Threshold element and method of designing the same

Information

  • Patent Grant
  • 6338157
  • Patent Number
    6,338,157
  • Date Filed
    Friday, January 29, 1999
    25 years ago
  • Date Issued
    Tuesday, January 8, 2002
    22 years ago
Abstract
A threshold element is implemented by output wired CMOS circuit. The corresponding threshold function is transformed to a form including the ratio of beta values of n- and p-transistors. Inputs are allocated to the n- and p-transistors in accordance with the ratio. The logic circuitry can be implemented with half of transistors in the similar implementation in a conventional CMOS circuitry, since every character entry corresponds to a single transistor.
Description




BACKGROUND OF THE INVENTION




1. Field of the Invention




The present invention relates to a method of designing a threshold element, and to a threshold element designed using such a method. In particular, the invention is adapted to develop a threshold element based on an output wired CMOS (complementary metal oxide semiconductor) logic.




2. Description of the Prior Art




During the last 40 years, the tides of interest to threshold and majority logics rose and waned periodically. This was caused, on the one hand, by new circuit elements of threshold nature and, on the other hand, by the tasks which used threshold functions to specify functional blocks. There are a lot of such tasks, for example, those of reliability, threshold coding, AD/DA-conversion, filtration, etc. Of special note are neural networks, the behavior of formal neutrons in which is described by threshold functions, starting from the classic work by McCulloch and Pitts.




A threshold function is defined as









y
=


Sign


(





i
=
1

n




w
i



x
i



-
T

)


=

{



1






if









i
=
1

n




w
i



x
i




-
T


0










0





if









i
=
1

n




w
i



x
i




-
T

<
0










(
1
)













where w


i


is the weight of the i-th input and T is the threshold. Thus, a threshold element should consist of an adder and a threshold comparator.




Recently, new circuits of CMOS threshold elements have been suggested and studied. A νCMOS is based on the CMOS-inverter with floating gate and capacity inputs. Another CMOS is based on output wired CMOS inverters. The latter are the subject of the present specification.





FIG. 1

illustrates the structure of a threshold element consisting of output wired CMOS inverters. The threshold element comprises inverters


1


,


2


, . . . , n having inputs x


1


, x


2


, . . . , x


n


. Outputs of the inverters


1


,


2


, . . . , n are wired. The output voltage Vout appears at the wired outputs. A decision inverter


10


converts the output voltage Vout into binary data which is the final result


y


. The threshold value of the decision inverter


10


is determined in proper manner to truly binary-code the result of the majority. The inverters


1


,


2


, . . . , n realize an adder, while the decision inverter


10


realizes a comparator, in this circuit. The threshold element allows outputs from the inverters


1


,


2


, . . . , n to cause the tug of war, in which the final result


y


becomes “1” if more binary code “1” is in put and becomes “0” if more binary code “0” is input. Thus, the inverters as a whole establish a majority circuit.




Since threshold functions are a subset of Boolean functions, any threshold function can naturally be represented by a superposition of operations (circuit elements) of any functionally full basis. The question is whether circuits can be built that would implement threshold functions in a simpler way than the traditional circuits do. If the answer is positive, it would be possible to simplify the implementations of arbitrary logic functions having a bigger number of basic elements. In CMOS-implementation, every character entry corresponds to two (p- and n-) transistors.




SUMMARY OF THE INVENTION




It is therefore an object of the present invention to provide a threshold element capable of providing the function blocks with transistors less than two for each of the inputs, and a method of designing the same.




According to the present invention, there is provided a method of designing a threshold element, comprising: converting to the form including a ratio related to p- and n-transistors a threshold function which outputs the result of comparison between a predetermined threshold value and the sum of weighted inputs; and allocating the inputs to the p- and n-transistors in accordance with the ratio, so as to design a threshold element corresponding to the threshold function.




The ratio could be a parameter to determine which of p- and n-transistors more affects on the output when the p- and n-transistors are turned on, since the ratio is related to the p- and n-transistors. In other words, the ratio could be a parameter to determine the condition of the above-mentioned tug of war.




A threshold element of the present invention can be designed using the above-mentioned designing method. When a threshold function is converted to the form including a ratio related to p- and n-transistors, respective inputs are ti allocated to the p- and n-transistors in accordance with the ratio.











