"Switched Current Memory Cell Using Floating Gate MOSFET"
J. of Active and Passive Electronic Devices, Vol. 2, pp. 261–269 c 2007 Old City Publishing, Inc. Reprints available directly from the publisher Published by license under the OCP Science imprint, Photocopying permitted by license only a member of the Old City Publishing Group Switched Current Memory Cell Using Floating Gate MOSFET M ANEESHA G UPTA∗ AND R AHUL S URI Electronics and Communication Engineering Department, Netaji Subhas Institute of Technology, New Delhi-110045 An implementation of switched-current memory cell in ﬂoating gate MOSFETs (FGMOS) technology is presented. The sampled and hold circuit built using FGMOS based memory cell has better performance due to the increased gate to source capacitance inherent to the FGMOS technology. Noise analysis shows improved performance of the proposed memory cell over conventional memory cell. Pspice simulations have been used to validate the theoretical predictions. Key words: Analog current mode circuits, switched current, ﬂoating gate MOSFET. 1 INTRODUCTION Switched current technique is being extensively used in the design of differentiator and integrator circuits, Switched current ﬁlters, Analog/Digital converters, Sigma-Delta converters [1, 2]. The switched current circuits can be build through MOSFETs without the need of any special design process which makes it ideally suitable for mixed ICs for implementation in standard CMOS technology. The non-ideal behavior of switched current memory cell results mainly from clockfeedthrough  switch charge injection , non zero output conductance  and settling time. Several design techniques have been used to remove these imperfections [6, 11]. Many researchers have analyzed the cause and effects of noise in current memory cells [3, 4] and proposed the measures to reduce it but there haven’t been many practicable solutions ∗ Corresponding author: E-mail: maneesha.gupta email@example.com 261 262 G UPTA AND S URI FIGURE 1 Floating Gate MOSFET (a) Layout (b) Equivalent Model (c) Symbol. that have been implemented. The FGMOS‘s have been used in switched current memory cell  for the low voltage applications. This paper aims at introducing a new implementation for switched current memory cell using multiple input ﬂoating gate MOSFETs (FGMOS) for improved noise performance. FGMOS can be used instead of MOSFETs in other switched current circuits so that a new class of switched current circuits can be evolved. The enhancement proposed for the memory cell is superior as it retains the same number of transistors with improved noise performance. 2 FGMOS AS CMC A multiple input FGMOS is shown in Figure 1. It consists of an n-MOS having a electrically ﬂoating gate electrode. The ﬂoating gate, which is built on the conventional Poly1, is capacitively coupled to the input control gates built on Poly2. If the total initial charge on the ﬂoating gate is removed by ultraviolet light [8, 9], then the voltage at the ﬂoating gate of a FGMOS can be expressed as: n 1 Vfg = Ci Vi (1) CT i=0 n CT = Ci (2) i=0 where C0 is the capacitor formed between the ﬂoating gate and the substrate, and C1, C2 , C3 . . . Cn are the capacitors between the control gates and the ﬂoating gate. V0 is the substrate potential and Vi represents the voltages S WITCHED C URRENT M EMORY C ELL U SING F LOATING G ATE MOSFET 263 FIGURE 2 Floating Gate MOSFET as used in the memory cell. (a) Equivalent Model (b) Symbol. of the input control gates. Figure 1b shows the symbol used for FGMOS and Figure 1c show the equivalent model for FGMOS. The FGMOS memory cell employs 4-input ﬂoating gate transistors (Figure 2). While one of the input control gate is used for analog signal processing as in the case of a MOSFET in a conventional memory cell, the other three control gates are grounded. Since in this structure, the source of the FGMOS is grounded the net gate to source capacitance (Cgs) increases as compared to Cgs, in a conventional memory cell. The substrate of the FGMOS is at ground, therefore all the capacitances are in parallel. Thus, in FGMOS the effective gate to source capacitance Cnet increases. Cnet can however be further increased by using more number of input control gates for FGMOS. The increase in Cnet can be exploited to reduce noise level in the memory cell. 