1
Abstract - The basic building blocks for Resonant Tunnel-
ling Diode (RTD) logic circuits are Threshold Gates (TGs) instead
of the conventional Boolean gates (AND, OR, NAND, NOR) due to
the fact that, when designing with RTDs, threshold gates can be
implemented as efficiently as conventional ones, but realize more
complex functions. Recently, RTD structures implementing Multi-
Threshold Threshold Gates (MTTGs) have been proposed which
further increase the functionality of the original TGs while main-
taining their operating principle and allowing also the implemen-
tation of nanopipelining at the gate level. This paper describes the
design of n-bit adders using these MTTGs. A comparison with a
design based on TGs is carried out showing advantages in terms of
latency, device counts and power consumption.
Index Terms - Resonant Tunneling Diodes, MOBILE, Multi-
threshold Threshold gate, nanopipelining.
I. INTRODUCTION & BACKGROUND
Resonant tunnelling diodes (RTDs) are very fast non linear
circuit elements which exhibit a negative differential resis-
tance (NDR) region in their current-voltage characteristics
(see Figure 1a) which can be exploited to significantly
increase the functionality implemented by a single gate in
comparison to conventional technologies, thus reducing cir-
cuit complexity [1]. Circuit applications of RTDs are mainly
based on the MOnostable-BIstable Logic Element
(MOBILE) [2], [3]. The MOBILE, shown in Fig. 1a, is a ris-
ing edge triggered current controlled gate which consists of
two RTDs connected in series and driven by a switching bias
voltage . When is low both RTDs are in the on-
state (or low resistance state) and the circuit is monostable.
Increasing to an appropriate maximum value ensures
that only the device with the lowest peak current switches
(quenches) from the on-state to the off-state (the high resist-
ance state). Output is high if the driver RTD is the one which
switches and it is low if it is the load. Peak currents are pro-
portional to RTDs areas. Assuming equal current densities,
the smallest RTD is the one which switches. Logic function-
ality is achieved by embedding an input stage which modifies
the peak current of one of the RTDs. An input stage com-
posed of the series connection of an RTD and a FET transis-
tor has been proposed [4]. This is the idea of the inverter
MOBILE shown in Fig. 1b. Only when a logic one is applied
to the gate of a transistor its associated RTD contributes to
the NDR
2
current. RTDs areas are selected such that the rela-
tion of the peak currents of NDR
1
and NDR
0
depends on
whether the external input signal is ?1? or ?0?. Transistor
are sized such that RTDs are the current limiting devices in the
series connection of the RTD and the transistor. For
high the output node maintains its value even if the input
changes. That is, this circuit structure is self-latching allow-
ing to implement pipeline at the gate level without any area
overhead associated to the addition of the latches which
allows very high throughtoutput.
1.This effort was partially supported by the Spanish Government
under project TEC2004-02948/MIC.
V
bias
V
bias
V
bias
RTD
1
RTD
2
V
out
-1 -0.5 0.5 1
-2
-1
1
2
V
RTD
(V)
I
RTD
(mA)
peak
current
current
valley
peak
voltage
load
driver
NDR
1
NDR
2
x
1
V
bias
V
out
w1
A
1
A
2
A
21
Figure 1. MOBILE circuits, a) basic MOBILE,
b) MOBILE inverter.
(a)
(b)
V
bias
V
in
V
bias
Using Multi-Threshold Threshold Gates in RTD-based Logic Design.
A Case Study
1
H?ctor Pettenghi, Mar?a J. Avedillo, and Jos? M. Quintana
Instituto de Microelectr?nica de Sevilla, CNM, Sevilla, SPAIN
E-mail: {hector, avedillo, josem}@imse.cnm.es
? TIMA Editions / ENS 2005 ISBN: 2-916187-02-2
Threshold logic [5] is a computational model widely used
in the design of RTD based circuits [4]. That is, the basic
building blocks for RTD logic circuits are Threshold Gates
(TGs) instead of the conventional Boolean gates (AND, OR,
NAND, NOR). This is due to the fact that when designing
with RTDs, threshold gates can be implemented as effi-
ciently, in terms of performance and complexity, as conven-
tional Boolean gates but realize more complex functions.
A TG or linear separable function is defined as a logic
gate with n binary input variables, , one
binary output , and for which there is a set of real
numbers: threshold T and weights , such that
its input-output relationship is defined as iff
, and otherwise. Sum and product are
the conventional, rather than the logical, operations. The set
of weights and threshold can be denoted in a more compact
vector notation way by .
TGs are efficiently realized by resorting to the operation
principle of the clocked series connection of a pair of RTDs
(MOBILE). The functionality of the MOBILE inverter can
be extended to the TG as depicted in Figure 2 showing the
circuit structure for a generic threshold gate of two inputs
, where represents a positive weight and
represent a negative one [4]. The areas and are deter-
minated by the threshold to be implemented, and the areas
and B are selected according to the technology and to the
required design trade-offs involving speed, power and robut-
ness against parameter variations. Threshold gates with a
larger number of inputs can be easily realized extending this
topology.