BRIEF DESCRIPTION OF THE DRAWINGS




The above and other objects, features and advantages of the present invention will become apparent from the following description of the preferred embodiment in conjunction with the accompanying drawings, wherein:





FIG. 1

illustrates a threshold element comprising conventional output wired CMOS inverters;





FIG. 2

illustrates the background leading to a method of designing a threshold element in accordance with an embodiment of the present invention;





FIG. 3

illustrates a threshold element realizing a certain threshold function in accordance with the theorem which the present inventor has found;





FIG. 4

illustrates the structure of the threshold element corresponding to xyz arithmetic;





FIG. 5

illustrates the structure of the threshold element corresponding to x(y+z) arithmetic;





FIG. 6

illustrates the structure of the threshold element corresponding to x+y+z arithmetic;





FIG. 7

illustrates the structure of the threshold element corresponding to xy{overscore (z)} arithmetic;





FIG. 8

illustrates the structure of the threshold element corresponding to x{overscore (yz)} arithmetic;





FIG. 9

illustrates the structure of the threshold element corresponding to {overscore (xyz)} arithmetic;





FIG. 10

illustrates the structure of the threshold element corresponding to xyz+x{overscore (yz)} arithmetic;





FIG. 11

illustrates the structure of the threshold element corresponding to {overscore (x)}yz+{overscore (xyz)} arithmetic;





FIG. 12

illustrates the structure of the threshold element corresponding to xyz+{overscore (xyz)} arithmetic;





FIG. 13

illustrates the structure of the threshold element corresponding to a two-input C-element;





FIG. 14

illustrates the structure of the threshold element corresponding to a three-input C-element;





FIG. 15

illustrates the structure of the threshold element corresponding to an n-input C-element; and





FIG. 16

illustrates the structure of the threshold element corresponding to a full adder.











DESCRIPTION OF THE PREFERRED EMBODIMENT





FIG. 2

illustrates the background leading to a designing method of the present invention. A p-transistor or p-channel MOSFET


20


and an n-transistor or n-channel MOSFET


21


are serially connected between a power supply and a ground. The gate


22


of the p-transistor


20


is grounded, and the gate


23


of the n-transistor


21


is pulled up to the power supply. In the static mode where both transistors


20


,


21


are completely open, the voltage V at the drains of the p- and n-transistors


20


,


21


is determined by the ratio α=βn/βp where βp and βn are a drain current coefficient or a beta value of p- and n-transistors


20


,


21


respectively. The voltage V drops when α increases and grows when α declines. Assuming that p- and n-transistors take the same characteristic in the respective inverters


1


,


2


, . . . , n in the conventional circuit shown in

FIG. 1

,






α
=





i
=
1

n



x
i






i
=
1

n




x
_

i













can be obtained. The threshold function of the majority circuit shown in

FIG. 1

can simply be expressed by








y


=Sign(α−1)  (2)






because










Sign


(

α
-
1

)


=

Sign


(





i
=
1

n



x
i


-




i
=
1

n




x
_

i



)








=

Sign


(





i
=
1

n



x
i


-




i
=
1

n



(

1
-

x
i


)



)








=

Sign


(





i
=
1

n



x
i


-

n
2


)















can be established. The ratio α is the principle to a threshold element of the present invention.




We will explain below the theorem which the present inventor has found to reduce the number of transistors implementing a threshold circuit.




Theorem




Any threshold function of expression (1) can be represented as follows so as to include a ratio,










F


(
X
)


=

Sign


(






i

S





w
i



x
i







i

S






w
_

i



x
i




-
1

)






(
3
)













where S is a certain subset of indexes


i


, such that






Σ


iεS




w




i




=T


  (4)






Proof




Since x


i


is either of “0” or “1”, the following expression













i
=
1

n




w
i



x
i



-
T




(
5
)













takes a discrete value in relation to variable combinations of x


i


. Taking a value


a


closest and larger than “0” among the discrete values, a threshold function never loses its function even if T is replaced with (T−a). This operation allows set


S


of the index


i


to meet the expression (4).




If a set


S


of the index


i


meets the expression (4), then the expression (5) may be transformed as follows:














i
=
1

n




w
i



x
i



-
T

=





i

S





w
i



x
i



-

(

T
-




i

S





w
i



x
i




)








=





i

S





w
i



x
i



-




i

S





w
i



(

1
-

x
i


)










=





i

S





w
i



x
i



-




i

S





w
i




x
_

i











F


(
X
)


=

Sign


(





i
=
1

n




w
i



x
i



-
T

)








=

Sign
(





i

S





w
i



x
i



-




i

S





w
i




x
_

i




)







=

Sign


(






i

S





w
i



x
i







i

S





w
i




x
_

i




-
1

)















This is the end of the proof. However, the above expression can be rearranged to the following expression.







