• 沒有找到結果。

The design of CMOS continuous-time VHF current and voltage-mode lowpass filters with Q-enhancement circuits

N/A
N/A
Protected

Academic year: 2021

Share "The design of CMOS continuous-time VHF current and voltage-mode lowpass filters with Q-enhancement circuits"

Copied!
11
0
0

加載中.... (立即查看全文)

全文

(1)

614 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 31, NO. 5 . MAY 1996

The Design

of

CMOS Continuous-Time VHF

Current and Voltage-Mode Lowpass Filters with

Q-Enhancement Circuits

Chung-Yu W u , Member, IEEE, and Heng-Shou Hsu

Abstract-In this paper, very high frequency (VHF) current and voltage biquadratic lowpass filters implemented directly by the linear wideband finite-gain current and voltage amplifiers, respectively, are proposed and analyzed. A new Q-enhance- ment circuit which consists of a finite-gain wideband tunable voltage amplifier and a Miller capacitor is also proposed. It can increase the maximum-gain frequency f, and enhance the max- imum-gain quality factor Q, of the VHF lowpass filters. Ex- perimental results have successfully verified the capability of the proposed new filter implementation method in realizing both VHF current and voltage lowpass filters with maximum-gain frequencyf, tunable in the range of 148 MHz to 92 MHz. It is also shown from experimental results that the VHF current lowpass biquad with the Q-enhancement circuit has the maxi- mum-gain frequency f, near 185 MHz and the maximum-gain quality factor Q, up to 18.5. A fourth-order Chebyshev cur- rent lowpass filter with the cut-off frequency of 190 MHz has been successfully designed by using the current biquads with Q-enhancement circuits.

I. INTRODUCTION

N disk-drive read-channel systems, lowpass filters with

I

the passband frequency up to 50 MHz are required. In this specific application, continuous-time technique has been proposed and widely used to implement the required filters [l]-[S]. In many very high frequency (VHF) filter- ing applications, continuous-time filters rather than sam- pled-data filters are also preferred. Thus many efforts have been contributed to develop continuous-time VHF filters including both current-mode and voltage-mode filters

[5],

t61, [81,

t

121,

t

191-t2 11.

In the construction of continuous-time current-mode filters, current integrators derived from the switched-cur- rent (SI) integrator [11]-[13] or current-mode G,-C inte- grators [14]-[16] are the basic building blocks. These im- plementation methods can realize the current-mode filters in the frequency range of several tens megahertz. On the other hand, the basic building blocks of many .voltage- mode continuous-time filters are formed by loading the

Manuscript received May 4 , 1995; revised October 22, 1995. This research is supported by the National Science Council (NSC), Tai- wan, R . O . C . , under Grant NSC83-0416-E-009-016.

The authors are with Integrated Circuits and Systems Laboratory, De- partment of Electronics Engineering and Institute of Electronics, Engi-

neering Building 4th. National Chiao Tung University, 1001 Ta-Hsueh Rd, Hsinchu, Taiwan 300, Republic of China.

Publisher Item Identifier S 0018-9200(96)03412-9.

capacitors at the output of the transconductance amplifiers [9]-[lo]. The filters using this kind of integrator as the basic cells are the well-known G,-C filters [1]-[8] which can be operated in the VHF range in current CMOS tech- nologies.

To form a biquadratic lowpass filter, two current inte- grators or G,-C voltage integrators as well as some con- stant-gain stages are connected together in a feedback structure. For certain filter specifications, a high quality factor is needed [17]-[18]. A design method which uses multiple-output nonlinearized operational transconduc- tance amplifiers (OTA’s) as building blocks has been de- veloped for the implementation of high Q and high-fre- quency second-order filters [ 181. The second-order filters can be cascaded to form high-order filters. Recently, a new design method which uses the transresistance- C(R,-C) differentiator as the second-order bandpass filter and the basic building block of high-order filters has also been developed and successfully applied to the realization of VHF bandpass filters with center frequencies up to 100 In this paper, simple unity-gain current and voltage am- plifiers constructed by G, and R, amplifier stages are pro- posed and analyzed. Considering the device capacitances of the MOS transistors as filter elements, both unity-gain current and voltage amplifiers can be used directly to re- alize current-mode and voltage-mode lowpass tunable bi- quadratic filters, respectively. The biquads can be further used to build high-order filters. Moreover, a new circuit called the Q-enhancement is proposed. Applying the cir- cuit to lowpass current biquads, the maximum-gain fre- quency fM and the maximum-gain quality factor QM can be enhanced efficiently.

In realizing biquadratic or high-order lowpass VHF fil- ters, the proposed circuits have the advantages of compact structure and easy design. The proposed Q-enhancement circuit has the advantage of efficient increase of the qual- ity factor Q. It also makes the controlling voltage ( Vcq) of the quality factor separated from the controlling voltage (Vcn = - Vcp) of the center-frequency wo for easy tuning, although extra power dissipation and chip area are re- quired for the Q-enhancement circuit. Moreover, the tun- ing of the proposed Q-enhancement circuit can be easily achieved with good stability by a controlling gate voltage. MHz [19]-[21].

(2)

wu A N D m u : DESIGN OF CMOS CONTINUOUS-TIME FILTERS 615

In Section 11, the circuit structures of the wideband fi- nite-gain current and voltage amplifiers are proposed and analyzed. In Section 111, VHF current and voltage bi- quadratic lowpass filters design based on the unity-gain amplifiers are described. The special Q-enhancement cir- cuit used to increase both f M and QM of the VHF lowpass biquad is proposed and analyzed. High-order filter design with the biquads as basic building blocks is also demon- strated. The experimental results are presented and dis- cussed in Section V. In Section V, conclusions are given.