More recently, RTD structures implementing Multi-
Threshold Threshold Gates (MTTGs) [6] have been proposed
which further increase the functionality of the original TGs
while maintaining their MOBILE operating principle and
associated advantages [7], [8], [9], [10].
MTTGs are a generalization of the conventional TGs. For-
mally, A k-threshold MTTG is a logic element with n binary
input variables, , one binary output , and
for which there is a set of real numbers: thresholds
, , and weights , such that its
input-output relation is defined as: iff
, with , ;
output is equal to zero otherwise. As in the TGs, the set of
weights and thresholds can be denoted in a more compact
vector notation way by .
There are a number of theoretical properties for MTTGs
which suggest they can be extremely powerful in digital design
[6]. In particular, any n-input switching function can be realized
by a single MTTG with at most (2
n
-1) thresholds simply
assigning . Moreover, usually exponential weights
and exponential number of thresholds are not required. Thus,
from a logic point of view, the MTTG is a powerful concept.
This paper describes the design of n-bit adders using RTD-
based MTTGs. The rest of the paper is organized as follows.
Section II describes an adder implemented by means of TGs
and reported by Pacha et al [4]. Section III describes the pro-
posed adder using MTTGs and the realization of tthese more
powerful gates on the basis of the concept of the MOBILE.
Section IV compares three-bit adders designed with each of the
architectures.
II. ADDERS BASED ON TGS
Figure 3a shows the logic diagram of a nanopipelined carry
propagation n-bit adder based on TGs reported in [4]. It is
comprised of the chain of full adders (FAs) and MOBILE fol-
lowers (latches) to support pipeline. The first FA has been sim-
plified. Since the carry output of the FA is a threshold function,
it is implemented by a single threshold gate. The sum output of
a FAs is a three-input EXOR function, which is not linear sep-
arable, and thus require a network of TGs to be realized. It is
implemented using a two stage network with four gates as
depicted in Figure 3a. Figure 3b-e depict the TG based realiza-
tions for the two-input EXOR (Figure 3b), the FA (Figure 3c),
the two-input AND (Figure 3d) and latches (Figure 3e). Bias
signals to operate cascaded MOBILE-type circuits are also
shown (Figure 3f). A four phase (evaluation, hold, reset and
wait) overlapping clocking scheme is used. Note that second
stage (V
bias2
) evaluates while the first stage (V
bias1
) is in the
hold phase. For a number of logic levels greater than three,
four bias signals are required. In one clock period all the gates
are activated. Data can be processed at a frequency determined
by the operation speed of four chained MOBILE gates.
x
i
i 1 ? n,,=(),
yn1+()
w
1
w
2
? w
n
,,,
y 1=
w
i
x
i
i 1=
n
?
T? y 0=
w
1
w
2
? w
n
; T,,,[]
w
k
w
l
; T,[]w
k
w
l
A
1
A
2
A
u
A
1
B+()A
u
V
bias
y
x
l
x
k
w
k
A
u
w
l
A
u
Figure 2.- Circuit structure for . w
k
w
l
; T,[]
A
2
B+()A
u
x
i
i 1 ? n,,=(), y
nk+()
T
i
i 1 ? k,,=() w
1
w
2
? w
n
,,,
y 1=
T
2j 1?
w
i
x
i
T
2j
<
i 1=
n
?
? T
j 1+
T
j
> j 1 ? k 2?,,=()
y
w
1
w
2
? w
n
; T
1
? T
k
,,,,,[]
w
i
2
i
1?=
? TIMA Editions / ENS 2005 ISBN: 2-916187-02-2
III. ADDERS BASED ON MTTGS
Figure 4a depicts the circuit topology we have proposed
for a generic MTTG with positive weights defined by
, i.e., two inputs and two thresholds [8].
The specific function implemented by such a circuit depends
on the areas of the RTDs. The concept of RTD-based
MOBILE is extended to a circuit consisting of three or more
RTDs in series. The switching sequence in series-connected
RTDs begins also with the RTD with the smallest peak cur-
rent [11]. If this peak current can be controlled by external
inputs, this sequence can be varied and different functions
can be obtained at the output nodes.