F


(
X
)


=

1
-

Sign


(






i

S





w
i




x
_

i







i

S





w
i



x
i




-
1

)













where S is a certain subset of indexes


i


, such that










i

S




w
i


=

T
-
1











This rearrangement may be employed in designing a circuit of a three-input C-element as described later.





FIG. 3

illustrates a threshold element based on a certain threshold function of the expression (1). As shown in

FIG. 3

, the threshold element comprises p-transistors


30


,


31


, . . . , 3k for receiving x


i


for the index


i


included in the above-mentioned set


S


defined by the expression (4), and n-transistors


40


,


41


, . . . , 4m for receiving x


i


for the index


i


outside the set


S


. All of the drains from the p- and n-transistors are commonly connected. A decision inverter


50


compares the voltage Vout at the connected drains with a threshold value T so as to output the final result


y


.




Taking account of the ratio α in the expression (2), the numerator of the fractional component in the expression (3) is considered to correspond to the n-transistors, and the denominator of the fractional component is likewise to the p-transistors. This tells that a single transistor covers a single input. Accordingly, the threshold element of

FIG. 3

can be constructed with a half of transistors employed in a conventional threshold element of output wired CMOS inverters shown in

FIG. 1

if the decision inverter


50


is neglected.




The utilization of the above-mentioned designing method allows a threshold element of a three-input Boolean calculation such as xyz, x(y+z), xy+yz+zx, x+yz, and x+y+z to be realized with four transistor and a single inverter, namely, six transistors in total.




xyz Arithmetic




A Boolean calculation of xyz can be expressed by a threshold function, F=Sign(x+y+z−3), which may be rearranged to






F
=

Sign


(


ε


x
_

+

y
_

+

z
_



-
1

)












where ε is a transistor of a smaller beta value, which is called a “weak transistor” hereinafter. It is adapted to prevent the floating of the output voltage Vout and brought in for convenient designing. It should be noted that transformation can be effected to the threshold function in variable manners.





FIG. 4

illustrates a threshold element corresponding to xyz implementation. The x-, y- and z-inputs are connected to the bases of p-transistors


100


,


101


,


102


of the same beta value. The drains of p-transistors


100


,


101


,


102


are commonly connected to the drain of an n-transistor


103


corresponding to a weak transistor ε which keeps open. Four transistors


100


-


103


form an NAND-gate


104


. The analog output voltage Vout appears at the connected drains of four transistors


100


-


103


. A decision inverter


105


is disposed at the stage subsequent to the NAND-gate


104


so as to binary-code the output voltage Vout. The whole circuit forms an AND-gate


106


.




x(y+z) Arithmetic




A Boolean calculation of x(y+z) can be expressed by a threshold function, F=Sign(2x+y+z−3), which may be rearranged to the following expression.






F
=

Sign


(



z
+
ε



2






x
_


+

y
_



-
1

)













FIG. 5

illustrates a threshold element corresponding to x(y+z) implementation. The x- and y-inputs are connected to the bases of p-transistors


110


,


111


, while the z-input is connected to the base of an n-transistor


112


. The drains of p- and n-transistors


110


-


112


are commonly connected to the drain of an n-transistor


113


corresponding to the weak transistor ε which keeps open. The beta value of the p-transistor


110


which receives the x-input is set at twice the beta value of the other transistors


111


,


112


so as to reflect the weight of the transistor


110


. The output voltage Vout appears at the connected drains of four transistors


110


-


113


and is supplied to a decision inverter


114


.




xy+yz+zx Arithmetic




This is a three-input majority circuit. A Boolean calculation of xy+yz+zx can be expressed by a threshold function, F=Sign(x+y+z−2), which may be rearranged to the following expression.






F
=

Sign


(



z
+
ε



x
_

+

y
_



-
1

)












The structure of the corresponding threshold element is the same as of FIG.


5


. Here, the beta value of the n-transistor


110


which receives the x-input is set at the same level as the other transistors


111


,


112


.




x+y+z Arithmetic




A Boolean calculation of x+y+z can be expressed by a threshold function, F=Sign(x+y+z−1), which may be rearranged to the following expression.






