Sei sulla pagina 1di 8

ENGINEERING REPORTS

A Transistor Ladder Voltage-Controlled Filter


Implemented on a Field Programmable
Analog Array

STEPHEN H. NEASE , AARON D. LANTERMAN , AND JENNIFER O. HASLER


(stephen.nease@gatech.edu) (lanterma@ece.gatech.edu) (jennifer.hasler@ece.gatech.edu)

School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332

Before fast, inexpensive digital electronics became commonplace, analog techniques domi-
nated the field of music synthesis. Over time, much of the art has moved to the digital domain,
which allows for greater flexibility than traditional analog systems. Many enthusiasts still ar-
gue that the warmth of analog systems is difficult to recreate digitally. Recent developments
concerning Field-Programmable Analog Arrays may permit the creation of analog synthe-
sizers that enjoy the ease of rapid reconfigurability previously associated only with digital
synthesizers, allowing musicians to characterize individual circuit elements and try different
topologies. This paper presents an implementation of a particularly famous voltage-controlled
filter design, namely, the transistor ladder filter. The original formulation, which was based on
BJTs, readily maps to MOSFETs operating in the subthreshold region. The implementation is
aided by the use of high-level software design tools in Simulink.

0 INTRODUCTION by Robert Moog [1], originally implemented as a ladder


consisting of discrete bipolar junction transistors (BJTs),
Music production is one of the few areas of technology naturally maps to an FPAA with metal-oxide semiconduc-
where analog signal generation and processing techniques tor field-effect transistors (MOSFETs) operating in the sub-
continue to capture the imagination of the public. This is in threshold region.
sharp contrast with consumer products such as cell phones
and televisions, where most users would not be surprised
at finding most processing being conducted using digital 0.1 Music Synthesis with Analog Circuits
techniques, to the extent that end users think about such Almost all music synthesizers developed in the 1960s
matters at all. Companies such as Korg, Moog, and Nova- and 1970s consisted primarily of analog circuitry [2]. Even
tion continue to produce analog music synthesizers. Electric when microprocessors became commonplace in synthe-
guitarists help keep vacuum tubes in production, and some sizers in the 1980s, the audio generation and processing
mixing console manufacturers continue to use discrete op remained analog in instruments such as the seminal Se-
amps instead of their more cost effective integrated circuit quential Circuits Prophet 5, with the processor reading
(IC) counterparts. The allure of vintage audio gear has front panel controls and using Digital-to-Analog Convert-
lead to computer software plug-ins that attempt to imi- ers (DACs) to generate control voltages for the analog cir-
tate classic pieces of analog equipment with digital signal cuitry. A typical subtractive synthesis chain consisted of a
processing, often deliberately including the nonlinearities voltage-controlled oscillator (VCO), a VCF, and a voltage-
that the original designers generally strove to remove. Such controlled amplifier (VCA), with pitch, timbre, and volume
software emulations have improved over the years, but de- control voltages generated by slowly varying envelopes and
mand for true analog sound remainsand the price of low-frequency oscillators. Some synthesizers that boasted
this sound remains high, due to the large number of dis- digital oscillators, such as the Palm Products GmbH Wave
crete components typically involved. 2.3 and the Sequential Circuits Prophet VS, retained ana-
Modern field programmable analog arrays (FPAAs) of- log filters and amplifiers. These instruments arose in an
fer the potential of implementing analog designs that pro- era in which oscillators could be readily constructed with
vide musicians with greater flexibility at reduced cost. As digital counters driving look-up tables of single-cycle wave-
an illustration, this paper demonstrates how the lowpass forms, whereas the digital multipliers needed to implement
voltage-controlled filter (VCF) developed and popularized real-time digital filters were cost prohibitive. Whether the

