Sie sind auf Seite 1von 4

A PROCESS-INDEPENDENT THRESHOLD VOLTAGE INVERTER-

COMPARATOR FOR PULSE WIDTH MODULATION APPLICATIONS


Meng-Tong Tan, Joseph S. Chang and Yit-Chow Tong
School of Electrical and Electronic Engineering
Nanyang Technological University
Singapore 639798, Singapore
Email: p6917985a@ntu.edu.sg (Meng-Tong Tan)

ABSTRACT We derive the output voltage V Racross


~ load RLas:
This paper presents a novel self-tuning inverter-
VR,.= 2 ( 2 k - l ) + 2 0 ~ 0 ~ ~ 0 , t - (1)
comparator for a pulse width modulator. The inherent
threshold voltage of the inverter-comparator is made where k is the duty cycle of the PWM signal at zero-
independent of temperature and process variations by input, D is the modulation index and a.\ is the frequency
means of a self-tuning mechanism. The measured duty of the input signal.
cycle error of the PWM at zero-input is typically 0.5%; From equation (I), we note that one of the major
an error 5 2% is typically specified. This is achieved drawbacks of such a configuration is the need for the
without any post-fabrication trimming. The PWM is duty cycle k of the PWM at zero-input to be as close to
realized by a very simple circuit and can be readily 50% as possible. This requirement is essential because
fabricated in a low-cost digital CMOS process. The the duty cycle error is translated into an output DC static
prototype IC inet all design specifications and is suitable current (or voltage) flowing into the load, and the power
for micropower low-voltage applications including efficiency of the amplifier will subsequently be
hearing instruments. substantially reduced if this DC current is large.
The simplified schematic of a simple PWM design
1. INTRODUCTION [3] is shown in Figure 2. In this circuit, a triangular-like
In portable micropower low-voltage electro- carrier is generated by integrating a 50% duty cycle
acoustical instruments including auditoryprostheses and square wave Q with a simple RC integrator (R4 and Cm).
hearing instruments (hearing aids) [l-81, the critical The input modulating signalV, is added to the carrier via
parameters include low voltage and micropower resistor R3. The comparator used in this design is a
operation, and a small IC area realization. To meet the simple CMOS digital inverter (MN2 and MP2) with an
micropower requirements of hearing aids, the Class D inherent threshold voltage, V T H which , is dependent on
amplifier is often used [2-81 as the final power amplifier. process and temperature variations. When the sum of
This is because Class D amplifiers feature high the input signal and carrier is higher than VTHc, the
efficiency (typically >80% over a large range of output of the inverter will transit from high to low, and
modulation indices (equivalent to signal swing)) [2] and vice-versa. The resultant output signal from the
is well suited for low voltage operations. comparator is a Pulse Width Modulated signal (PWM
A block diagram of a typical Class D amplifier is signal).
depicted in Figure 1. It consists of a .Pulse Width
Modulator (PWM) and an output stage. Due to the low I V
supply voltage (1.3V), the output stage is usually
configured as bridge output to increase the power (by s
four times over a single-ended output).

Figure 2 A current-art Pulse Width Modulator.

To reduce the output DC static current, a laser is


Pulse Width MOdulalor
used to trim one of the two bias resistors, R I or Rz, to
cL,L,,= LC filter adjust the DC bias of the carrier to VTHc. This post-
cis,* D 0 Y t p " t stag.
R b = Load ResinLance
fabrication laser trimming process is, however,
expensive and labor intensive.
Figure 1 Block diagram of a full-bridge output Class D
Other methods reported to reduce the output static
power amplifier.
current include a complex and slower (switching speed)