F
=

Sign


(



y
+
z
+
ε







x
_



-
1

)













FIG. 6

illustrates a threshold element corresponding to x+y+z implementation. The x-input is connected to the base of a p-transistor


120


, while the y- and z-inputs are connected to the bases of n-transistors


121


,


122


. The drains of the p- and n-transistors


120


-


122


are commonly connected to the drain of an n-transistor


123


corresponding to the weak transistor ε which keeps open. Four transistors


120


-


123


form an NOR-gate


124


having analog output. The output voltage Vout appears at the connected drains of four transistors


120


-


123


. A decision inverter


125


is disposed at the stage subsequent to the NOR-gate


124


so as to binary-code the output voltage Vout. The whole circuit forms an OR-gate


126


.




x+yz Arithmetic




A Boolean calculation of x+yz can be expressed by a threshold function, F=Sign(2x+y+z−2), which may be rearranged to the following expression.






F
=

Sign


(



y
+
z
+
ε


2






x
_



-
1

)












The structure of the corresponding threshold element is the same as of FIG.


6


. Here, the beta value of the p-transistor


120


which receives the x-input is set at twice the beta value of the other transistors


121


,


122


.




The description has been made to five examples of Boolean calculation circuits. According to the present invention, a three-input Boolean calculation circuit can be formed with a combination of transistors equal to or less than eleven and inverters equal to or less than two, namely, transistors equal to or less than fifteen in total. A three-input Boolean calculation may be established by a combination of any of four bases such as xyz, xy{overscore (z)}, x{overscore (yz)}, and {overscore (xyz)}, so that the following examples can be implemented.




xyz Arithmetic




As described above, the threshold element of

FIG. 4

corresponds to xyz implementation. Six transistors are used in total. As of {overscore (xyz)}, the threshold function can be expressed by F=1−Sign(x+y+z−3), so that an inverter may be added to the later stage of the AND-gate


106


of FIG.


4


.




xy{overscore (z)} Arithmetic





FIG. 7

illustrates a threshold element corresponding to this logic. This can be implemented with the z-input of

FIG. 4

inverted by an inverter


130


. Eight transistors are used in total. The AND-gate


106


in

FIG. 4

is also used in FIG.


7


. The threshold function may be expressed by F=Sign(x+y+{overscore (z)}−3).




Since







xy






z
_


_










can be expressed by F=1−Sign(x+y+{overscore (z)}−3), the corresponding threshold element may be implemented by inserting an inverter after the AND-gate


106


in FIG.


7


.




x{overscore (yz)} Arithmetic




Since a Boolean calculation of x{overscore (yz)} is equal to









x
_

+
y
+
z

_

,










the corresponding threshold function may be expressed by F=Sign(x+{overscore (y)}+{overscore (z)}−3)=1−Sign({overscore (x)}+y+z−1).

FIG. 8

illustrates a circuit utilizing the OR-gate


126


in FIG.


6


. When the OR-gate


126


is used, the x-input is inverted by an inverter


131


and the output from the OR-gate


126


is inverted by an inverter


132


. Ten transistors are used in total.




Since







x






y
_







z
_


_










can be expressed by F=1−Sign(x+{overscore (y)}+{overscore (z)}−3), the corresponding threshold element maybe implemented by inserting an inverter after the inverter


132


in FIG.


8


.




{overscore (xyz)} Arithmetic




Since a Boolean calculation of {overscore (xyz)} is equal to








x
+
y
+
z

_

,










the corresponding threshold function may be expressed by F=Sign({overscore (x)}+{overscore (y)}+{overscore (z)}−3)=1−Sign(x+y+z−1).

FIG. 9

illustrates a circuit utilizing the OR-gate


126


in FIG.


6


. When the OR-gate


126


is used, the output from the OR-gate


126


is inverted by an inverter


133


. Eight transistors are used in total.




Since








x
_







y
_







z
_


_










can be expressed by F=1−Sign({overscore (x)}+{overscore (y)}+{overscore (z)}−3), the corresponding threshold element may be implemented by adding an inverter after the inverter


133


in FIG.


9


.




xyz+x{overscore (yz)} Arithmetic





FIG. 10

illustrates a circuit of xyz+x{overscore (yz)} implementation. First, A=x{overscore (yz)} is made.




