methodology
Heiner Alarc on Cubas
and Jo ao Navarro
LSITEC-S ao Paulo
Email: heiner.alarcon@lsitec.org.br
AbstractThis paper presents the design of an OTA-Miller
amplier of the rst integrator of a Switched-Capacitor Multi-
bit Sigma-Delta Modulator. The rst integrator OTA is the
most critical block in Sigma-Delta Modulator due to the high
bandwidth, high Slew Rate and low noise requirements. The
rst integrator OTA specications are obtained from the Sigma-
Delta Modulator designing for low power consumption. The
gm/ID methodology is used on the OTA design of the rst
integrator to reduce the power consumption. This methodology
is also applied in the other OTAs of the Sigma-Delta Modulator.
The Chopper technique is also implemented to reduce the input
referred noise of the rst integrator. The SDM with the designed
OTAs using the gm/ID methodology is simulated by the Spectre
simulator. Implemented in 0.18 CMOS technology, the SDM
achieves a 96 dB SNR for 20 kHz signal bandwidth and a power
consumption of 2.77 mW for a 1.8 V supply.
Keywords:/
).
The Miller architecture is chosen to implement the OTA 1.
In this work, a methodology design for the OTA 1 of a
Third-order Four-bit CIFF (Cascade of Integrators Feedback
Form) SDM is presented. In section II, the SDM circuit is
presented. In section III, the OTA 1 specications are obtained.
In section IV the OTA 1 design is developed. Finally in section
V and VI, the results of the OTA 1 and the Sigma-Delta
modulator and the conclusions are presented respectively.
II. SDM CIRCUIT IMPLEMENTATION
The Third-order Four-bit CIFF SDM shown in the gure 1 is
composed by three SC integrators, a capacitive resettable ana-
log adder with offset cancellation, a four bits ash quantizer,
and a four bits weighted-DAC. The OTA ampliers are applied
in the three DC integrators (OTA 1, OTA 2 and OTA 3) and in
the analog adder (OTA 4). Two chopper modulators are added
at the rst integrator to reduce the in-band icker noise [2].
A Fully differential structure is selected to reduce the charge
injection generated by the switches and to increase the input
value range and the PSRR [3]. A bottom plate technique [1]
is also implemented to reduce the parasitic switch capacitance
error. The switches are implemented with carefully designed
transmission gates to reduce the charge injection and time
constant.
III. OTA 1 SPECIFICATIONS
The input OTA specications are obtained from the SDM
system design. The SDM is designed for high SNR (about
96 dB), so it is necessary to establish the noise budget of
the system. The noise sources can be divided in internal
and external sources. The internal sources are due to the
quantization noise (
). The external
sources (
2
,
2
+
2
,
+
2
+
2
(1)
In equation 1,
,
is the maximum allowed input
signal RMS and
,
is the sum of the intrisic RMS noise of
the OTAs and KT/C RMS noise. As the
,
determines the
978-1-4673-4900-0/13/$31.00 c 2013 IEEE
Fig. 1. Third-order Four-bit SDM
SDM power consumption, it is assumed that
,
contribuits to
80% of the total noise. The
and the
contribuit
to 1%, 9% and 10%, respectively [4]. For
is possible to
determine the SQNR=20(
,
/
) should be equal to
2 MHz for a bandwidth slightly greater than 20 kHz. Using
the found OSR and the modulator order (L), the modulator
coefcients can be obtained using the Matlab SDM toolbox
[5]. These coefcients are listed in the table I.
TABLE I
SDM COEFFICIENTS
3
2
1
1
2
3
4
1
2
3
3.4 2.4 0.7 0.8 0 0 1 0.8 1.4 1.1
The coefcients are scaled to limit the output swing to 20%
of the full-scale at each output integrator and are rationalized
to implement the capacitance feedback of each integrator [4]
[5].