J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September 611


NEASE ET AL. ENGINEERING REPORTS

The FPAA explored in this paper is organized into three


functional blocks. The first is the Computational Ana-
log Block (CAB), which is a physical grouping of ana-
log circuits that act as computational elements. These el-
ements include n and p channel field-effect transistors
(nFETs and pFETs), operational transconductance ampli-
fiers (OTAs), capacitors, Gilbert multipliers, and more. A
simple schematic depiction of the connectivity of CABs is
shown on the right side of Fig. 2 a. These CAB compo-
nents can be connected together to form more complicated
subcircuits, which can be further interconnected to create
arbitrary analog systems. This paper illustrates the imple-
mentation of a VCF on such an FPAA.
The interconnection of CAB components is accom-
plished with the FPAAs second functional block, the switch
matrix (SM). A SM and its connection to a CAB is shown
in Fig. 2 a. A SM is a collection of floating-gate pFETs that
connect rows and columns of routing lines. A floating-gate
Fig. 1. Two methods of analog music synthesis. Currently, many FET is a transistor whose gate has no DC path to ground.
discrete components must be obtained and soldered together on The SMs are arranged such that they allow local routing
a printed circuit board (PCB). This process is costly and time- between elements inside a single CAB as well as global
consuming. The Field Programmable Analog Array, however,
offers an analog platform that can be quickly and easily re- routing between CABs. A depiction of the global routing
programmed to connect many different analog circuit elements scheme is shown in the overall system block diagram in
for fast prototyping. The elements are controllable at a single- Fig. 2 b.
transistor level or a large circuit level, enabling fine-grained con- The third functional block is the programmer (also de-
trol over the desired sound. picted in Fig. 2 b), which selects a floating-gate device in
the switch matrix and controls the processes of tunneling
and injection to add or remove charge to the floating gate.
oscillator itself is primarily analog or digital, the distinc- This allows each device to be turned completely on, turned
tive essence of many classic synthesizers lies in their VCF completely off, or operated somewhere in-between. This
implementations. flexibility means that switch elements can be used for com-
The affordability of polyphonic analog instruments in putation as well as routing. This is especially beneficial in
the late 1970s and early 1980s depended on the availability audio applications, since transistors set to a constant bias
of custom music synthesis ICs. Although many of these are often necessary. A die photo of the fabricated chip is
were fairly simple for instance, the SSM2040 consists of shown in Fig. 2 c.
just four operational transconductance amplifiers (OTAs),
four voltage buffers, and an exponential current-to-voltage
0.3 FPAA Test Platform and High-Level Tools
converter these ICs significantly simplified the design
and lowered the cost of such instruments. Unfortunately, All measurements in this paper were taken on a PCB
most of these dedicated music ICs have been long out of designed to test various circuits compiled on the FPAA [3].
production. This system is depicted schematically in Fig. 2 b. The board
Although the idea of field-programmable analog arrays consists of DACs, analog to digital converters, and several
has been around for some time, it is only in recent years other peripherals, which are controlled by an Atmel micro-
that they have become powerful enough to be considered controller. It also contains audio jacks for communicating
for facilitating complex analog music synthesis. As shown audio between standard analog audio sources. The Atmel
in Fig. 1, a single modern FPAA chip could potentially chip receives commands from a PCB running MATLAB
replace dozens of low-scale integration analog ICs from the via a USB connection.
1980s, and their rapid reprogrammability could potentially Over the years, our lab has developed tools that make
provide musicians with architectural flexibility along the the design of circuits on FPAAs easier. Programming of
lines of the original modular synthesizers pioneered by the FPAA is controlled through a suite of MATLAB soft-
Robert Moog and Donald Buchla in the 1960s. ware that interfaces to the microcontroller. We developed
a Simulink frontend to the programming scripts called
Sim2Spice [4]. Users first create a Simulink block diagram
0.2 Field-Programmable Analog Arrays containing components of the FPAA connected together to
Our implementations of FPAAs are mixed-signal Com- create the desired circuit. Sim2Spice then parses the dia-
plementary Metal-Oxide Semiconductor (CMOS) chips gram and generates a netlist that the programming scripts
that allow analog components to be connected together in can target to the FPAA.
an arbitrary fashion, allowing for rapid testing and mea- One advantage of Sim2Spice is that users can create their
surement of many different circuit designs. own Simulink blocks that represent higher-level collections

612 J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September


ENGINEERING REPORTS A TRANSISTOR LADDER VOLTAGE-CONTROLLED FILTER IMPLEMENTED ON A FIELD PROGRAMMABLE ANALOG ARRAY

(a) CAB and SM detail (b) Basic block diagram (c) Die Photo

Fig. 2. (a) Basic block diagram of experimental setup. The entire system is controlled by MATLAB running on a PC. This communicates
with a microcontroller which sends commands to the devices on the board. The FPAA consists of Computational Analog Blocks (CABs)
interconnected by local and global Switch Matrices (SMs). (b) Detail of a portion of a Computational Analog Block. (c) Die photo of
the RASP 2.9a.