11. WIDEBAND AMPLIFIERS DESIGN

Fig. 1 shows the proposed linear wideband current am- plifier. In the circuit of Fig. 1, a symmetric push-pull common-gate amplifier is formed by the transistors M P 2 , MN1. MP1, and MN2 to perform a simple I to I/ conver-

sion and offer a low input impedance to the current am- plifier. Basically, this stage can be treated as a transre- sistance (R,) amplifier whose outputs are connected to the gates of MOSFET’s in the next stage. In the second stage, the transistors MPR, M N R , M P 3 , MN3, M P 5 , and MN5 form a pair of tunable G,n amplifiers. The triode-operated transistors MPR and M N R serve as the tunable resistors to make the G, values of the transconductance amplifiers tunable. The left G, amplifier formed by MP3 and MN3

is connected as a shunt-shunt feedback to the open-loop

R , amplifier to achieve lower input impedance, lower out- put impedance, and wider amplifier bandwidth. Thus, the performance of the close-loop R, amplifier can be en- hanced. The right G,n amplifier receives the voltage signal from the input R, amplifier and converts it into the output current.

Due to the symmetric push-pull structure of the input R, amplifier, the small-signal voltages at nodes 1 and 3 are nearly equal. Thus they can be merged into one node in the small-signal equivalent circuit. Similarly, the nodes 4 and 5 can be merged together. The resultant small-sig- nal equivalent circuit of the current amplifier is shown in Fig. 2 where g, and gd are the transconductance and the drain conductance of the MOS devices, respectively. As- sume g,

>>

gd for all MOS transistors, the dc gains of the open-loop R, amplifier and the left feedback G, am- plifier can be expressed as

1 1

(1)

gdnl

+

gdn2 + gdpl + gdp2

R,

=

Similarly, the gain of the right output G, amplifier is ( 3 )

g d r

where gdr is the output conductance of the transistors MPR and MNR operated in the linear region.

MPZ MN1 MP1 MN2 MPR MP3 MP5 MN3 MN5 MNR

H*M

Fig. 1. The new linear wideband current amplifier.

Iin 2 6 Iout g.2+ + c 2 c 4 g m k g m p l + g ” l g d l = gdpl+gdnl gm3 = gmp3+gmn3 gd13 = gdp3+gdn3 gd2 i:gdpZ+gdnZ gm5 = gmpS+gmnS gd5 = gdpS+gdnS gdlr = gdprcgdnr C2 = Cgdp2+C~;dnl+Cdbp2+Cdbnl+Cgsp3+C~~3+C~pS+Cgd~+Cgdpl+Cd~~nZ+Cdbpl+Cgsn3+CgsnS C3 = Cgsnl+Cslbnl+Cdbp3+Cppl+~sbpl+Csbpl+Cdbn3 C4 = CgdpS+CgdnS CS = CgdpS+Cg:dn3 C6 = CdbpS+CdbnS

Fig. 2. The small-signal equivalent circuit of the current amplifier shown in Fig. 1.

From Fig. 2 , the gain, input resistance, and output re- sistance of the current amplifier in Fig. 1 can be easily derived as

(3)

616 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 31, NO. 5 , M A Y 1996 Vin ., 2 MP4 I t - Vcn (b)

Fig. 3. (a) The new linear wideband voltage amplifier. (b) The block dia- gram of the linear wideband voltage amplifier.

where Gm3 and G,, are given in (2) and ( 3 ) , respectively. As may be seen in (4), the current gain of the current amplifier depends upon the G, ratio of the two GI, ampli- fiers. Thus the current gain may be unity or finite value, depending upon the transistor geometric ratio of the two G, amplifiers. According to the HSPICE simulation re- sults, the values of rln and rout in the designed current am- plifier are 7.42 ohm and 5.75k ohm, respectively.

As shown in Fig. 1, the two triode-operated MOS- FET's MPR and M N R are connected between the power supplies and the source nodes of the transistors in both G,, amplifiers. They act as common tunable resistors for the two G, amplifiers. Their resistances can be adjusted by simply changing the control gate voltages V,, and V(:p. These tunable source series resistances change the gate- source voltages of the MOS transistors and thus change the G, values of the G, amplifiers. If the dimensions of

MN5 (MP5) and MN3 ( M P 3 ) are the same, the current gain almost remains unity independent of the tuning.

The circuit diagram and the block diagram of the pro- posed linear wideband voltage amplifier are shown in Fig. 3(a) and (b), respectively. In this voltage amplifier, the input voltage signal is converted to the current signal through the G, amplifier and then the current signal is converted to the voltage signal through the close-loop R , amplifier. The input tunable G, amplifier has a CMOS inverter structure with the triode-operated MOSFET's MPR and M N R controlled by the voltages Vcp and V,,, respectively. The close-loop R , amplifier has a comple- mentary source follower to reduce the output resistance.

Vout - - g m l = gmplcgmnl gdl = gdpl+gdnl gd2 = gdp2+gdn2 gd3 = gdp3+gdn3 gd4 = g d p l c g d d g&=gdnS+gdp5 C2 = CgdnZ+CdbnZ+Cgdpl+Cdbpl+Cgdp4+Cgbp4+Cgdp2+CdbpZ+Cgdnl+Cdbnl+Cgd~4+Cgbn4 C3 = Cdbn3+Cdbp3+CgspI+Cgsnl+Csbnl+Csbpl C4 = Csbn4+Csbp4+Cgsn3+Cgsp3+Cgbn3+Cgbp3 C6 = CdbpSiCdbn5 C5 = Cgsn4+Cgs@+Cgdn3+Cgdp3 C7 = CgdpS+Cgdn5 gm3 = gmp3+gmn3 gm4 = gmp4+gmn4 gm5 = gmpS+gmnd

Fig. 4. The small-signal equivalent circuit of the voltage amplifier shown

i n Fig. 3.