The rst integrator has the greatest noise contribution, so
it is xed as
2
1
= 0.9
2
,
. The input referred noise of the
OTA 1 is xed in 10%
2
1
, because of the used Chopper tech-
nique. The rst integrator / noise can be expressed as
4/(
1
2) [4], where C2 is the integration capacitance
of the rst stage.
The OTA non-idealities are commanded by the nite DC
gain (
is desirable.
The
/(
/6).
is the maximum
output value peak to peak of the integrator and
is the
oversampling period.
The remaining time of the integration phase is taken by the
settling time, then
= (()
)/(2
factor is equal to
1/(1 +
), being
a SDM coefcient.
An important parameter on the OTA 1 design is the output
load capacitance,
=
2(2
1
+
2
+2
1
+
(2)
where
), 2,
,
and
Methodology
The OTA 1 is designed using the /
methodology.
In this methodology the design parameters are expressed in
function of /
[7]. The /
and
vs
= 420 and
= 355
amplier. To increase the Common Mode Gain it is used two
CMFB (Common Mode Feedback). A SC symmetrical load
architecture is applied for the two CMFBs. This architecture
stabilizes faster than traditional SC-CMFBs.
Fig. 3. Fully Differential Two Stage OTA including CMFB
Most of the current consumption is due to the second stage
current,
7
(shown in gure 3). This current can be expressed
in function of
7
/
7
,
2
, as can be seen in the equation 3.
7
=
2
(
7
/
7
)
(3)
The Phase Margin (PM) depends on the
2
and can be
expressed by the equation 4.
= 90
1
(
2
) (4)
PM must be designed for 85
to assure stability, so
4
transistors are designed for the same inversion region of
the
7
transistor to ensure low systematic offset [1].
In order to reduce the power consumption without increase
the current mismatch, it is necessary a current copy factor of 3
as maximum in the current mirrors. For minimize the current
consumption, the following relation is assumed.
7
= 3
5
= 6
1
(5)
where
1
is the differential pair current and
5
is the
current of the transistor
5
. The compensation capacitance
can be obtained from the equation 6.
=
2
1
(6)
The current
1
can also be expressed in function of
and
1
/
1
as it is shown in the equation 7 [1].
1
=
2
1
/
1
(7)
To dene the inversion region of
1
, the parameter
1
/
1
is varied between 0 and 30 in the equation 7. The
former curve is compared with the obtained current (
1
=
7
/6) being both curves shown in the gure 5.
From the gure 5 we can deduce that
1
/
1
should
not be greater than 20, because
1
would be less than the
specied current by 5. Finally,
5
,
6
and
9
transistors are
designed in strong inversion reducing the mismatch between
current mirrors.
The values of
1
/
1
,
5
,
,
7
and
7
/
7
for
the OTA 1 are summarized in the table III.
0 5 10 15 20 25 30
0
0.2
0.4
0.6
0.8
1
1.2
1.4
x 10
3
gm1/ID1
I
d
1
(
A
)
I
D7
/6
fta=7,3MHz
Fig. 5.
1
for xed
of 1 in function of
1
/
1
TABLE III
SUMMARIZED DESIGN RESULTS FOR THE OTA 1
1
/
1
5
7
7
/
7
1 20 120 20 360 20
V. RESULTS
The shown in the gure 1 was designed using the
0.18m CMOS7RF CMOS technology of IBM. The power
supply is 1.8 and the maximum input is 1 (50% of Full
Scale). The input OTA results listed in the table IV, including
input referred offset (
),
are obtained using the Spectre simulator.
TABLE IV
SUMMARIZED RESULTS FOR THE OTA 1
[] 10 24 24 20
10
1
10
2
10
3
10
4
10
5
10
6
160
140
120
100
80
60
40
20
0
A
m
p
l
i
t
u
d
e
[
d
B
]
Frequency[Hz]
SDM Output Spectrum
Input Referred Noise PSD
Fig. 6. Output SDM frequency response compared with the input referred
PSD noise
VI. CONCLUSIONS
A /