of circuits. These blocks can also be paramaterized so that The original transistor ladder filter exploited the dynamic
users can change aspects of the circuit from the higher-level resistance of BJTs; we similarly exploit the dynamic resis-
interface. tance of MOSFETs, with the Vcutoff controlling the bias cur-
rent produced by the topmost transistor in Fig. 3. This bias
current varies the dynamic resistance of the MOSFETs and
1 TRANSISTOR LADDER VCF hence the effective cutoff frequency of the filter stages, just
IMPLEMENTATION as in the original BJT-based filter. Most implementations
This study used one of our most recent FPAA designs, of the transistor ladder filter, both commercial and hobby-
RASP 2.9a. It is a larger version of RASP 2.8b [5], and ist, have used NPN transistors; implementations using PNP
closely related to RASP 2.9v [6]. transistors exist, but are relatively rare. Hence, our ladder
A conceptual schematic of our FPAA implementation schematic may look upside down relative to the transis-
of the transistor ladder filter is shown in Fig. 3, and its tor ladder schematics most enthusiasts are accustomed to
Simulink implementation is shown in Fig. 4. The original seeing.
circuit was implemented with bipolar circuit technology, We used a simple two-OTA attenuator at the input to
so it had to be retooled for CMOS technology. nFETs are reduce the audio signal to a level such that the filter acts
most naturally analogous to NPN BJTs, and pFETs are most approximately linearly.1 Our experiments used a similar
naturally analogous to PNP BJTs. two-OTA amplifier at the output to compensate for this
We chose to implement the ladder using CAB pFETs. initial attenuation.
We decided to use CAB elements (rather than floating-gate The transfer function for this topology has been derived
elements) because they allow for a rapid exploration of the in detail elsewhere [7]. In that analysis, BJTs were used
parameter space in a prototyping context. In future versions rather than MOSFETs. Fortunately, MOSFETs acting in the
of the ladder, the CAB pFETs can be replaced with floating- subthreshold2 regime have exponential current-to-voltage
gate pFETs. The benefit of floating-gate FETs is that their (I-V) characteristics, similar to BJTs. For a pFET in satura-
biases are stored in a nonvolatile manner through floating- tion with terminals referenced to the bulk and drain effects
gate programming, so no explicit gate potentials need to be ignored, this characteristic simplifies to
applied from external circuits. This enables a wide variety I = I0 e(Vs Vg )/UT , (1)
of different biasing conditions across a single chip. The
downside in this context is that developing the calibration
procedure for a ladder based on floating-gate elements can
1
take longer to accomplish. Since we were testing topology Pushing the input level to a point that the filter slightly distorts
the signal is not necessarily bad, as subtle nonlinear effects are
and not calibration, CAB pFETs were deemed sufficient
commonly thought to contribute to the fatness of the transistor
for our purposes. We chose CAB pFETs (rather than CAB
ladder filter sound.
nFETs) because only pFETs are present in floating-gate 2
MOSFETs are most commonly thought of as having square-
form on our target FPAA, and we would like to use floating- law, instead of exponential, I-V characteristics. Unfortunately,
gate devices in this circuit eventually. The floating-gate some undergraduate electronics textbooks mischaracterize the
implementation and calibration procedure are left as future subthreshold regime, which is so useful in many low-power IC
work. designs, as simply being off.