0-7803-5682-9/99/$10.000 1 999 I E EE. I201


differential-input comparator in BiCMOS process [4] Table 1 Inherent VTHCdeviations and the resultant
and a complex differential approach [ 5 ] . output static current due to threshold voltages
To eliminate this post-fabrication trimming variations of the nMOS and PMOS transistors.
process, we had also proposed two methodologies- Threshold Voltage, O/P Static
namely, a self-error correction method by digital means Case V T H (VI % “ ~ x Current
Deviations
[6] and a self-tuning PWM based on Master-Slave nMOs MOS (PA)*
architecture 171 to reduce the output static current. In 1 0.751 0.756 0 0
both designs, the inherent threshold of the inverter- 2 0.68 0.65 -0.5 -20.8
comparator (similar to the one used in Figure 2) is 3 0.88 0.65 -8.7 -362.5
dependent on process variations (see Section 2 later), 4 0.68 0.85 7.4 308.3
and a self-tuning mechanism is used to adjust the DC 5 0.88 0.85 -0.7 -30
bias of the carrier to V T ~ ~ .
A better approach to reduce the output DC static
current is to design an inverter-comparator with stable
inherent threshold V T Hset ~ to a pre-determined voltage.
In this application, V T ~should
, be set to half of the
supply voltage ( W D o ) because at ? h V D D , the leading and
trailing edge of the carrier will be symmetrical. A
programmable threshold voltage inverter for CMOS ,
-

programmable logic circuits was recently proposed [9].


However, the maximum inherent threshold deviation of 5
the inverter from the mean threshold is 8.5% (or
equivalent to an unacceptable output static current of
8.5% full load current), too large for high efficiency
Class D amplifier applications. The IC area overhead of
the digital circuits will also be large if a large number of
bits are used to program the inherent threshold voltage.
In this paper, we proposed a novel self-tuning
inverter-comparator for the PWM to realize the above Figure 3 DC transfer characteristics of the inverter-
approach. The inverter-comparator based on Master- comparator due to process variations.
Slave architecture is capable of self-tuning its own
inherent threshold voltage VTHCto %VDD- independent From Table 1 and Figure 2, we note that the
of process and temperature variations. Post-fabrication inherent threshold voltage varies from -8.7% to 7.4%
adjustment or calibration is therefore unnecessary. This (with respect to the typical value in Case 1) due to the
circuit design methodology employs very simple tuning threshold variations of the nMOS and PMOS transistors.
circuitry and can be readily fabricated in a low cost In the worst case, an unacceptably large DC static
digital CMOS process. current of 362.5pA flows into the load, thereby
substantially reducing the power efficiency of the Class
2. AN INVERTER AS A VOLTAGE D amplifier.
COMPARATOR
In a micropower low voltage Pulse Width 3. A PROPOSED SELF-TUNING INVERTER-
Modulation design, the simple digital inverter is one of COMPARATOR FOR THE PWM
the best approaches to realize the voltage comparator for The schematic of the proposed self-tuning inverter-
two reasons [lo]. First, the slew rate of the comparator comparator for the PWM is depicted in Figure 4. The
is not limited by the biasing current. Second, it offers Master circuit comprises a simple bias circuit (?I,,, and
the maximum possible charging and discharging current R2& and a self-tuning inverter-comparator which is
for a given supply voltage. However, one of the major capable of self-tuning its own inherent threshold to the
drawbacks of this inverter-comparator is that the voltage applied at its input (node DC-M). The self-
inherent threshold voltage VTHCis dependent on the tuning inverter-comparator consists of an inverter
threshold voltage variations of the nMOS transistor (MPlm and MNlm), and an nMOS and a PMOS
(nMOST) and the PMOS transistor (pMOST). transistor (MPlam and MNlam) at the supply rails. The
The VTHcdeviations of the inverter due to the inverter performs the logic switching whereas the two
nMOS and PMOS transistors threshold voltage transistors at the supply rails are used as variable
variations in a typical CMOS process are tabulated in resistors to compensate for the inherent threshold
Table 1. The resultant output static current at the load of voltage variations of the inverter. The two transistors at
the Class D amplifier is also tabulated in Table 1. For a the supply rails are biased in the triode region, and their
better graphical illustration of the V T H C variations, we output resistances is varied by the output voltage of the
have also plotted the corresponding DC Input/Output inverter-comparator. The transistors of the Slave
transfer characteristics of the comparator in Figure 2. inverter-comparator (comprising MPlas, MPls, MNls
and MNlas) are designed with the same dimensions as

