their systems, then the cleanest interface is on the responsibility
boundary.
The Performance Question
As soon as we accept the idea that digital signal processing is an integral part of an
instrumentation system, we need a new set of performance metrics. Conventional
systems were specified with such parameters as accuracy, resolution, bandwidth, and
similar. For reconfigurable systems we need to add parameters such as throughput,
182
latency time, real-time capability, batch processing depth, processing power and
similar.
Another important aspect regarding reconfigurable systems is that performance
must always be associated with firmware revision. New features or even new
applications can be downloaded to a system without changing hardware.
Evolution Towards a Clean Digital Interface
The following three figures show a possible interface evolution scenario for
reconfigurable instrumentation devices. The aim is to bring the level of "cleanliness"
of the analog interface to the digital domain and in this way facilitate proliferation and
smooth integration of cutting edge digital signal processing technologies into
instrumentation devices.
As noted before, analog interface was a norm for years. It has the disadvantage that
it does not allow integration of digital signal processing into instrumentation devices.
ANGLOS INTERFACE
RF
RF
FIGURE 3. Yesterday: analog interface.
The next step toward the evolution of a clean digital interface is interface at the
driver/backplane level. It has a significant advantage that it opens the possibility for
integrating digital signal processing into instrumentation devices. It is, however, a
challenge from the system integration point of view. The gray area of responsibility
between instrumentation and controls is significant and requires good collaboration
between the two groups.
DRIVER/BACKPLANE
LEVEL
INTERFACE
Analog front end control (gain, pilot signal, LO frequency,...)
RF-
Real time feedback hook
FIGURE 4. Today: driver/backplane interface.
183
A possible scenario, which brings a brand new perspective on the process of
developing reconfigurable instrumentation devices, is shown in figure 5. The enabling
technology is Ethernet. System integration in such configuration is simplified,
because individual cards can have their own processor, operating system and memory
and can communicate independently with other cards. Because nodes can be
operating-system agnostic, integration is no longer required at the driver/backplane
level but ascends to the network and transport layers, which means significant time
savings and simpler design models.
NETWORK/TRANSPORT
LAYER
INTERFACE
CompactPCI/Packet
Switching Backplane
(PICAA6 2.16)
RF-
NETWORK/TRANSPORT
UYER
INTERFACE
RF
RF-
RF
Hook for feedback
(b)
FIGURE 5. Tomorrow: (a) Ethernet within the chassis based on PICMG 2.16 standard for cPCI or (b)
stand alone solutions provide a brand new perspective on the process of developing reconfigurable
instrumentation systems.
184
TRENDS
Life Cycle Customization
Accelerators are complex machines. Their performance depends strongly on the
quality of instrumentation support. New requirements are generated throughout the
life cycle of any accelerator, which in turn requires better performance or even
additional functionality from instrumentation systems. Historically this resulted in
difficult, expensive and time consuming upgrades. Reconfigurable instrumentation
offer future proof solution that facilitates simple and low cost software customization.
Fixed-functionality product
Diff icu11 and expensive hardware customization cycle
Reconfigurable product
Simplejj andjjj !•jj costjj softwarej customizationj cycle
Time
FIGURE 6. Reconfigurable products offer simple and low cost customization throughout the product
life cycle.
Commercial Of The Shelf (COTS)
Specialized knowledge, skills and tools in different engineering areas are needed to
develop, manufacture, supply and provide technical support for the state of the art
reconfigurable products. The resource allocation is, in most of the cases, beyond the
capabilities of a single instrumentation group at particle accelerator facilities. In
addition, the development process for reconfigurable devices is complex and more
expensive that in the case of developing simple analog modules. Standardization in
manufacturing allows suppliers of COTS modules to achieve better quality and price
per module by distributing development cost over a larger volume.
On the other hand, customization is better done by users. They know the issues,
they know the machine and they can tailor the solutions to their specific needs. In
order to do that efficiently they need good tools and support.
Quality Tools and Support
Two key prerequisites for a successful future of reconfigurable products are quality
tool(s) and good technical support. The quality tools should allow modeling and
185
simple, abstraction level customization of the product in laboratory. They should also
facilitate verification of developed models on a real hardware. Good technical support
helps new users to get acquainted with a system. It should also provide system life
cycle support regarding repairs, spare parts, firmware upgrades, etc. The COTS model
will be successful only if those two prerequisites are met.
CONCLUSIONS
Widening gap in the cost/performance trend between analog and digital integrated
electronics provide the foundation for reconfigurable instrumentation systems. These
systems offer in a single integrated solution benefits that were not achievable before
with conventional hard-wired analog modules. New system architectures that will
allow simple system integration, support life cycle customization by means of userfriendly tools will drive the reconfigurable beam instrumentation revolution.
REFERENCES
1. Joseph Mitola HI, Software Radio Architecture, Object-Oriented Approach to Wireless Systems
Engineering, New York: John Willey & Sons, Inc., 2000, pp. 1-31.
2. Gordon E. Moore, "Cramming more components onto integrated circuits", Electronics, Volume 38,
Number 8, April 19, 1965.
3. Walden, R., "Analog to digital converter survey and analysis", JSAC, New York: IEEE Press,
February 1999.
4. Ursic Rok and Raffaele De Monte, "Digital Receivers Offer New Solutions for Beam
Instrumentation", Proceedings of the 1999 Particle Accelerator Conference, New York, 1999, pp.
2253-2255.
5. John Peters., "PICMG 2.16 CompactPCI/Packet Switching Backplane Specification" TechFocus,
June 2001, pp. 92-93.
186