J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September 613


NEASE ET AL. ENGINEERING REPORTS

Fig. 3. Conceptual schematic of an implementation of ladder filter tailored for pFETs operating in the subthreshold regime. An input is
fed into the left side of the filter, and it passes through four filter stages. A differential output is taken between the left and right side of
the ladder with a negative feedback gain of K. In our implementation, feedback was achieved using a two-OTA voltage amplifier. We
also attenuated the signal before applying to the input to avoid distortion, and then amplified the output.

where I0 is a grouping of physical constants, Vs is the source inducing self-oscillation at k = 4. The ability of the filter
voltage, Vg is the gate voltage, is a physical constant to self-oscillate in a reasonably stable fashion can be cre-
representing the gate to channel coupling, and UT is the atively useful; various nonlinear and nonideal effects may
thermal voltage. Typical values for are around 0.7. cause practice to diverge from theory in interesting ways.
The transfer function is developed as in [7] by small- In this implementation, the amount of feedback is con-
signal analysis of these devices. The input pair converts an trolled by a second two-OTA amplifier hooked to the bot-
input voltage into current through its transconductance tom of the ladder in Fig. 4. The first OTA converts the
differential output voltage into an output current through
I
gm = = I, (2) its transconductance gm1 , and the second OTA converts
Vg UT that current into a single-ended voltage via its transcon-
where I is the bias current. The ladders gates are at fixed ductance gm2 . The transconductance of an OTA operated in
potentials, but their source voltages are changing as the subthreshold is defined as [9]:
input changes, so the source conductance is modeled as:
Ibias
I 1 gm = . (5)
gs = = I. (3) 2UT
Vs UT
Developing as in [7], the closed-loop transfer function is Writing Kirchhoffs Current Law at the node that they
share, the DC transfer function is
H (s) =  4 . (4)
2sC
+ 1 + k
gs Ibias1  
Vout = Vcm + Vleft Vright . (6)
Changing the bias current will modulate gs and therefore Ibias2
change the cutoff frequency.
The k factor determines the amount of negative feedback The effect of increasing the amount of negative feedback
from the output of the four stages back to the input. Each in this FPAA implementation is illustrated in Fig. 5. Note
stage approximately buffers the next, so when k = 0, the the peaking resonance behavior, as well as the changing DC
response is that of a cascade of four identical single-pole gain.
sections. Each section provides a 45 degree phase shift at Fig. 6 shows example time-domain system outputs for
its half-power cutoff frequency, for a total phase shift of three cases of square wave inputs: k = 1, 10 kHz input (top
180 degrees. Moogs insight [1] was that negative feedback graph), k = 4, 10 kHz input (middle graph), and k = 4,
could result in a musically useful enhancement around the 100 Hz (bottom graph). Each waveform on a graph corre-
cutoff frequency, with the DC gain decreasing with increas- sponds to a different value of Vcutoff . The DC components
ing k. A root locus analysis [8] shows that for k = 0, the of the waveforms are shifted to provide a convenient way
four poles lie at the same place on the real axis; as k is to show the outputs for several different cutoff frequencies
increased, the poles run along diagonal paths, with the two on the same graph. Varying resonance effects can be seen
rightmost poles hitting the imaginary axis and theoretically in the Gibbs-like oscillations at input transitions.

614 J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September


ENGINEERING REPORTS A TRANSISTOR LADDER VOLTAGE-CONTROLLED FILTER IMPLEMENTED ON A FIELD PROGRAMMABLE ANALOG ARRAY

Fig. 4. Simulink model used to take the data presented in this


paper. CAB pFETs and OTAs are used for this implementation.
For the k amplifier, the differential V-I OTAs bias current was
swept between 250 nA and 4 A. The I-V OTAs bias current was
left at 1A. The A and 1/A OTAs had biases of 200 nA and
1 A, depending on whether the gain was more or less than one.
The output buffers are biased at 5 or 10 A. Ladder capacitance
values are approximately 2 pF. The FG elements are effectively
short-circuits, used to circumvent the fact that Simulink does not
allow two block outputs to be connected together. Physically, they
are compiled to fully-on floating-gate routing elements.

5
k=4
k=2
0
k=1 Fig. 6. Examples of time-domain outputs. The trace colors in each
k=0.5 graph correspond to different values of Vcutoff , the voltage which
k=0.25
sets the circuits cutoff frequency. Vcutoff was swept between 1.8
5 and 2.2V. The y axis gives the output voltage. The amplitude of all
waveforms is correct, but the voltage offsets have been modified
Gain (dB)

for display purposes. The different graphs correspond to different


10
feedback amounts k and different input frequencies as indicated
in the graph titles.
15