Since







A
=



x
_

+
y
+
z

_


,










the x-input of the NOR-gate


124


is inverted by an inverter


139


. This is the first stage of the circuit. The threshold function would be: The x-, y- and z-inputs are respectively connected to the drains












F
=

Sign


(

xyz
+
A
-
1

)








=

Sign


(

x
+
y
+
z
+

3

A

-
3

)








=

Sign


(




3

A

+
ε



x
_

+

y
_

+

z
_



-
1

)









(
6
)













of p-transistors


140


,


141


,


142


at the second stage. The logic A is connected to the drain of an n-transistor


143


in parallel with a weak n-transistor


144


which keep open. The beta value of the n-transistor


143


which receives the logic A is set at thrice the beta value of the other transistors


140


-


142


. The drains of five transistors


140


-


144


are commonly connected, and the output voltage Vout appears at the connected drains. A decision inverter


145


is disposed after the drains, so that the whole second stage forms a composite circuit


146


corresponding to the expression (6).




The input A to the composite circuit


146


is an analog signal. The proper operation of the circuit


146


has been confirmed by circuit simulation if the characteristic of transistors are appropriately managed. Thirteen transistors are used in total.




Since







xyz
+

x






y
_



z
_



_










can be expressed by F=1−Sign(x+y+z+3x{overscore (yz)}−3), the corresponding threshold element may be implemented by adding an inverter after the circuit in FIG.


10


. Fifteen transistors are required in total.




{overscore (x)}yz+{overscore (xyz)} Arithmetic




The corresponding threshold function may be expressed by F=Sign({overscore (x)}yz+{overscore (xyz)}−1), a circuit of which is illustrated in FIG.


11


. The circuit eliminates the inverter


140


of

FIG. 10

for inverting the x-input at the first stage, while it includes an inverter


150


for inverting the x-input at the second stage. Thirteen transistors are used in total.




Since










x





_


yz

+


x
_







y
_



z
_



_










can be expressed by F=1−Sign({overscore (x)}+y+z+3{overscore (xyz)}−3), the corresponding threshold element may be implemented by adding an inverter after the circuit in FIG.


11


.




xyz+{overscore (xyz)} Arithmetic




The corresponding threshold function may be expressed by F=Sign(xyz+{overscore (xyz)}−1), a circuit of which is illustrated in FIG.


12


. The circuit simply eliminates the inverter


140


of

FIG. 10

for inverting the x-input at the first stage. Eleven transistors are used in total.




Since








x





yz

+


x
_







y
_



z
_



_










can be expressed by F=1−Sign(x+y+z+3{overscore (xyz)}−3), the corresponding threshold element may be implemented by adding an inverter after the circuit in FIG.


12


.




C-element




The present invention may be applied to the types of calculation other than simple Boolean calculations.

FIG. 13

illustrates the structure of a circuit called a two-input C-element with x- and y-inputs. C-elements are widely used as synchronizing devices in asynchronous circuits. It is regarded as a majority circuit with hysteresis. When C=0 in a two-input C-element, the x- and y-inputs both reaching “1” allows C=1. On the other hand, when C=1, the x- and y-inputs both coming to “0” allows C=0. The behavior of a two-input C-element is specified by a threshold function:







C


(

x
,
y

)


=


Sign


(

x
+
y
+
C
-
2

)


=

Sign


(



C
+
ε



x
_

+

y
_



-
1

)













Accordingly, the x- and y-inputs are connected to the bases of p-transistors


200


,


201


, respectively. The drains of a weak n-transistor


202


which keeps open and an n-transistor


203


which receives the output C at the base are commonly connected to the drains of the p-transistors


200


,


201


. The output voltage Vout appears at the connected drains. The output voltage Vout is supplied to a decision inverter


204


so as to provide a target C. The two-input C-element can be implemented with six transistors, while the similar implementation of a two-input C-element in a CMOS circuitry requires eight transistors.





