Proceedings of ESSCIRC, Grenoble, France, 2005
A Novel Current-Mode Very Low Power Analog CMOS Four Quadrant Multiplier Mirko Gravati(1), Maurizio Valle(1), Giuseppe Ferri(2), Nicola Guerrini(2) and Linder Reyes(3) (1) DIBE, University of Genova, Via All’Opera Pia 11/A, I16145 GENOVA, Italy. (2) DIE, Università dell’Aquila, Monteluco di Roio, I67040 L'Aquila, Italy (3) Instituto de Ingenieria Electrica, Facultad de Ingenieria, Universidad de la Republica, Montevideo, Uruguay
[email protected],
[email protected] ,
[email protected] ,
[email protected] ,
[email protected]
Abstract: In this paper, a novel current mode CMOS four-quadrant analog multiplier circuit is presented. The multiplication is implemented by four translinear loops with MOS transistors operating in weak inversion. Information carrying signals are differential balanced currents. The multiplier circuit has been implemented in a test chip in a standard 0.35 ȝm CMOS technology. The experimental measurements (dc bias current of 250 nA and a power supply of 2.0 V) show a bandwidth of 200 kHz and a THD figure value lower than 0.9 %. The multiplier features a wide signal dynamic range and linearity, low power consumption (the maximum power consumption is of 5.5·10-6 W) and very low area (18.7 10-3 mm2). The multiplier is suitable for a wide range of analog signal processing applications. Due to the low power and silicon area consumption, scalability and modularity can be also easily integrated in massive parallel systems.
1. Introduction The evolution of integrated circuit technology and future scenarios of ubiquitous and pervasive computing have stressed the need of very low power and low voltage circuits with high signal dynamic range and linearity. We will focus on analog multipliers as basic blocks of many analog and mixed-mode systems: e.g. RMS-DC converters, modulators, frequency synthesizers, massive parallel systems, etc. Current inputs – current outputs multipliers can achieve low power, low voltage and high dynamic range. To this aim, two different implementation approaches can be found: i) translinear loops with MOS transistors operating in the strong inversion region [1]; ii) loops with MOS transistors operating in the weak inversion region [2]. Both approaches are suitable for standard CMOS fabrication. In this paper we present a compact current mode CMOS four-quadrant analog multiplier circuit based on a novel topology which adopts the latter approach. The circuit exhibits wide signal dynamic range, high modularity and scalability. We designed a test chip using the AMI Semiconductor CMOS 0.35 Pm minimum channel length technology and we report main experimental results and 0-7803-9205-1/05/$20.00 ©2005 IEEE
compare them with those reported in [1]. The experimental results indicate that the circuit achieves very low power, low voltage, high linearity and silicon area efficiency. The paper is organized as follows. Section 2 will present the multiplier circuit, while Section 3 will introduce and discuss the experimental results. Conclusions are drawn in Section 4.
2. Translinear current mode MOS multiplier circuits 2.1
Modelling MOS translinear loops
In a previous paper [3] we presented a model of MOS translinear loops operating in weak inversion. We will use that model to introduce the novel circuit topology that will be presented in Section 2.2. Hence, we briefly introduce the model reported in [3]. We refer to the following expression of the channel current in a MOS transistor biased in weak inversion [4]:
I DC e
I DS
VGS VTH nIt
V DS ª «1 e It «¬
º » »¼
Eq. 1 where IDC is a current term, VTH is the threshold voltage, n is the weak inversion slope factor. The mismatch between devices causes random variations of the values of IDCi and VTHi [5]. From Eq. (1) and taking into account a generic transistor i of a translinear loop, we can define
Gi
G (VDS ) 1 e
VDSi
It
i
as a generic error term whose
value depends on the drain to source voltage value. If, due to mismatch between devices, the terms IDCi and VTHi experience variations (i.e. errors) of 'IDCi and 'VTHi respectively from their nominal/typical values, then we can write for a generic transistor i:
I DSi
1 'I DCi I DC e
VGSi VTH nIt
e
'VTHi nIt
Gi
Eq. 2 Please note that in Eq. (2), the term 'IDCi represents the percentage variation with respect of the nominal value; on the other hand 'VTHi represents an absolute variation.
495
Authorized licensed use limited to: Universidad de la Republica. Downloaded on July 30, 2009 at 16:52 from IEEE Xplore. Restrictions apply.
Paper 9.E.1
Proceedings of ESSCIRC, Grenoble, France, 2005
is a bias (reference) current. The output current is:
In other words:
1 'I DC I DC , VTH _ real
I DC _ real
VTH 'VTH
I OUT
Let us take into account the (basic) translinear loop shown in Fig. 1. IDS1
IDS2
M1
IDS3
M2
2.2
M4
Novel multiplier circuit topology
I O1
( I X I W ) I B . The transistors M7, M8, M11 and
M12
(translinear
operation: I O 2
Figure 1: Generic (alternate) translinear loop.
loop)
implement
the
following
X W
( I I ) I B . The currents IO1 and IO2
are summed at node n2; the result is the positive single ended term of the output current IOUT+:
After some mathematical computations one can obtain:
I DS 1 I DS 3
xwI B .