I202
the Master to obtain the same inherent threshold as the small AC gain (with good phase margin) to prevent
Master. undesired oscillation.
V"" The self-tuning inverter-comparator can also be
r - - - - - -T-- -, used for other applications which requires a stable
inherent threshold voltage. The inherent threshold can
be easily changed by changing the bias at nodeDC-M
I
I
m via the two bias resistors (RI,,, and Rz,) which have a
,U
+
matching accuracy better than 0.2%.
I'WM Output
The schematic diagram of the proposed self-tuning
PWM is depicted in Figure 6. The 3 succeeding inverter
stages connected to the output of the Slave inverter-
, - - - - - -=- - ..-I
- - Slave comparator are used to increase the voltage gain of the
inverter-comparator and also to provide sufficient drive
to drive the Class D output stage.

The inverter-comparator self-tunes its own inherent


for the

!
U
threshold in a DC feedback loop manner and operates as
follows. The desired inherent threshold voltages of the
Master and the Slave inverter-comparator at node
DC-Car and DC-M respectively, are set by biasing
node DC-M with a resistive divider comprisingRI, and
Rz,. In this design, the inherent threshold of the inverter
comparator is set to %VD, as discussed in Section 1. I
I
._-
With its input set to % V ~ Dthe , Master inverter-
comparator will generate the required output voltage to Figure 6 The proposed self-tuning PWM.
tune the resistance of the two transistors at the supply
rails. Through this negative feedback mechanism, the At the input of the comparator, there is a possibility
inherent threshold of the Master and Slave inverter- that any unwanted noise present would result in chatter
comparator will be tuned to 1/2VoD, independent of or high frequency oscillation at the output PWM signal.
temperature and process variations. To eliminate this chatter, two capacitors, C, and C,b, are
Consider a case when the inherent threshold of the added to improve the noise immunity at the input of the
Master inverter-comparator is below '/Voo. The output inverter-comparator (node DC-Car) [3]. The positive
voltage of the Master comparator will be lower than feedback action of capacitor C , also has an added
?4VD,, resulting in an increase in the drain-to-source advantage of improving the rise and fall times of the
resistance of the nMOST (MNlam) and a decrease in PWM signal at the output of the inverter-comparator.
the drain-to-source resistance of the pMOST (MPlam).
Due to the changes in resistance at its rails, the inherent 4. SIMULATION AND EXPERIMEMTAL
threshold of the comparator will rise towards '/zVDD. RESULTS
This negative feedback action will continue (i.e. the The proposed self-tuning PWM is simulated and
output voltage of the Master comparator will increase) fabricated using a 1.2 pm CMOS process. For the
until the inherent threshold is approximately ' ~ " D D . comparator inherent threshold deviations given earlier in
Similarly, when the inherent threshold is above 'NOD, Table 1, the PWM embodying the proposed inverter-
the inverter-comparator will tune its inherent threshold comparator as depicted in Figure 6 was simulated. The
to % VDo through the negative feedback mechanisms. percentage deviation of VTHCfrom the nominal %VDo
For stability, a Miller capacitor C,. is introduced and the resultant output bias current at the load of the
between the input and output of the Master inverter- Class D amplifier are tabulated in Table 2. The
comparator to create a dominant pole at node DC-M. corresponding DC transfer characteristics of the self-
This dominant pole will provide a large DC gain for the tuning inverter-comparator are plotted in Figure 5 for
inverter-comparator to minimize the tuning error and a comparison with Figure 2.

I 203
Measurements show that by means of the self-
tuning mechanism, the 50% duty cycle error at zero-
input is typically 0.5% (or equivalent to -1lpL.A output
Case Threshold Voltage, % VTHc O/P Static bias current at 1.3V operation). This compares well
VTH(VI Deviations Current with the design in Figure 2 which has a zero-input duty
I nMOS I PMOS I I (pN* cycle of 2% after employing the expensive post-
1 1 0.751 I 0.756 I 0.10 I 4.0 fabrication trimming. The total current consumption of
2 0.68 0.65 0.18 7.5 the proposed prototype PWM is 20 pA which is only
3 0.88 0.65 0.25 10.5 0.5% of the full load current, thereby ensuring quiescent
4 0.(ix
__ 0.85
. .. 0.08 3.5 ~~

micropower operation. If realized in a low threshold


5 1 0.88 I 0.85 1 0.06 I 2.5 voltage process, the circuit consumes only lOpA at
* vDD=2.5v, RL=60W,Carrier Freq=40 H z . vDD=1.3v. This makes the proposed PWM suitable for
......................................... ~

micropower low voltage applications including hearing


aids.