FIG. 14

illustrates the structure of a three-input C-element with x-, y- and z-inputs. When C=0 in this element, the x-, y- and z-inputs all reaching “1” allows C=1, while the x-, y- and z-inputs all coming to “0” allows C=0 when C=1. The behavior of a three-input C-element is specified by a threshold function:







C


(

x
,
y
,
z

)


=


Sign


(

x
+
y
+
z
+

2

C

-
3

)


=

1
-

Sign


(




x
_

+

C
_



y
+
z
+
C


-
1

)














Accordingly, the x-input and output C are connected to the bases of p-transistors


210


,


211


, respectively. The y- and x-inputs and output C are connected to the bases of n-transistors


212


,


213


,


214


, respectively. All drains of n- and p-transistors


210


-


214


are commonly connected to form the output voltage Vout. Since either of p-transistor


211


or n-transistor


214


should inevitably open, a weak transistor is not necessary to keep the level of the output voltage Vout. The output voltage Vout is supplied to a decision inverter


215


so as to give a target C. The three-input C-element can be implemented with seven transistors, while the similar implementation of a three-input C-element in a conventional CMOS circuitry requires ten transistors.





FIG. 15

illustrates the structure of an n-input C-element. The corresponding threshold function can be generalized as:







C


(
X
)


=

1
-

Sign


(





C
_






i
=
1


n
-
1





x
_

i



+

0.5







x
_

n





C





i
=
1


n
-
1




x
i



+

0.5






x
n




-
1

)













Accordingly, the x


1


- to x


n


-inputs are connected to p-transistors


151


-


15




n


and n-transistors


161


-


16




n


. The drains of all the p-transistors


151


-


15


(n−1) are commonly connected while the drains of all the n-transistors


161


-


16


(n−1) are commonly connected.




A p-transistor


170


and an n-transistor


171


are serially connected between the drains of the p- and n-transistors


15


(n−1),


16


(n−1) corresponding to the x


n−1


-input. The drains of the p- and n-transistors


170


,


171


are connected to the drains of the p- and n-transistors


15


n,


16


n corresponding to the x


n


-input, which in turn are connected to a decision inverter


180


. The feedback from the output of the decision inverter


180


is connected to the drains of the serially connected p- and n-transistors


170


,


171


. When C=1 in the n-input C-element, all inputs assuming “0” allows C=0, while all inputs reaching “1” allows C=1 when C=0.




Full Adder





FIG. 16

illustrates a threshold element of a full adder as another example. The full adder receives a carrier C


i−1


from the ahead stage, and sums x


i


- and y


i


-inputs, to give the sum σ


i


and the carrier C


i


. Threshold representation of a full adder is:














C
i

=

maj


(


x
i

,

y
i

,

C

i
-
1



)



;








C
_

i

=

Sign


(




C

i
-
1


+
ε



x
i

+

y
i



-
1

)









(
7
)













and






σ


i


=Sign(


x




i




+y




i




+C




i−1


−2


C




i


−1)=Sign(


x




i




+y




i




C




i−1


+2


{overscore (C)}




i


−3);


















σ
_

i

=

Sign


(





C
_


i
-
1


+


C

_
_


i




x
i

+

y
i

+


C
_

i



-
1

)






(
8
)













According to the representation (7), a circuit


300


generating {overscore (C)}


i


includes a p-transistor


301


receiving the carrier C


i−1


at the gate. A weak p-transistor


302


is disposed in parallel to the p-transistor


301


for keeping open. The x


i


- and y


i


-inputs are respectively connected to the gates of n-transistors


303


,


304


. The drains of these transistors


301


-


304


are commonly connected. The signal {overscore (C)}


i


appears at the connected drains, to form C


i


through an inverter


320


.




On the other hand, according to the expression (8), a circuit


310


generating {overscore (σ)}


i


includes p-transistors


311


,


312


receiving {overscore (C)}


i


, and C


i−1


at the gate. The x


i


- and y


i


-inputs and {overscore (C)}


i


are respectively connected to the gates of n-transistors


313


,


314


,


315


. The drains of these transistors


311


-


315


are commonly connected. The signal {overscore (σ)}


i


appears at the connected drains, to form σ


i


through an inverter


321


.




As described above, a full adder of the present invention employs nine transistors (with two inverters), while a similar implementation of a full adder in a conventional CMOS circuitry requires twenty transistors (plus two inverters). The scale of the circuitry can be reduced in the present invention.