2.9a could support 12 pFET-based ladder VCFs, assuming


20
perfect utilization of all the available CAB pFETs in the
chip; experience suggests that between 8 and 10 may be
25
10
1
10
2
10
3 4
10
5
10 more realistic due to imperfect utilization. However, the
Frequency (Hz) chip also contains numerous nFETs, so one could invert the
topology (making it look more like a traditional NPN-based
Fig. 5. Measured magnitude frequency response for different
amounts of negative feedback k. ladder filter), yielding another 810 filters.
The RASP 2.9a CABs also contain MITEs (Multiple-
Input Translinear Element Networks), which are multiple-
input floating-gate pFETs; these could be potentially used
1.1 How Many VCFs Would Fit? to facilitate additional VCFs. Even more VCFs might be
The question of how many such VCFs could fit on any implemented if the floating-gate routing switches could be
given FPAA platform is complicated. Of course, it would pressed into double duty as legs of the ladder, although
depend on what other kinds of components one might want that would likely be the most challenging approach.
to put on any given FPAA, but it would also depend on how In any case, even relatively straightforward implementa-
much effort could be put into making clever use of FPAA tions could readily support the VCF needs of a polyphonic
resources and developing efficient routing schemes. RASP synthesizer with a single FPAA chip.

J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September 615


NEASE ET AL. ENGINEERING REPORTS

1.2 Differential amplifier implementation issues most well-known VCF structures. Other VCF topologies
The gain of the two two-OTA3 amplifiers used in the feed- could be readily explored as well. For instance, another
back path and the final output of the filter is controlled by common structure with a cascade of four single-pole filters
floating-gate transistors, whose value must be programmed; with a negative feedback loop uses OTA-C filters for the
this is inconvenient if smooth real-time changes are desired. stages.
The built in OTAs on RASP 2.9a have their gain con- While this paper focuses on the VCF subsystem of a
trols hardwired to these floating gates. It would be more synthesizer, FPAAs are capable of creating entire synthesis
convent to directly control both the feedback amount and systems. The other two subsystems needed for a simple
the differential amplifier driving the output; the latter then synthesizer are the voltage-controlled oscillator (VCO) and
plays the subtractive synthesis role of a VCA. One could voltage-controlled amplifier (VCA), and both have already
construct OTAs using the stand-alone generic pFETs and been demonstrated on FPAAs. A VCO was implemented
nFETs that are available in each CAB (three of one type of as part of a sigma-delta modulator [13], and a VCA based
FET and two of another type could be made equivalent to on a Gilbert Multiplier was demonstrated on one of the
one of the dedicated OTAs). Alternatively, one could use first FPAAs [5]. Such subsystems will likely need to be
the Gilbert multipliers that are available in each CAB; these modified to conform to audio synthesis specifications. For
may seem like overkill, since they are four-quadrant multi- instance, VCO implementations that would be controllable
pliers and only two-quadrant multiplication is needed, but over a wide pitch range and remain stable with temperature
they may be an attractive option if they are not being used changes might pose a significant challenge.
for other purposes. These issues are left as directions for These subsystems could easily be connected together
future work. by creating a Simulink block for each subsystem and then
connecting them all together in a single block diagram.
A set of interface specifications for Simulink blocks in
2 CONCLUSIONS AND FUTURE WORK
an analog macromodeling context has already been pro-
New FPAAs based on floating-gate transistors offer a posed [14]. The specification makes three recommenda-
flexible platform for the exploration of analog music syn- tions for Simulink blocks. First, blocks should have voltage
thesis. These approaches could empower an analog music inputs and outputs (current-mode computation should be
synthesis renaissance due to the advent of large-scale, re- done within each block). Second, blocks should make use
configurable analog chips. As demonstrated in numerous of vectorized notation, so a single block can represent a
applications [10]-[12], FPAAs have the benefit that they large number of parallel blocks with different properties.
can deliver the functionality of custom integrated analog Finally, analog biasing should be done within each block.
CMOS systems, but they also have the flexibility of FP- If all three of these recommendations are followed, cre-
GAs. An FPAA is a reconfigurable very large-scale inte- ating various synthesis systems with different configura-
grated (VLSI) system, so its impact and design are the same tions of VCOs, VCAs, and VCFs will be relatively easy to
as a custom IC, but it allows a range of potential circuits to accomplish.
be compiled on a single device. Since our VCF is imple- Any practically useful implementation of such
mented on an analog VLSI substrate, its design required the systems including the VCF implementation detailed in
same considerations as those in custom IC design. For ex- this paper would require the development of automatic
ample, the bias currents of the two-OTA amplifiers needed calibration procedures. For example, producers should be
to be high enough such that the k gain factor does not able to specify the VCOs cutoff frequency fc rather than the
significantly degrade over frequency. Additionally, we re- voltage Vc which sets that frequency. The calibration proce-
duced the inputs amplitude and shifted its bias point so dure consists of three steps. First, an analog parameter (Vc )
that it was within the linear regime of the ladder and kept is swept over its valid range, and the systems behavior (fc )
the ladder FETs in saturation. The classic tradeoff between is measured. Next, a lookup table from parameter to behav-
power and performance, the consideration of linearity, and ior is stored. Finally, the Simulink block sets the parameter
the accounting for transistor operating regimes are all ubiq- based on the desired behavior. This could be automated by
uitous themes in both custom IC designs and FPAA system writing scripts that automatically sweep parameters, mea-
designs, and therefore circuit design intuition is essential sure behavior, and store characteristics. Simulink blocks are
when using either platform. capable of using the lookup tables to change their parame-
This paper focused on VCFs because they are one of the ters, so music producers do not need to be familiar with the
most characteristic aspects of analog music synthesis, and it internals of the block they simply need to specify what
focused on the transistor ladder filter because it is one of the behavior they want. Combining this automatic calibration
procedure with the Simulink block specification could yield
3 highly flexible and usable analog synthesis systems. More
The construction two two-OTA amplifiers looks like an error
technically inclined users could explore and modify the in-
but is correct. Each amplifier uses two OTAs; the first turns voltage
into a current, and the second essentially acts like a resistor turning ternal structures of the blocks as desired, to whatever level
current back into a voltage. The total gain of the two-OTA structure of detail interests them.
is determined by the ratio of the gains of the individual OTAs; Implementations need not necessarily employ Simulink,
having both gains available allows designers to make complex which is an expensive program; customized software with
performance tradeoffs. a look-and-feel along the lines of DSP-oriented modular