In this paper we present a novel circuit topology that is shown in Figure 2. The transistors M1, M2, M3 and M4 (translinear loop) implement the following operation:
IDS4
M3
I OUT I OUT
J I DS 2 I DS 4
I OUT
Eq. 3 where we have approximated n to the value of one. In the
I X I W ( I X I W ) I B .
I O1 I O 2
Iout-
Iout+
(n1)
(n2)
Iw+
IB
M1
Io1
M2
M3
I x+
I x-
M4
M5
Io3
M6
Iw-
IB
M7
Io4
M8
I x+
I x-
M10
M9
M11
Io2
M12
(np)
+ VPOL
-
Mp2
Mp1
previous equation, we introduced a “non linearity” factor J which takes into account the effects of mismatch between the devices belonging to the translinear loop. The term J is defined as follows: 1 'I DC1 1 'I DC3 G1G 3 e 'VTH 1 'VTH 3n'It VTH 2 'VTH 4 J 1 'I DC2 1 'I DC4 G 2G 4 Eq. 4 The non linearity factor J is given, besides by the spread of the technological parameters 'IDCi and 'VTHi, by the bias point value through the terms
Gi 1 e
Figure 2: Novel multiplier circuit topology In a similar way, the current term IO3 (the result of the operation of the translinear loop made by transistors M1, M2, M5 and M6) is summed at node n1 to the current term IO4 (the result of the operation of the translinear loop made by transistors M7, M8, M9 and M10). The result is the negative single ended term of the output current IOUT-: I OUT
VDSi
It
(1 x)( I B 2), I X (1 w)( I B 2), I W
I X I W ( I X I W ) I B .
It has been verified through experimental measurements that the translinear loops are rather insensitive to the value of VPOL to a large extent (i.e. [0.5 y1.2] V). Then, to simplify the circuit, one can substitute the bias voltage generator VPOL with a transistor working in the linear region (transistor Mb1, see Figure 3)
. In the
following we will consider all terms Gi equal to 1. The error given by this approximation is fairly low: in fact if, let say, VDS is equal to only 100 mV, the error is in the order of magnitude of about 0.05%. Please note that the non linearity term J depends also on the topology of the circuit and on the layout design (i.e. matching structures). In particular, J = 1 in the case of ideal matching between the devices of the translinear loop. In the following subsections we will apply the previous model to the novel four quadrant current mode translinear multiplier circuit. In the following we will consider input (IX and IW) and output (IOUT) signals as differential and balanced current mode signals:
I X I W
I O3 I O 4
Iw+
IB
(np)
Mb1
(1 x)( I B 2) (1 w)( I B 2)
Mp1
Figure 3: Circuit implementation of VPOL. Due to the spread of technological parameters, each translinear loop introduces a non linearity term Ji (i=1:4, See Eq. (4)).
where x and w are the input information carrying variables (please note that -1 d x d+1, -1 dw d+1) and IB 496
Authorized licensed use limited to: Universidad de la Republica. Downloaded on July 30, 2009 at 16:52 from IEEE Xplore. Restrictions apply.
Proceedings of ESSCIRC, Grenoble, France, 2005
The output current can be expressed as:
I OUT
I offset > Ax x Aw w AI xw@
where:
AI
Ax
1,0
Eq. 5
0,5
J1 J 2 J 3 J 4 , IB , 4
Iout/ IB
J 1 J 2 J 3 J 4
I offset
IB 4
x = -1 x = -0.6 x = -0.3 x=0 x = +0.3 x = +0.6 x = +1
0,0
-0,5
J 1 J 2 J 3 J 4 , Aw J 1 J 2 J 3 J 4 . -1,0
The proposed multiplier circuit topology is more symmetric and exhibits the following advantages over the standard current mode MOS Gilbert multiplier ([6]): a) The expression of the output current (see Eq. (5)) does not present any higher order term of the inputs (i.e.
x 2 , x 3 .. w 2 , w3 .. ) even in the case that mismatch is taken into account. b) If the non linearity terms Ji assume similar values (i.e. in the case of matching inside and between the transistors of translinear loops) then the terms AI, Ax, Aw, Ioffset tend to decrease and the overall linearity increases. C) In the expressions of AI, Ax, Aw, Ioffset, terms in the form of JiJj (izj) are not present.
-0,5
0,0
0,5
1,0
w [-1:1]
Figure 5: Measured DC transfer characteristics Fig. 6 shows the DC measured transfer characteristics of the multiplier in the case: x input is on the x axis, and the w input is used as parameter. 1,0
0,5
Iout/ IB
3. Experimental results
-1,0
w = -1 w = -0.6 w = -0.3 w=0 w = +0.3 w = +0.6 w = +1
0,0
-0,5
We designed a test chip by using the AMI Semiconductor CMOS 0.35 Pm minimum channel length double metal, double poly technology. The microphotograph of the multiplier is shown in Fig. 4. The multiplier area is of about 170 Pm u 110 Pm.