It also has a feedback tunable G, amplifier connected from the source follower output to the input of the close-loop RI,, amplifier so that the output resistance can be further decreased.

The small-signal equivalent circuit of the voltage am- plifier in Fig. 3 is shown in Fig. 4. Assuming g,

>>

gd for all transistors, the voltage gain and the output resis- tance of the voltage amplifiers can be derived similarly as

those of the current amplifier. They can be expressed as

The voltage amplifier has an infinite input resistance. The HSPICE simulated value of rout of the voltage amplifier is as low as 2.84 ohm.

111. VHF BIQUAD A N D HIGH-ORDER LOWPAS FILTER DESIGN

A . VHF Current Biquadratic Filter

To analyze the high-frequency behavior of the current amplifier, the small-signal equivalent circuit of Fig. 2 with device capacitances is used. Assuming g ,

>>

gd for all transistors and neglecting of higher-order non-dominant terms, the transfer function can be expressed as

iout(s)

-

-D

H(s) = 7 - -

(4)

WU A N D HSU: DESIGN OF CMOS CONTINUOUS-TIME FILTERS

B gml(C2

+

C,)

+

G m 3 G (11)

c

S m l G w l 3 (12)

D

=

g m , G m s . (13)

Apparently, the transfer function expressed in (9) is the form of current lowpass biquad with the center frequency coo and the quality factor Q expressed as

The maximum-gain frequency f M denotes the frequency at which Iff( j27rf)

I

has a peak and the maximum-gain qual- ity factor Q M denotes the gain at f M . Both fM and Q M are

defined by the formulas

1

1 - 3

It is seen from (14) and (15) that the center frequency

wo and the quality factor Q can be tuned by adjusting the control voltages V,, and V,,, to change Gm3. Since G,,* is also simultaneously tuned as described in Section 11, the gain of the current filter, which is equal to Gms/Gm3 from (12) and (13), remains almost unchanged during the tun- ing.

In the small-signal equivalent circuit of Fig. 2, there exist parasitic zeros which are usually very far away from

fM so that they can be neglected in deriving (9). According

to the HSPICE simulation results on the current biquad of Fig. 1 with V,, = - Vcp = 2.5 V and without any extemal capacitor, the maximum-gain frequency f' is as high as 225 MHz and the nearest LHP zero is 14.6 GHz which is about 70 times larger thanf,. However, to generate a far- ther LHP zero from f M even under worst-case process variations or operation temperature, three external linear capacitors denoted as Cfixed may be added to the nodes of 1, 2, and 3 of the current biquad of Fig. 1. This decreases f M and keeps the nearest LHP zero about 100 times larger thanfM. Since the value of Cfixcd is not critical, it can be realized by a simple MOS device.

With V,, = -Vcp = 2.5 V and Chxed = 0.5 pF at the nodes 1, 2, and 3 of Fig. 1, the HSPICE simulated max- imum-gain frequency fM (center frequency f o ) and maxi- mum-gain quality factor Q M (pole quality factor Q ) are

153.69 MHz (218.49 MHz) and 1.15 (1.002), respec- tively. The corresponding transfer function can be ex- pressed as

-1.866 X IOi8

(1 8) H(s) = s 2

+

1.37 x 109.r

+

1.885 x 10l8

617

From the IHSPICE simulated output list files, the node capacitances and transistor transconductances of the cur- rent VHF filter can be calculated. The callculated values including thiree external linear capacitors are C2

z

3.6 pF, C 3

=

2.43 pF, C4 z 0.0274 pF, C5 z 0.0274 pF, gml z 0.0024 mho, and Gm3

=

0.00525 mho.

Since C2

>>

C4, C 3

>>

C5, and Grm3 E gm3, (14) and (15) can be further simplified as

Using these simplified formulas, the centier frequency f o and pole quality factor Q can be easily calculated by hand. The calculated center frequency& (cal) is about 191 MHz, and the pole quality factor Q (cal) is about 1.2149. As compared with the HSPICE simulation results, the devia- tions are 12.6 % and 2 1.2 % , respectively. Since these de- viations can be compensated by tuning, (1'9) and (20) can be used as the design equations to design the geometric dimensions of the MOS transistors.

Assume thiat the channel lengths and widths of the MOS transistors A4P2, MN2, MPR, and MNR are fixed and the geometric dimension of the MOS transistor MN3 (MP3) is equal to that of MN5 (MP5) under the consideration of unity current gain. Equations (19) and (20) can be re- written as

g m 3

=

woQC2 (21)

where

+

? C p n < w P l -t wNl)ws

+

Cpn(wP3

+ wN3)wd-

(24) In the above formulas, Cp, is the source-bulk (drain-bulk)

junction capacitance which is dependent upon junction bias voltage. W, is the typical distance between junction edge and poly gate edge. Further expressing g m l and g,,

(5)

618 IEEE J O U R N A L OF SOLID-STATE CIRCUITS, VOL. 31, NO. 5, MAY 1996 as functions of device currents and dimensions, (21) and

(22) can be written as

I I

where p N ( p P ) is the surface mobility. If the channel lengths of the MOS transistors are fixed and the device parameters, the biasing currents 11 and 13, and the desired center frequency f o and quality factor Q are given, the channel widths W P 3 , W N 3 , WPl , and WNl can be calculated from (25) and (26).

B. VHF Voltage Filter

To analyze the high-frequency behavior of the voltage amplifier in Fig. 3, the small-signal analysis of the volt- age amplifier including device capacitances is performed by using the small-signal equivalent circuit in Fig. 4. As- suming.gm

>>

gd for all transistors and neglecting of higher-order nondominant terms, the transfer function can be expressed as (27), shown at the bottom of the page.

From (27), it is seen that the transfer function is the form of voltage lowpass biquad with the pole frequency

wo and the pole quality factor Q expressed as

The tuning method of w o and Q is the same as that of the current-mode lowpass biquad.