Figure 4b depicts a more efficient alternative realization of
two input two threshold MTTGs [10]. Circuit in Figure 4a
operates on the basis of current comparison: the NDR with
the lowest peak current switches off. Inputs modulate the
peak currents of both the bottom (NDR
2
) and the middle
(NDR
1
) NDRs, although by a different amount since the
areas of the RTDs of their associated input stages are differ-
ent. Thus, concerning NDR
1
and NDR
2
, and assuming, with-
out loss of generality, A
11
> A
21
(A
12
> A
22
), the same result
of the current comparison could be obtained if only NDR
1
is
modulated with input stages with area A
11
- A
21
(A
12
- A
22
)
for input variable x
1
(x
2
). That is what has been done in Fig-
ure 4b. However, area relationships of NDR
1
and NDR
2
with
NDR
0
must be kept for circuit in Figure 4b to implement the
same functionality that circuit in Figure 4a. To achieve this,
input stages of area equal to A
21
(A
22
) for x
1
(x
2
) common to
both have been added to the proposed circuit. Note that the
area of some RTDs is reduced with respect to the solution in
Figure 4a, thus allowing to reduce the width of their associ-
ated transistors.
w
1
w
2
; T
1
T
2
,,[]
&
exor
[111;2]
S
0
A
0
B
0
A
1
B
1
C
2 C
n-1 C
n
A
n-1
B
n-1
[111;1]
[111;3]
[1-11;1]
S
1
FA
S
n-1
FA
C
1
V
bias1
x
2
x
3
x
1
V
bias1
x
2
x
3
x
1
V
bias1
x
2
x
3
x
1
V
bias2
sum
[1, 1, 1;2]
[1, 1, 1;1]
[1, 1, 1;3]
[1, 1, -1;1]
(c)
V
bias1
x
1
x
2
V
bias1
x
1
x
2
V
bias2
y
[1, 1;1]
[1, 1;2]
[1, -1;1]
V
bias1
x
1
x
2 y
Figure 3. n-bit adder based on TGs. (a) logic diagram. (b) RTD-based realization of two-input EXOR. (c) RTD-based realiza-
tion of FA. (d) RTD-based realization of two-input AND. (e) Latch. (f) Bias scheme for MOBILE circuits.
evaluation hold reset wait
evaluation hold reset wait
V
bias1
V
bias2
VBIAS
V
bias1
x
2 y
(a)
(b)
(d) (e)
(f)
latch
carry
? TIMA Editions / ENS 2005 ISBN: 2-916187-02-2
Figure 5a depicts the logic diagram of the proposed nan-
opipelining carry-propagation n-bit adder. Note that it
uses modified full adders in which complemented inputs
are processed. Each of them comprises only two logic
gates (only one logic level): the three-input mayority and
the three-input EXNOR. The first one is a TG ([1, 1, 1; 2])
and the second one is an MTTG which can be imple-
mented by a single RTD gate extending the topology
above described to four RTDs in series. The FA modifica-
tion has been done to avoid using a three input EXOR which can
not been implemented with the MTTG topology described.
Additional latches (implemented as MOBILE buffers) and
inverting latches (implemented as MOBILE inverters) are
also required to support pipeline. Due to the fact that the pro-
posed adder needs complemented inputs in the modified FAs, it
is required to use an inverter in the last stage of the input latches.
Note that he circuit showed in Figure 3a requires a two-level
network for the full adder while the architecture in Figure 5a
Figure 4.- Circuit structures for generic two-input two-threshold MTTG with positive weights. w
1
w
2
; T
1
T
2
,,[]
A
12
A
22
A
11
A
21
A
1
+B
A
2
+B
A
0
+B
y
2
V
bias
x
2
x
1
y
1
NDR
0
NDR
1
NDR
2
V
bias
A
21 A22
A
1
+B
A
2
+B
RTD
RTD
1
2
y
1
A
11
A
21
? A
12
A
22
?
A
0
+B
y
2
NDR
0
NDR
1
NDR
2
(a)(b)
&
exor
exor
[111;2]
Modified FA
S
0
A
0
B
0
A
1
B
1
S
1
exor
[111;2]
Modified FA
S
n-1
C
2
C
n-1
C
n
A
n-1
B
n-1
C
1
x
2
x
3
x
1
x
3
x
2
x
1
x
1
x
2
x
3 y
V
bias
x
1
x
2x
2
x
1
y
1
y
2
V
bias
y
x
2
x
3
x
1
V
bias
x
1
y
V
bias
x
1
x
2
Figure 5. n-bit adder based on MTTGs. (a) logic diagram. (b) RTD realization of two-input EXOR. (c) RTD realization of
three-input EXNOR. (d) RTD realization of Majority [111;2]. (e) inverting latch. (f) RTD realization of two-input NAND.
V
bias
latch
inverting latch
(a)
(b)
(c) (d)(e)(f)
? TIMA Editions / ENS 2005 ISBN: 2-916187-02-2
implements the modified full adder with a single level of gates.
This means the latency of the proposed adder is reduced.