-1,0
-1,0
-0,5
0,0
0,5
1,0
x [-1:1]
170Pm
Figure 6: Measured DC transfer characteristics
3
2
Absolute error [nA]
110Pm
Translinear loops
One can note that the multiplier exhibits linear behaviour with respect to both inputs. The static linearity error is shown in Fig. 7.
Figure 4: Microphotograph of the multiplier The transistors sizes are reported in the Table 1. Transistor W [Pm] /L [Pm]
Mp1-Mp2 Mb1-Mb2
1/6
3/10
M1-M2-M3-M4M5-M6-M7-M8M9-M10-M11M12
1
x = -1 x = -0.6 x = -0.3 x = +0.3 x = +0.6 x = +1
0
-1
-2
-1,0
-0,5
0,0
0,5
1,0
w [-1:1]
60/1
Figure 7: Linearity error
Table 1: sizes of transistors (see Fig. 2 and Fig. 3) In all the following measurement results, IB was set at 250 nA; IOUT+ and IOUT- vary in the range [0nA y 250nA], while IOUT varies in the range [-250nA y +250nA]. Fig. 5 shows the DC measured transfer characteristics of the multiplier. The w input is on the x axis, and the x input is used as parameter.
The measured -3 dB Band Width is of 200 kHz. The maximum power consumption is of 5.5 10-6 W. The total harmonic distortion (THD) of the output current versus the input signal level is shown in Fig. 8.
497
Authorized licensed use limited to: Universidad de la Republica. Downloaded on July 30, 2009 at 16:52 from IEEE Xplore. Restrictions apply.
Proceedings of ESSCIRC, Grenoble, France, 2005
multiplier. With respect to other well known topologies, the proposed circuit is linear with respect to both inputs We compared the proposed circuit topology with the standard MOS current mode Gilbert multiplier ([6]) and we found that our multiplier achieves higher linearity. When compared to state-of-the-art voltage- translinear CMOS multiplier circuits, such as that proposed in [1], the performances of the multiplier proposed here are similar in terms of linearity (see Table 2). Nevertheless, the power consumption and silicon area are impressively lower also taking into account the smaller minimum channel length given by the technology.
1,0
0,8
THD [%]
0,6
x input w input
0,4
0,2
0,0 20
40
60
80
100
120
140
160
180
200
Modulated input current [nA]
Figure 8: THD of the output waveform for different input amplitudes. The magnitude spectrum of the output current waveform is shown in Fig. 9: the w input is a 1kHz sinusoidal waveform with a peak value of 112 nA, (W = 0.45) and the x input is a DC input with a value of 210 nA (X = 0.84).
Multiplier
[1]
Technology (CMOS)
2.4 ΐm
This paper 0.35 ΐm
Supply voltage [V]
3.3
2.0
THD as gain cell [%] (measured)
1.5
0.9
Max. rel. error as multiplier [%] (measured)
1.9
5
BW as gain cell
3 MHz
200 kHz
(simulated)
(measured)
Area [mm2]
0.24
18.7 10-3
Power consumption [ΐW]
600
5.5
Table 2 Our long term goal is to design multiplier circuit topologies able to effectively operate at very low current values [7].
Figure 9: Magnitude spectrum of the output current waveform. Figure 10 shows the experimental results obtained by modulating a 4 kHz high frequency sinusoidal signal with a peak value of 160nA, (w = 0.64) with a 100 Hz low frequency sinusoidal signal with a peak value of 40 nA (x = 0.16).
Figure 10. Output current waveform in the case of waveforms modulation.
4. Conclusions
References: [1] A. J. Lopez-Martin, A. Carlosena, “Current-Mode Multiplier/divider Circuits based on the MOS Translinear Principle”, Analog Integrated Circuits and Signal Processing, Vol. 28, pp. 265 – 278, 2001. [2] A.G. Andreou and K. A. Boahen, “ Translinear Circuits in subtrsheold CMOS”, Analog Integrated Circuits and Signal Processing, Vol. 9, pp. 141 – 166, 1996. [3] M. Gravati and M. Valle, “Modelling mismatch effects in CMOS translinear loops and current mode multipliers”, submitted to ECCTD05, Cork, Ireland, 29 august – 1 september 2005. [4] Y. Tsividis, “Mixed Analog-Digital VLSI Devices and Technology”, Mc Graw Hill, 1996. [5] B. Razavi, “Design of Analog Integrated Circuits”, McGraw Hill, 2001. [6] F. Diotalevi, M. Valle, “An analog CMOS four quadrant current-mode multiplier for low power artificial neural networks implementation”, ECCTD’01, Helsinki, Finland, 28 – 31 august 2001, pp. III – 325 – III 328 156. [7] B. Linares-Barranco and T. Serrano-Gotarredona, “On the Design and Characterization of Femtoampere CurrentMode Circuits”, IEEE J. of Solid State Circuits, Vol. 38, No 8, August 2003.
In this paper we present the circuit topology of a novel analog CMOS four quadrant current mode 498
Authorized licensed use limited to: Universidad de la Republica. Downloaded on July 30, 2009 at 16:52 from IEEE Xplore. Restrictions apply.