With V,, = -Vcp = 2.5 V , the HSPICE simulated maximum-gain frequency fM (center frequency

fo)

and maximum-gain quality factor QM (pole quality factor Q)

are 149.5 MHz (197.7 MHz) and 1.18 (1.044), respec-

Fig. 5 . The block diagram of the new Q enhancement circuit

tively . Thus, the corresponding transfer function can be expressed as

-1.521 X 10l8

s2

+

1.189 x 109s

+

1.543 X 10l8' (30)

H(s) =

C. New Q-Enhancement Circuit

As can be seen from (14) and (15), if the capacitance of C 2 at the node 1 or node 3 is reduced, the center fre-

quency f o and the quality factor Q can be increased. If the reduction quantity can be adjusted,

fo

and Q can be tuned. Here a new circuit is proposed to reduce C 2 and enhance the Q value.

If a capacitor CM is connected across the input and the output of a voltage amplifier with the voltage gain A,, as

shown in Fig. 5, the equivalent capacitance seen from the input node of the amplifier is CM(1 - A ( , ) , known as the

Miller Effect. If the voltage gain of the amplifier is posi- tive and larger than one, the effective input capacitance seen from the amplifier input node is negative, which can reduce the overall capacitance at the input node.

Based upon the above circuit concept, the node capac- itance C2 at the node l(3) can be reduced by connecting the input of a positive-gain wideband voltage amplifier to the node l(3) and keeping its output floating. The Miller capacitor CM is connected between the node l(3) and the floating output node. The positive voltage gain of the am- plifier can be adjusted to tune& and Q.

The proposed positive-gain wideband voltage amplifier is shown in Fig. 6. The amplifier is constructed by cas- cading two NMOS common-source amplifiers. In the first (second) amplifier, the transistor MN1 (MN3) acts as the common-source stage and the transistor MN2 (MN4) serves as the enhancement load. The voltage gain of the single-stage voltage amplifier is the transconductance ra- tio of the transistors MN1 (MN3) and MN2 (MN4). The triode-operated transistor MNR (MNRl) which is con- nected between the source of the transistor MN1 (MN 3) and the power supply V,,, acts as a tunable resistor with the control gate voltage Vcq. By varying Vcq, the g, value

(6)

WU AND HSU: DESIGN OF CMOS CONTINUOUS-TIME FILTERS 619 W (um) L(um) MNR MNl MN2 MNRl MN3 MN4 400 98 49 400 250 80 1.0 1.0 1.0 1.0 1.0 1.0

of the transistor MN1 (MN3) is changed and the gain of the amplifier is also changed. Thus, the gain of the volt- age amplifier is adjustable by varying the control gate voltage Vc.q.

The special Q-enhancement circuit of Fig. 5 is applied to the nodes 1 and 3 of the current amplifier shown in Fig. 1. Thus the current filter with

fo

and Q enhancement is constructed. According to HSPICE simulation results, if Vdd = - VT,, = 2.5 V and changes from 0 V to 2.5 V, the dc gain of the voltage amplifier varies from 1.30 to 1.74, the -3 dB frequency varies from 1.56 GHz to 1.27

GHz, and the power dissipation of the Q-enhancement

circuits varies from 190 mW to 200 mW.

When the control voltage Vc4 changes from 0 Volt to 2.5 Volt and the Miller capacitance CM = 1.5 pF, the HSPICE simulated maximum-gain frequency f M (center frequency fo) changes from 144 MHz (166.2 MHz) to 205.6 MHz (207 MHz) whereas the maximum-gain qual- ity factor QM (pole quality factor Q,,) from 1.44 (1.429)

to 14.2 (14.1). With suitable transistor geometric ratio of the voltage amplifiers in the Q-enhancement circuits, the gain of the amplifiers can be as high as 1.87 and the -3 dB frequency as high as 1.3 GHz under 218 mW power dissipation. Thus, the maximum-gain quality factor QM of the current filter with Q enhancement can be as high as 145.54. If the gain of the amplifier is increased further, the effect of the Q-enhancement circuit will overcompen- sate the node capacitances. In this case, the RHP poles will be generated and the system will be unstable. How- ever, suitable choosing CM can avoid the unstability. Fig.

7 shows the HSPICE simulation results of the maximum- gain frequency fM and the maximum-gain quality factor

QM of the current filter with Q enhancement versus the power dissipation of the Q-enhancement circuits. Note that both Vc4 and V,,(V,,) can be used to iteratively tune f M and QM of the filter so that the desired values can be

reached. n 2.200 10

I

2ooO1o8 - 1800 10'

I

-

E -3 1.600108

-

Power Dissipation (Watts)

Fig. 7. The HSPICE simulation results of the f M and QM of the current

filter with Q enhancement versus the power dissipation of the Q-enhance- ment circuits.

D.

High-Order Filter Design

Since the ratio of routlr,, is large (small) enough, the proposed current (voltage) biquads can be cascaded di- rectly to form the high-order current (Voltage) filters. As a demonstrative example, a fourth-order Chebyshev cur- rent lowpass filter is designed by using the proposed VHF current biquadratic lowpass filter as the basic building block. Here, the current filter with Q enhaincement is used as the basic block since this kind of filter is controlled by the two independent tuning control voltages Vc4 and V,,(V,). Thus, the desired values of f o and Q can be reached by iterative tuning.