IV. COMPARISON
Three-bit adders have been designed to operate up to 750MHz
using each of the architectures in a non commercial but univer-
sity InP technology in which RTD and transistors can be co-
integrated [12]. Minimization of power has been the design tar-
get. Simulation results at the operating frequency of 750MHz
are depicted in Figure 6. Figure 6a shows the results for the pre-
viously reported adder, Figure 6b shows the results for the pro-
posed gate. Note that in this figure the carry is complemented
although it can be obtained without complementation changing
the majority gate [1, 1, 1; 2] of the last stage to [-1, -1, -1; -1]
without increasing of the latency.
For three bit adders operating at 750 MHz the design based
in TGs has 15% more devices, its latency is 33% larger and
consumes 20% more power than the proposed in this work
based on MTTGs.
V. CONCLUSIONS
In this paper a novel nanopipelined n-bit adder based on
RTDs and HFET devices has been described. A comparative
analysis has been carried between this proposed design and an
adder previously reported. The proposed design has several
advantages: lower number of transistors, smaller total area of
RTDs, lower latency and lower power consumption. Nanopipe-
lined architectures for multipliers and divisors recently reported
[13] can take advantage of these advanced proposed adders.
REFERENCES
[1] P. Mazumder, S. Kulkarni, M. Bhattacharya, J.P. Sun
and G.I. Haddad: ?Digital Circuit Applications of Res-
onant Tunneling Devices?, Proceedings of the IEEE,
Vol. 86, no. 4, pp. 664-686, April 1998.
[2] K.J. Chen, K. Maezawa and M. Yamamoto: ?InP-Based
High Performance Monostable-Bistable Transition Logic El-
ements (MOBILEs) Using Integrated Multiple-Input Reso-
nant-Tunneling Devices,? IEEE Electron Device Letters,
Vol. 17, no. 3, pp. 127-129, March 1996.
[3] K. Maezawa and T. Mizutani ?A New Resonant Tun-
neling Logic-Gate Employing monostable-bistable
transition?, Japan Journal applied Phys. Letters, Vol.
37, pp. 142-144, 1993.
[4] C. Pacha et al.: ?Threshold Logic Circuit Design of
Parallel Adders Using Resonant Tunnelling Devices,?
IEEE Trans. on VLSI Systems, Vol. 8, no. 5, pp. 558-
572, Oct. 2000.
[5] S. Muroga, Threshold Logic and Its Applications, New York:
Wiley, 1971.
[6] D. R. Haring: ?Multi-Threshold Threshold Elements,? IEEE
Trans. on Electronic Computers, Vol. EC-15, No. 1, pp. 45-
65, February 1966.
[7] M.J. Avedillo, J.M. Quintana, and J.L. Huertas, ?Beyond
Threshold Logic: Multi-threshold Threshold Gates,? 11th
MEL-ARI/NID Workshop, Toulouse (France), Feb. 2003.
[8] M.J. Avedillo, J.M: Quintana, H. Pettenghi, P.M. Kelly, and
C.J. Thompson, ?Multi-threshold Threshold Logic Circuit
Design Using Resonant Tunneling Devices,? Electronics
Letters, Vol. 39, pp. 1502-1504, 2003.
[9] H. Pettenghi, M.J. Avedillo, J.M: Quintana, ?Useful Logic
Blocks Based on Clocked Series-Connected RTDs? Proceed-
ings. IEEE Conference on Nanotechnology, pp. 593-595, 2004.
[10] H. Pettenghi, M.J. Avedillo, J.M: Quintana, ?New Circuit To-
pology for Logi Gates Based on RTDs?, Proceedings. IEEE
Conference on Nanotechnology, pp. 437-440, 2004
[11] K.J. Chen, T. Waho, K. Maezawa and M. Yamamoto: ?An
Exclusive-OR Logic Circuit Based on Controlled Quenching
of Series-Connected Negative Differential Resistance De-
vices?, IEEE Electron Device Letters, Vol. 17, no. 6, pp.
309-311, June 1996.
[12] W. Prost et al.: EU IST Report LOCOM no. 28 844 Dec. 2000.
[13] P. Gupta, N. K. Jha: ?An Algorithm for Nano-pipelining of
Circuits and Architectures for a Nanotechnology?, Proceed-
ings Design, Automation and Test in Europe Conference,
Vol. 2 pp. 904-909, 2004.
Figure 6.- Simulations results for three-bit adders. (a) TG based.
(b) MTTG based.
0
0
0
0
0
0
0
0
1
0
1
0
1
0
1
1
1
1
Vbias
0
Vbias
1
Vbias
2
A
0
B
0
A
1
B
1
A
2
B
2
S
0
S
1
S
2
C
3
Vbias
1
Vbias
2
Vbias
3
A
0
B
0
A
1
B
1
A
2
B
2
S
0
S
1
S
2
C
3
Vbias
0
1
0
1
1
1
1
1
0
1
1
1
1
0
0
0
0
0
0
(b)
(a)
? TIMA Editions / ENS 2005 ISBN: 2-916187-02-2