616 J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September


ENGINEERING REPORTS A TRANSISTOR LADDER VOLTAGE-CONTROLLED FILTER IMPLEMENTED ON A FIELD PROGRAMMABLE ANALOG ARRAY

synthesis programs such as Max/MSP, Pure Data, National [7] Timothy E. Stinchombe. Analysis of the moog tran-
Instruments Reaktor, or the Nord Modular patch editor sistor ladder and derivative filters. http://www.timstinch
might be more appropriate. combe.co.uk/synth/Moog ladder tf.pdf, 2008.
[8] T. Stilson and J. Smith. Analyzing the moog vcf with
considerations for digital implementation. In Proceedings
3 REFERENCES of the International Computer Music Conference, Hong
[1] R.A. Moog. Electronic, high-pass, and low-pass fil- Kong, 1996.
ters employing the base to emitter diode resistance of bipo- [9] C. Mead. Analog VLSI and neural systems. Addison-
lar transistors, 1969. Wesley, 1989.
[2] Robert A. Moog. Voltage controlled electronic music [10] Scott Koziol, Paul Hasler, and Mike Stilman. Robot
modules. J. Audio Eng. Soc, 13(3):200206, 1965. path planning using field programmable analog arrays. In
[3] Scott Koziol, Craig Schlottmann, Arindam Basu, Robotics and Automation (ICRA), 2012 IEEE International
Stephen Brink, Csaba Petre, Brian Degnan, Shubha Ra- Conference on, pages 17471752. IEEE, 2012.
makrishnan, Paul Hasler, and Aurele Balavoine. Hardware [11] Shubha Ramakrishnan, Arindam Basu, Leung Kin
and software infrastructure for a family of floating-gate Chiu, Paul Hasler, David Anderson, and Stephen Brink.
based fpaas. In Circuits and Systems (ISCAS), Proceedings Speech processing on a reconfigurable analog platform. In
of 2010 IEEE International Symposium on, pages 2794 Subthreshold Microelectronics Conference (SubVT), 2012
2797. IEEE, 2010. IEEE, pages 13. IEEE, 2012.
[4] C.R. Schlottmann, C. Petre, and P. E. Hasler. A high- [12] Stephen Nease, Suma George, Paul Hasler, Scott
level simulink-based tool for fpaa configuration. Very Large Koziol, and Stephen Brink. Modeling and implementation
Scale Integration (VLSI) Systems, IEEE Transactions on, of voltage-mode cmos dendrites on a reconfigurable analog
(99):1018, 2010. platform. Biomedical Circuits and Systems, IEEE Transac-
[5] A. Basu, S. Brink, C. Schlottmann, S. Ramakrishnan, tions on, 6(1):7684, 2012.
C. Petre, S. Koziol, F. Baskaya, C.M. Twigg, and P. Hasler. [13] Richard B Wunderlich, Farhan Adil, and Paul
A floating-gate-based field-programmable analog array. Hasler. Floating gate-based field programmable mixed-
Solid-State Circuits, IEEE Journal of, 45(9):17811794, signal array. Very Large Scale Integration (VLSI)
2010. Systems, IEEE Transactions on, 21(8):14961505,
[6] C.R. Schlottmann, S. Shapero, S. Nease, and P. 2013.
Hasler. A digitally enhanced dynamically reconfigurable [14] C.R. Schlottmann and J. Hasler. Highlevel model-
analog platform for low-power signal processing. Solid- ing of analog computational elements for signal processing
State Circuits, IEEE Journal of, 47(9):21742184, Sept. applications. Very Large Scale Integration (VLSI) Systems,
2012. IEEE Transactions on, PP(99):11, 2013.