The fourth-order Chebyshev lowpass filter has 0.5 dB ripple in the passband and a cut-off frequency 190 MHz. From the filter handbook, the transfer function of the fourth order lowpass Chebyshev filter is

1.063563~:

-

- ( s 2

+

0.350706w,p

+

1.063%)

0.356461~;

( r ' i - 0 . 8 4 6 7 9 6 ~ ~ ~

+

0 . 3 5 m ) . (30) In the implementation of the transfer function H C f , ( s ) , a biquadratic lowpass filter with Q = 2.94 and

fo

= 196 MHz is needed. Similarly, for the transfer function H C f 2 ( s ) , a lowpass biquad with Q = 0.705 andfo = 113.4 MHz is designed. In designing the two biquads, (25) and (26) are used to determine the device dimensions of the current filters. Then tuning is used to obtain the desired With V,, = -Vrp = 1.0 V and Vc4 = 1.3 V, the HSPICE simulated center frequencyfo of the current filter with Q enhancement is 197 MHz and the pole quality fac- tor Q is 3.0. Thus this biquad can be used to implement the transfer function H c f , ( s ) . With V,, = - Vcp = -0.65 V and Vcq == -0.3 V, the HSPICE simulated center fre- quency f o is 114.6 MHz and the pole quality factor Q is f o and

Q.

(7)

... 0. I ... ---.. 2 0 . 0 -. ... - L q o , o s o . o ... ; ... : ... ... : ... : ... 6 0 . 0 E ... i ... . -. a n ~ ... ~ ... . , 9 0 ... 1 . 6 0 ~ ... . ~ . . - l . g o e ... ... Frequency (a) ... 2 . 0

_

... Frequency (bj

Fig. 8. (a) The HSPICE simulated frequency response of the fourth-order Chebyshev lowpass filter. (bj The passband frequency response of the fourth-order Chebyshev filter.

0.694. This biquad is used to implement the transfer func- tion Hcf2(s). Finally, these two lowpass biquads are cas- caded directly and then the fourth-order Chebyshev low- pass filter with the cut-off frequency 190 MHz is designed. The HSPICE simulated frequency response of the fourth-

order Chebyshev lowpass filter is shown in Fig. 8(a)

where the passband frequency response is separately shown in Fig. 8(b).

IV. EXPERIMENTAL RESULTS

The proposed lowpass current and voltage biquads have been designed and fabricated in 0 . 8 pm N-well double- poly-double-metal CMOS technology. Suitable layout techniques have been used to reduce the extra circuit par- asitics. Also some extra on-chip circuits have been added to ease the measurement of the fabricated filters in the VHF range.

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 31, NO. 5 , M A Y 1996

VSS

VHF Filter

V38

Fig. 9. The on-chip measurement circuit schemes for the fabricated VHF voltage filter.

Fig. 10. The measured frequency response (gain and phase) of the fabri- cated VHF voltage lowpass biquad filter with CL = 0.2 pF, when the con- trol voltages V,,, = - V C p = 2.5 V.

The on-chip measurement circuit schemes of the fab- ricated voltage-mode lowpass filters are shown in Fig. 9. Here, a CMOS wide-band gain stage serving as an output buffer for the VHF filters is used to decrease the loading effect of the parasitic capacitances at the output pad. Be- sides, the on-chip reference path with a single output buffer is designed. Thus its frequency response can be measured for the compensation of the measured frequency response of the filters and the buffer so that the real filter response can be obtained. After compensating the re- sponse of the reference path, the measured frequency re- sponse of the VHF voltage lowpass biquad is shown in Fig. 10. For CL = 0.2 pF and V,, = -Vcp = 2.5 V, the measured maximum-gain frequency f M is about 141.8 MHz and the maximum-gain quality factor QM is 1.19. Compared with the HSPICE simulation results offM (sim)

= 149.05 MHz and QM (sim) = 1.18, the deviations 5.15 % and 0.84 % , respectively. These deviations mainly

(8)

WU AND H S U : DESIGN OF CMOS CONTINUOUS-TIME FILTERS Simulated Values (HSPICE) 62 1 Experimental Values 1.600 lo8 1.4 1.35 fM 1.2 Q M 1 1 5

-

Qu (mea) 1.1 1 0 5 U Qu (rlm) 1.OOO 108 - 8OOO10’ - 0 0 5 1 1.5 2 2.5 3 6 OOO 10’ Vcn= -Vcp

Fig. 1 1 . Comparison of the simulated and measured results o f the fabri- cated VHF voltage lowpass biquad with CL = 0 2 pF versus dlfferent con- trol voltages V,,, and V,,,.

Power Dissipation ( Vcn=-Vcp=2,5V) 20 mW Capacitance Loading

1

0.2pF

1

0.2pF

1

I

Max-gain Frequency fM :Vcn= -Vcp=2.5V) 149.05 MHz 141.8 MHz Max-gain Quality Factor QM (Vcn=-Vcp change

I

(Vcn= -vcp change 1.18/1.04

1

1.19/1.05 from 2.5V to 0.2V)

I

Max. Output Swing for 1 % THD

(Vcn= -Vcp=2.5 V) 3 8 . 9 m V m s

Total Passband Noise

I

I

123uVrms

I

Dynamic Range

( vcn=-vcp=2.5v)

I

result from the impreciseness in the parasitic capacitances estimation due to process variations.

The frequency deviations may be post-tuned by adjust- ing the control voltage Vcn(VCP) of the unity-gain voltage amplifier. Fig. 11 shows the comparison of the simulated and measured f M and QM of the fabricated VHF voltage

lowpass biquad versus the control voltages Vcn and Vcp. The tuning range of the maximum-gain frequency is as high as 49 MHz. The equivalent maximum output signal

of the biquad with f M = 141.8 MHz is about 38.9 m V m s . The measured total passband noise is about 123 uVrms. Therefore, the dynamic range is about 50 dB. The power consumption of the VHF voltage lowpass filter, not in- cluding the output buffer is about 20 mW for k 2 . 5 V power supply. Table I summarizes the comparison results

--+-

I

--

Fig. 12. The on-chip measurement circuit schemes o f the fabricated cur- rent lowpass biquad filter.

between measured characteristics and HSI’ICE simulated results o f the VHF voltage-mode lowpass biquad.