Claims
  • 1. A method of designing a threshold element, comprising:transforming to a form including a ratio of a parameter related to p- and n-transistors a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of weighted inputs; and designing a threshold element corresponding to the threshold function by allocating the inputs to the p- and n-transistors in accordance with the ratio, wherein said parameter is a drain current coefficient of the transistor, and wherein said threshold function is transformed to a form of Sign ((the ratio of drain current coefficient) −1).
  • 2. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, and wherein said transformed form includes Sign ((said ratio) −1).
  • 3. The threshold element according to claim 2, wherein said parameter is a drain current coefficient of the transistor.
  • 4. A threshold element comprising.a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of x(y+z) based on said threshold function comprising F=Sign(2x+y+z−3).
  • 5. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of xy+yz+zx based on said threshold function comprising F=Sign (x+y+z−2).
  • 6. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of x+yz based on said threshold function comprising F=Sign (2x+y+z−2).
  • 7. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of xy{overscore (z)} based on said threshold function comprising F=Sign (x+y+{overscore (z)}−3).
  • 8. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of x{overscore (yz)} based on said threshold function comprising F=Sign (x+{overscore (y)}+{overscore (z)}−3).
  • 9. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of {overscore (xyz)} based on said threshold function comprising F=Sign ({overscore (x)}+{overscore (y)}+{overscore (z)}−3).
  • 10. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of xyz+x{overscore (yz)} based on said threshold function comprising F=Sign (xyz+x{overscore (yz)}−1).
  • 11. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of {overscore (x)}yz+{overscore (xyz)} based on said threshold function comprising F=Sign ({overscore (x)}yz+{overscore (xyz)}−1).
  • 12. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of xyz+{overscore (xyz)} based on said threshold function comprising F=Sign (xyz+{overscore (xyz)}−1).
  • 13. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of {overscore (xyz)} based on said threshold function comprising F=1−Sign (x+y+z−3).
  • 14. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of xy⁢ ⁢z__based on said threshold function comprising F=1−Sign (x+y+{overscore (z)}−3).
  • 15. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a Boolean calculation of x⁢ ⁢y_⁢ ⁢z__based on said threshold function comprising F=1−Sign (x+{overscore (y)}+{overscore (z)}−3).
  • 16. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a two-input C-element with x- and y-inputs based on said threshold function comprising C=Sign (x+y+C−2).
  • 17. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a three-input C-element with x-, y- and z-inputs based on said threshold function comprising C=Sign (x+y+z+2C−3).
  • 18. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p-and n-transistors, the threshold element implementing an n-input C-element with x1-, x2-, . . . , xn-inputs based on said threshold function comprising C=1-Sign⁡(C_⁢∑i=1n-1⁢x_i+0.5⁢ ⁢x_nC⁢∑i=1n-1⁢xi+0.5⁢ ⁢xn-1).
  • 19. A threshold element comprising:a p-transistor receiving an input allocated in accordance with a predetermined ratio; and an n-transistor receiving an input allocated in accordance with the ratio, wherein said ratio is included in a transformed form of a threshold function which outputs a result of comparison between a predetermined threshold value and a sum of the inputs, said ratio representing a parameter related to the p- and n-transistors, the threshold element implementing a full adder with xi- and yi-inputs, a carry-in Ci−1, a carry-out Ci, and a sum result σi based on said threshold function comprising C_i=Sign(Ci-1+εxi+yi-1)⁢ ⁢andσ_i=Sign(C_i-1+C__ixi+yi+C_i-1) where ε is a value corresponding to a transistor introduced for convenient designing.
Priority Claims (1)
Number Date Country Kind
10-054079 Jan 1998 JP
US Referenced Citations (3)
Number Name Date Kind
5706226 Chan et al. Jan 1998 A
5872030 Huang Feb 1999 A
6198336 Varshavsky Mar 2001 B1
Non-Patent Literature Citations (6)
Entry
Bohossian et al, “Programmable Neural Logic,” IEEE, Oct. 1997, pp. 13-21.*
Siu et al, “Rational Approximation Techniques for Analysis of Neural Networks,” IEEE, Mar. 1994, pp. 455-466.*
Varshavsky, “β-Driven Threshold Elements,” IEEE, Feb. 1998, pp. 1-7.*
Varshavsky, “CMOS Artificial Neuron on the Base of β-Driven Threshold Element,” IEEE, Oct. 1998, pp. 1857-1861.*
Varshavsky et al, “Implementability Restrictions of the Beta-CMOS Artificial Neuron,” IEEE, Sep. 1999, pp. 401-405.*
Busta et al, “Hardware Implementation of a “Wired-Once” Neural Net in Thin-Film Technology on a Glass Substrate” IEEE, Apr. 1990, pp. 1093-1045.