J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September 617


NEASE ET AL. ENGINEERING REPORTS

THE AUTHORS

Stephen H. Nease Aaron D. Lanterman Jennifer O. Hasler

Stephen Nease received the B.S. degree in electrical en- Engineering Richard M. Bass/Eta Kappa Nu Outstanding
gineering from Rose-Hulman Institute of Technology, Junior Teacher Award in 2006, and was named a 2009 Hes-
Terre Haute, IN, in 2009. He received the M.S. and Ph.D. burgh Award Teaching Fellow. He has served as associate
degrees from the Georgia Institute of Technology, Atlanta, editor for radar for the IEEE Transactions on Aerospace
GA in 2011 and 2014, respectively. He is currently a post-
doctoral scholar at CITEC in the University of Bielefeld.
and Electronic Systems.
r
His interests include neuromorphic circuits and systems,
analog circuit design, and reconfigurable mixed-signal Jennifer Hasler is a Professor in the School of Electrical
CMOS systems. and Computer Engineering at Georgia Institute of Technol-
r ogy. Dr. Hasler received her M.S. and B.S.E. in Electrical
Engineering from Arizona State University in 1991, and
received her Ph.D. from California Institute of Technology
Dr. Lanterman grew up in St. Louis, MO and attended in Computation and Neural Systems in 1997. Her current
Washington University, where he finished a triple major research interests include low-power electronics, mixed-
consisting of a B.A. in music, B.S. in computer science, signal system ICs, floating-gate MOS transistors, adaptive
and a B.S. in electrical engineering in 1993. He stayed information processing systems, smart interfaces for sen-
on for graduate school, receiving an M.S. (1995) and D.Sc. sors, cooperative analog-digital signal processing, device
(1998) in electrical engineering. His graduate work focused physics related to submicron devices or floating-gate de-
on target recognition for infrared imagery as part of the vices, and analog VLSI models of on-chip learning and
U.S. Army Center for Imaging Science. After graduation, sensory processing in neurobiology. Dr. Hasler received the
he joined the Coordinated Science Laboratory at the Univ. NSF CAREER Award in 2001, and the ONR YIP award
of Illinois at Urbana-Champaign, where he managed a large in 2002. Dr. Hasler received the Paul Raphorst Best Paper
project on covert radar systems which exploit illuminators Award from IEEE Electron Devices Society in 1997, the
of opportunity such as commercial television and FM ra- CICC Best Student Paper Award in 2006, the ISCAS Best
dio signals. He joined the Georgia Tech faculty in the fall of Sensors Paper award in 2005, a best paper award at SCI
2001. He received the School of Electrical and Computer 2001.

618 J. Audio Eng. Soc., Vol. 62, No. 9, 2014 September

Potrebbero piacerti anche