The on-chip measurement circuit schernes of the fab- ricated current-mode lowpass biquad are shown in Fig. 12. The voltage lowpass filter in Fig. 3(b) is used as the interface circuit to measure the fabricateld VHF current filter. The voltage lowpass filter circuit can be separated into two parts. The first part is the G, amplifier, and the second one is the close-loop R, amplifier. The input volt- age signal is converted to the current siginal through the G, amplifier and flows into the current filter. The filter output current is then converted to voltage signal by the close-loop F:, amplifier and the voltage signal is sent to the output pad through a CMOS voltage bluffer.

Moreover, the on-chip reference interface circuit path constructed by voltage lowpass filter and voltage buffer is used to compensate the measured frequency response of the current filter and the interface circuits. After compen- sating the response of the reference path, the measured frequency response of the VHF current lowpass biquad is shown in Fig. 13. For CL = 1.4 pF and V,, = -Vcp = 2.5 V, the measured maximum-gain frequency f M is about 147.8 MHz and the maximum-gain quality factor Q M is

1.49. As compared with the HSPICE simulation results

O f f M (sim) = 153.7 MHz and QM (sim) = 1.15, the de-

viations are 3.84 % and 29.5 % , respectively.

Fig. 14 shows the comparison of the simulated and measured f M and QM of the fabricated VHF current low- pass biquad versus the control voltages V,,, and Vcp. When

V,, = -Vcp changes from 2.5 V to 0.2 V, the measured

f M ( Q M ) of the fabricated VHF current lowpass biquad

changes from 147.8 MHz (1.49) to 102.5 MHz (1.13). The tuning range of the maximum-gain frequency is as high as 45 MHz. Fig. 15 shows the total harmonic dis- tortion measurement (1 % THD) of the VI-IF current low- pass filter atid the on-chip measurement circuit of Fig. 12 In Fig. 15, the measured maximum output voltage swing for 1 % THD is 35.4 mVrms. Since the total transresis- tance of the close-loop R, amplifier and the CMOS volt- age buffer is about 384.8 ohm, the equivalent maximum output current signal of the biquad withf,,, = 147.8 MHz

(9)

622

I 400 IO8

1 200 lo8

1 OOO IO*

fM

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 31, NO. 5, MAY 1996

- 2 9 -fu(mea)

-

~ ~ ( s i m ) - Q M ( w ~ ) I21 2 5

/ -

+Qu(sim) QM - - t

Fig. 13. The measured frequency response (gain and phase) of the fabri-

Variations of QM

(Vcn= -Vcp change

from 2.5V to 0.2V) cated VHF current lowpass~with CL = 1.4 pF, when ;he control voltages

V,,, = - V I i’ = 2 . 5 V.

1.15 / 1.05 1.49/ 1.13

Fig 16. Noise characteristics of the fabricated VHF current lowpass bi- quad filter and the on-chip measurement circuit of Fig. 12.

1.600 IO8 I

,

3.3

TABLE I1

BIQUAD W I T H (V,,,, = -V,% = 2 . 5 V)

SIMULATED 4 N D MEASURED RESULTS OF THE VHF CURRENT LOWPASS

(HSPICE) Capacitance Loading 1.4pF 1.4pF 1 3 o---o---_o--- 0 9 8.OOO 10’ 1 1 5 2 2 s 3 Vcn= -Vcp

Fig. 14. Comparison o f the simulated and measured results of the fabri- cated VHF current lowpass biquad with CL = 1.4 pF versus different con- trol voltages V,,, and V,,,.

Fig. 15. The total harmonic distortion measurement (1 % THD) of the VHF current lowpass biquad filter and the on-chip measurement circult of Fig. 12.

is about 92 uArms. The distortion is partially generated by the output common-source amplifier whose linearity is not inherently high.

Fig. 16 shows the noise characteristics of the fabricated

ax-gain Frequency fM 153.7 M ~ z

I

147.8MHz Max-gain Quality Factor QM

I

1.15

1

1.49 (Vcn= -VCP=~SV)

I

Tunable Range of fM

I

I

(Vcn=-Vcp change

1

153.7 Il07.9 MHz

I

147.8 / 102.5 MHz

I

from 2.5V to 0 . W Max. Output Swing for 1% THD

(Vcn= -Vcp=2.5 V) 92 u A m s

1

Total Passband Noise

I

1

347.8nAms

Dynamic Range

( VCn=-Vcp=2,5V)

I

48.5dB

Power Dissipation

I

(VCn=-Vcp=2.5V)

1

I

21.8mW

VHF current lowpass biquad and

the

on-chip measure-

ment circuit of Fig. 12. The total passband noise includ- ing that in the on-chip measurement circuit is about 347.8 nArms and the filter dynamic range should be larger than 48.5 dB. The power consumption of the VHF current lowpass biquad is about 21.8 mW for k 2 . 5 V power sup- ply. The measured characteristics and the simulation re- sults of the VHF current lowpass biquad filter are sum- marized in Table 11.

Using the same on-chip measurement circuit schemes for the current-mode lowpass biquad, the characteristics of the fabricated current biquad with the Q enhancement circuit can be measured. The measured frequency re-

(10)

WU AND HSU: DESIGN OF CMOS CONTINUOUS-TIME FILTERS

Factor QM 14.2

623

18.5

TABLE 111

SIMULATED AND MEASURED RESULTS OF THE VHI’ CURRENT Q

ENHANCEMENT LOWPASS BIQUAD WITH (Vdd = - V , , = 2.5 V)

(Vcn=-Vcp=Vcq=2.5V) Tunable Range of fM (Vcn=-Vcp=Vcq change 205.6 / 144.5 MHz 185.7 / 150.7 MHz from 2.5V to 0 V) Variations of QM (Vcn = -Vcp=Vcq (change 14.2 / 1.61 18.5 / 1.44 from 2.5V to 0 V)

Max. Output Swing for 1 % THD (Vm= -VCp =Vcq=2.5V)

Total Passband Noise 441.5 nArms

Dynamic Range Wcn

87.3 uAms

45.9 dB

= -Vcp=VCq:=2.5V) Fig. 17. The measured frequency response (gain and phase) of the fabri-

cated VHF current lowpass with Q enhancement biquad filter with CL =

1.4 pF, when the control voltages V,y = V,,, = - V C p = 2.5 V.