3 FGMOS MEMORY CELL The CMC designed with FGMOS in place of MOSFET is presented (Figure 3). During the phase φ1 , the input current Iin adds to the bias current J and the sum of the two currents ﬂows into the initially discharged gate to source capacitance, Cnet of the ﬂoating gate transistor. As a result the capacitance Cnet begins to charge and as the voltage across it exceeds the threshold voltage, the ﬂoating gate transistor T1 conducts. When Cnet becomes fully charged, the entire current J + Iin ﬂows into the drain of the FGMOS. When phase φ2 occurs, the current J + Iin is sustained in the drain because the gate-source voltage is maintained constant by virtue of the charge stored in the capacitor. Thus during φ2 the output current i 01 is the 264 G UPTA AND S URI FIGURE 3 FGMOS Memory Cell. difference of the bias current and the drain current written as: I01 = J − ( J + Iin ) = −Iin (3) In order to obtain a memory of the input current throughout the entire clock period φ1 and φ2 , the circuit can be modiﬁed by adding another FGMOS as shown in Figure 4. The output current I02 ﬂows during the sampling phase φ1 and the holding phase φ2 by current mirror action. If the two FGMOS transistors are matched then, I02 = −Iin (4) FIGURE 4 FGMOS Memory Cell Modiﬁed. S WITCHED C URRENT M EMORY C ELL U SING F LOATING G ATE MOSFET 265 4 NOISE REDUCTION Noise in switched CMC results from the transistor switch and the copying transistor. For a MOS, noise has two components, thermal noise and 1/ f noise. For a transistor in saturation with transcondutance, gm the thermal noise current power spectral density is given as follows  i t2 2gm = 2kT −∞ f ∞ (5) f 3 The 1/ f noise component has a power spectral density that decreases as frequency increases and is expressed as: i t2 2 K f n gm = (6) f Cox W L| f | Where K f n is a process dependent constant, Cox is the gate capacitance per unit area and W and L are the transistor’s width and length respectively. Now we can write , K fn 2 SM ( f ) = + 2kT (7) Cox W L| f | 3gm While current is being sampled by copier cell, noise from both the switch and the transconductor changes the voltage across capacitor. When the switch is turned off, a sample of this noise is stored on the capacitor. It can be shown that the power spectral density of noise on the sampling capacitor due to switch transistor and copy transistor is given respectively as (g0 /gm )2 SC R ( f ) = SR ( f ) (8) 1 + ( f / f 0 )2 1 SC M ( f ) = SM ( f ) (9) 1 + ( f / f 0 )2 It is to be noted that the transconductor noise alters the voltage on the capacitor much more than the switch noise. Since go /gm 1, (go and gm are trans-conductances of the switch and the copy transistor respectively) switch noise can be completely neglected and total PSD is expressed as SC ( f ) = SC R ( f ) + SC M ( f ) (10) 1 SC ( f ) = SM ( f ) (11) 1 + ( f / f 0 )2 An expected value of mean square output noise current can be obtained by integrating the current noise spectral density over the cell’s own 266 G UPTA AND S URI bandwidth  ∞ ∞ ∞ i TOT = 2 ST OT ( f )d f = SISD ( f )d f + SIH ( f )d f D 0 0 0 ∞ =2 SISD ( f )d f (12) 0 Here SiSDrepresents the current noise spectral density at the output and is given as SISD = gm SC ( f ) 2 (13) If nT s < t < (n + 1)Ts is considered then only the baseband noise holds. ∞ 2 1 4kT i T OT =2 2 gm df (14) 1 + ( f / f0 ) 2 3g m 0 4kT π = 2gm 2 f0 (15) 3gm 2 The loop bandwidth f o = gm /(2πC) is substituted to get, 2 kT i T OT = 2gm 2 (16) 3C With a current copier the low frequency 1/ f noise is removed and has been neglected in evaluating the integral. The expected current noise at the cell’s output can be written as 2kT σni = gm (17) 3C Thus, one of the ways of reducing the output noise current is to increase the gate to source capacitance. Another superiority of the proposed memory cell is that the increase in the capacitance can be adjusted as per the bandwidth requirements by varying the number of input control gates for the ﬂoating gate transistor. Greater the number of grounded input control gates, greater is the effective gate to source capacitance. Thus, reduction in output noise is achieved with the proposed memory cell without any additional MOSFET or external capacitance. 