2.400 10’ I 340 2 o o o 1 ~ 8 -Qu(Ma) , ,

,1

id

QM f M 15 10 5 1600 lo8 1 400 IO8 1 200 lo8 0 -05 0 0.5 1 15 2 2 5 3 Vcq=Vcn= -Vcp

Fig. 18. Comparison of the simulated and measured results of the fabri- cated VHF current lowpass with Q enhancement biquad filter with CL =

1.4 pF versus different control voltages V,,,, V C p , dnd V<y

sponse of the VHF current filter with Q enhancement is shown in Fig. 17. For CL = 1.4 pF, CM = 1.5 pF, and

Vcq = V,, = - Vcp = 2.5 V, the measured maximum-gain frequency

f,,

is about 185.7 MHz and the maximum-gain quality factor QM can be enhanced to be 18.5. As com-

pared with the HSPICE simulation results f M (sim) =

205.6 MHz and

Q M

(sim) = 14.2, the deviations are 9.68% and 30%, respectively. Fig. 18 shows the com- parison of simulated and measuredfM and Q,, of the fab- ricated VHF current lowpass biquad with Q enhancement versus the control voltage V C q , V,,, and Vcp. When V,, =

V,, = - V c p changes from 2.5 V to 0 V, the measured

f M ( Q M ) of the fabricated VHF current lowpass biquad with

Q enhancement changes from 185.7 MHz (18.5) to 150.7 MHz (1.44), with the tuning range of the maximum-gain frequency as high as 35 MHz. The equivalent maximum output current signal withf, = 185.7 MHz is about 87.3 uArms. The totcll pdssbdnd noise is about 441.5 nArms and the dynamic range is about 45.9 dB. The power con- sumption of the VHF current filter with Q enhancement is about 221.4 mW for 1 2 . 5 power supply which is much larger than the current filter without Q enhancement. The increased Dower consumdon is mainlv due to the two

Simulated Values E x p e h ” t a l Value! (HSPICE)

I

Capacitance Loading

I

1.4pF

I

1.4pF pax-gain Frequency fM

I

205.6 M k

I

185.7 MHz (Vcn=-Vcu=Vca=2.5V)

I

Power Dissipation

I

(Vcn=-Vcp=Vcq=2SV)

1

221.4mW

Q-enhancement circuits. The measured characteristics and the simulation results of the VHF current filter with Q enhancement are summarized in Table 111.

V. CONCLUSIONS

CMOS tunable VHF current-mode and voltage-mode lowpass biquadratic filters designed by using unity-gain or finite-gain amplifiers have been proposed, analyzed, and fabricated. Both experimental and sirnulation results of the VHF lowpass biquadratic filters have successfully verified the performance. A new Q-enhancement circuit

consisting of a wideband tunable positive-gain voltage amplifier and a Miller capacitor is also proposed to en- hance both maximum-gain quality factor Q,, and maxi- mum-gain frequency f M of the filters. Experimental results have been slhown that the fabricated lowp,ass biquad with the Q-enhancement circuit has a highfM around 185 MHz and a Q M near 18. A fourth-order Chebyshev current-mode lowpass filter has been successfully desip,ned by cascad- ing two current-mode biquads. Further research on other applications of the biquads will be conducted in the future.

REFERENCES

[ l ] C . S . Park and R . Schaumann, “Design of a 4 M H z analog inte- grated CMOS transconductance-C bandpass filter,” IEEE J . Solid- State Circuits, vol. 23, pp. 987-996, Aug. 1988.

[2] H . Khorramabadi and P. R . Gray, “High frequency CMOS contin- uous-time filters,” IEEE J . Solid-State Circuits, vol. SC-19, pp. 939- 948, Dec. 1984.

[3] J . M . Khoury, “Design of a 15-MHz CMOS continuous-time filter with on-chip tuning,” IEEE J . Solid-state Circuits, vol. 25, pp. 1988-1997. Dec. 1991.

(11)

Y . P. Tsividis, “Integrated continuous-time filter design-An over- view,” IEEE J . Solid-State Circuits, vol. 29, pp. 166-176. Mar.

1994.

L. J . Pu and Y . P. Tsividis, “Transistor-only frequency-selective circuits,” IEEE J . Solid-State Circuits, vol. 2 5 , pp. 821-832, June

1990.

B. Nauta, “A CMOS transconductance-C filter technique for very high frequency,” I E E E J . Solid-State Circuits, vol. 27, pp. 142-153, Feb. 1992.

F. Krummencher and N. Joehl, “A 4-MHz CMOS continuous-time filter with on-chip automatic tuning,” IEEE J . Solid-State Circuits.

vol. 23, pp. 750-758, June 1988.

Y . T. Wang and A. A . Abidi, “CMOS active filter design at very high frequencies,” IEEE J . Solid-state Circuits. vol. 25, pp. 1562-

1574, Dec. 1990.

F. J. Femandez, “Linear CMOS transconductance elements,” US Patent No. 4 , 734, 654, Mar. 29, 1988.

M. A. Tan and R. Schaumann, “Simulating general-parameter LC ladder filters for monolithic realizations with only transconductance element and grounded capacitors,” IEEE Trans. Circuirs Syst.. vol.

36, pp. 299-307, Feb. 1989.

S . S. Lee, R. H. Zele, and D. J . Allstot, “CMOS continuous-time current-mode filters for high-frequency applications, ” IEEE J . Solid-

State Circuits, vol. 2 8 , pp. 323-329. Mar. 1993.

R . H. Zele, S. S. Lee, and D. J . Allstot, “A 3V-125 MHz CMOS continuous-time filter,” in Proc. IEEE Int. Symp. Circuits Sjsr., May

1993, pp. 1164-1 167.

T. S. Fiez, G . Liang, and D. J . Allstot, “Switched-currents circuit design issues,” ZEEE J . Solid-State Circuits, vol. 26, pp. 192-202,

Mar. 1991.

I. Ramirez-Angulo, M. Robinson, and E. Sanchez-Sinencio, “Cur- rent-mode continuous-time filters: Two design approaches,” IEEE Trans. Circuits Syst. IZ, vol. 39, pp. 337-34,1, June 1992.

S. L. Smith and E. Sanchez-Sinencio, “3V high-frequency current mode filters,” in Proc. IEEEInt. Symp. CircuitsandS~ut.. May 1993. pp. 1459-1462.

J . Ramirez-Angulo and E. Sanchez-Sinencio, “High frequency com-

pensated current-mode ladder filters using multiple output OTAs.” in Proc. IEEE Int. Symp. Circuits and Syst., May 1993, pp. 1412-

1415.

H. Khorramabadi and P. R. Gray, “High frequency CMOS contin- uous-time filters,” IEEE J . Solid-State Circuits, vol. SC-19, DD. 939-

IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 31, NO. 5 , MAY 1996

[I91 P. H . Lu, C . Y . Wu, and M. K. Tsai, “VHF bandpass filter design using CMOS transresistance amplifier,” in Proc. IEEE Inc. Symp. Circuits Syst., May 1993, pp. 990-993.

I201 -, “Design techniques for tunable transresistance-C VHF band- pass filters,” IEEE J . Solid-State Circuits, vol. 29, pp. 1058-1066,

Sept. 1994.

[21] C. Y . Wu and H. S . Hsu, “The continuous-time VHF lowpass filter design using finite-gain current and voltage amplifiers and special @enhancement circuit,” in Proc. IEEE Int. Symp. Circuits and Syst.,

vol. 5 , May 1994, pp. 771-774.

Chung-Yu Wu (S’76-M’88) was born in Chiayi,

Taiwan, Republic of China, in I950 He received the M S. and Ph D. degrees trom the Institute of Electronics, National Chiao-Tung University, Hsinchu, Taiwan, in 1976, and 1980, respec- tively

From 1980 to 1984, he was an Associate Pro fessor in the National Chiao-Tung university During 1984-1986 he was a Visiting Associate Professor in the Department of Electrical Engi neering, Portland State University, Oregon Since 1987, he has been a Professor in the National Chiao-Tung University He

h d S published more than 60 journal papers and 80 conference papers on

several topics. including digital integrated circuits, analog integrated cir- cuits, computer-aided design, ESD protection circuits, special semicon ductor de\ices, and process technologies He also has eight patents includ- ing four U S patents His current research interests focus on low-voltage, low-power mixed-mode integrated circuit deaign, hardware implementa- tion of visual dnd auditory neural systems, and RF integrated circuit de sign

Dr Wu is d member of Eta Kappa Nu and Phi Tau Phi

Heng-Shou Hsu was born in Hsinchu, Taiwan,

Republic of China, i n 1967 Hc received the B S

degree from the Department of Electronics Engi neering, National Chiao Tung University, Hsin- chu, Tdluan, in 1990

Presently, he 15 working toward the Ph D de- gree dt the Department of Electronics Engineering

. .

948, Dec 1984.

1181 J Ramirez-Angulo, E Sanchez Sinencio, and M Howe, “Largefog second-order filters using multiple OTAS,” IEEE Truns Cilcults

S y A t , vol 41, pp 587-592, Sept 1994 low voltage high-frequency filter designs and the Institute of Electronics, National Chiao- Tung University His current research intere\ts in- d u d e CMOS current mode, voltage mode, and

數據

Fig.  1.  The new  linear wideband current amplifier.
Fig.  3.  (a) The new linear wideband voltage amplifier.  (b)  The block  dia-  gram of  the linear wideband voltage  amplifier
Fig.  5 .   The  block  diagram  of  the  new  Q enhancement  circuit
Fig. 7.  The HSPICE  simulation  results  of  the  f M   and  QM  of  the  current
+5

參考文獻

相關文件

• In Shutter-speed priority mode, photographers sets the shutter speed and the camera deduces the aperture. • In Program mode, the camera decides both exposure and shutter

• In Shutter-speed priority mode, photographers sets the shutter speed and the camera deduces the aperture. • In Program mode, the camera decides both exposure and shutter

1997 年 IEEE ELECTRONICS LETTERS 曾有學者 A.Motamed 、 C.Hwang 以及 M.Imail 提出一篇 CMOS Exponential Current-to-Voltage Converter[7],主要 是利用

Abstract—We propose a multi-segment approximation method to design a CMOS current-mode hyperbolic tangent sigmoid function with high accuracy and wide input dynamic range.. The

Abstract - A 0.18 μm CMOS low noise amplifier using RC- feedback topology is proposed with optimized matching, gain, noise, linearity and area for UWB applications.. Good

FMEA, fail mode and effective analysis, which is one of a common method to analysis and find out the fail mode of the product is to dig out the unobservable problem or hidden

Finally, making the equivalent circuits and filter prototypes matched, six step impedance filters could be designed, and simulated each filter with Series IV and Sonnet program

GaN transistors with high-power, High temperature, high breakdown voltage and high current density on different substrate can further develop high efficiency,