5 SIMULATION RESULTS The proposed CMC has been simulated on PSPICE with the assistance of the FGMOS macro model shown in appendix. The parameters used for S WITCHED C URRENT M EMORY C ELL U SING F LOATING G ATE MOSFET 267 FIGURE 5 Time Domain Simulation results a) FGMOS Memory Cell b) Conventional Memory Cell c) FGMOS Cell modiﬁed. 268 G UPTA AND S URI FIGURE 6 Macro Model for FGMOS. the ﬂoating gate MOSFETS are: W = 16µm, L = 1.5µm, Cgso = 1 pF, Cgbo = 1pF, and the coupling capacitors are each 357fF. The current source has a value J = 100µA. For a sinusoidal input current, the current at the output of the FGMOS memory cell of Figure 3 is shown in Figure 5(a). It can be seen that the current I01 is a memorized version of the input current. It can be compared with the current I01 at the output of the conventional single MOSFET memory cell [6–10] shown in Figure 5(b) to conﬁrm identical operation of the two memory cells. Further, the current I02 at the output of the modiﬁed memory cell (Figure 4) is shown in Figure 5(c), for a sinusoidal input current, Iin . This current is memorized version of the input current and is available during both phases φ1 and φ2 . 6 CONCLUSIONS In this paper, a new FGMOS based implementation for switched CMC has been presented. It was shown that the CMC would work in the same manner as in case of a CMC. FGMOSs have been connected in such a way so that the total gate to source capacitance of the copy transistor increases. Noise analysis shows that the proposed CMC has better noise performance. The proposed CMC is completely implementable in CMOS technology and the designer has the control over the amount of increase in the capacitance. REFERENCES  Hughes, J. B., Bird, N. C., and Macbeth, I. C. (1989). “Switched Current –A new technique for analog sampled-data signal processing”, Proc. IEEE International symposium on Circuits and Systems, 1584–1587.  Hughes, J. B., Macbeth, I. C., and Patullo, D. M. (1990). “Switched Current Filters”, Proc. IEE Part G. S WITCHED C URRENT M EMORY C ELL U SING F LOATING G ATE MOSFET 269  Helfenstein, M., and Moschytz, G. S. (1998). “Improved two-step clock-feedthrough compensation technique for switched current circuits”, IEEE Transactions on Circuits and Systems-II: Analog Digital Signal Process., 45(6), 739–743.  Balachandran, G. K., and Philip E. Allen, (2002). “Switched-Current Circuits in Digital CMOS Technology with Low Charge-Injection Errors”, IEEE Journal of Solid-State Circuits, 37(10).  Toumazou, C., Hughes, J., and Patullo, D. (1990). “Regulated cascode switched current memory cell”, Electronic Letters, 26(5), 303–305.  Guggenbuhl, W., Di, J., Goette, J. (1994). “Switched-Current Memory Circuits for High-Precision Applications”, IEEE Journal of Solid-State Circuits, 29(9).  Steven Daubert, “Noise in Switched Current circuits”, Switched-Currents: an analogue technique for digital technology, Peter perigrinus Ltd., London, U.K., ISBN 0 86341 294 7, chapter 5.  Thomsen, A., Brooke, M. A. (1994). “Low control voltage programming of ﬂoating gate MOSFETS and applications”, IEEE trans. Circuits and Systems, 41(6), 443–452.  Chai, Y. Y., Johnson, L. G. (1994). “Floating gate MOSFET with reduced programming voltage”, Electronic Letters, 30(18), 1536–1537.  Alarcon, E., Iannazzo, M., Madrenas, J. (1997). “Noise and Speed Performance in Switched- Current Memory Cells”, Proceedings of the All Conference Design of Integrated Circuits and Systems 1997 (DCIS 97), 59–64, Sevilla, Spain.  Balachandran, G. K., and Allen, P. (2001). “A fully differential switched current memory cell with low charge-injection errors ”, IEE Proc. Ckts., Devices, Systems, 148(3), 157–164.  Igor Mucha, (1999). “Ultra low voltage class AB Switched current memory cell based on the ﬂoating gate transistors”, Analog Integrated circuit Circuits and Signal Processing, 20, 43–62. APPENDIX The macro model of the FGMOS used in the simulations is shown below. Large resistances are included in parallel with the capacitances to avoid DC convergence problems in simulations. The expressions pertaining to the model are 1) G 0 (V0 − V f g ) + G 1 (V1 − V f g ) + · · · + G n (Vn − V f g ) = 0 (1.1) n where G i = 1/Ri and G T OT = Gi i=1 2) Given G i = kCi then (C0 V0 + C1 V1 + · · · + Cn Vn Vfg = (1.2) C T OT Ri = 1/kCi where k is a constant chosen to provide a large value of Ri of the order of 1000G. Thus, their effect can be neglected.