5. CONCLUSIONS
..,
..~,._I ..^.,.
We have presented a novel self-tuning inverter-
comparator based on Master-Slave architecture for a
PWM. The measured duty cycle error at zero-input is
typically 0.5%. This is achieved without employing the
expensive post-fabrication trimming. The PWM is
realized by very simple circuit and can be readily
fabricated in a low-cost digital CMOS process. The
prototype IC met all design specifications, and is
suitable for micropower low-voltage applications
including hearing instruments.

REFERENCES
Figure 5 DC transfer characteristics of the self-tuning [I] J.S. Chang and Y.C. Tong, “A low-power time-
inverter-comparator due to process variations. multiplexed switched cap. speech spectrum analyser,”
IEEE J. Solid-state Circuits. 28, pp. 40-48, 1993.
Comparing Tables 1 and 2, we note that the self- [2] M.T. Tan et al., “Analysis and two proposed design
tuning PWM has effectively self-tuned the DC bias of methodologies for optimizing power efficiency of a Class
the carrier to the inherent threshold of the comparator - D amplifier output stage,” IEEE Int. Symp. Circuits &
independent of the threshold voltage variations of the Systems, 1998.
nMOS and PMOS transistors. This significant [3] M.C. Killion and G. Village, “Class D hearing aid
improvement will be better appreciated by comparing amplifier,”US Patent No. 4,689,819, 1987.
[4] H.A. Gurcan, “Class D BiCMOS hearing aid output
the DC transfer characteristics of the inverter-
amp,” USPatent No. 5,247,581, 1993.
comparator in Figures 2 and 5. The worst case [5] J.F. Duque-Carrillo et al., “VERDI: An acoustically
simulated output static current in Table 2 is 1 0 5 p A programmable & adjustable CMOS mixed-mode signal
(VDD=2.5V), about 35 times better than the worst case processor for hearing aid applications,” IEEE J. Solid-
output static current given in Table 1. State Circuits, 31, pp. 634-645, 1996.
The simulation and measurement results of the [6] M.T. Tan et al., “ A novel self-error correction pulse
self-tuning PWM for the remaining parameters are width modulator for a Class D amplifier for hearing
tabulated in Table 3. instruments,”IEEE Int. Symp. Circuits & Systems , 1998.
[7] M.T. Tan, J.S. Chang and Y.C. Tong, “A novel self-
Table 3 Simulation and measurement results of the tuning pulse width modulator based on master-slave
architecture for a Class D amplifier,”IEEE Int. Symp.
self-tuning PWM (vDD=2.5v, RL=600L2, Circuits & Systems, June 1999.
Carrier Freq=40 kHz). [8] F. Calliaset al., “A set of four IC’s CMOS technology for

I ’pea I
Results
Simulations I Measure-
Ments
a programmable hearing aid,” IEEE J. Solid-state
Circuits, 24, pp. 301-312, 1989.
[9] J. Segura et al., “A variable threshold voltage inverter for
Current <40@ 17.8 pA 20 pA CMOS programmable logic circuits,”IEEE J. Solid-State
Consumption Circuits, 33, pp. 1262-1265, 1998.
50%Duty <2% 0.25 %* 0.5 % [IO] E.A. Vittoz, “Dynamic analog techniques,” in Design of
.Cycle Error analog-digital VLSI circuits for telecom. C? signal
at zero-input processing, Edit. J. Franca and Y. Tsividis, Prentice Hall,
Total <5% 1% 2.5 % pp. 97-124, 1994.
Harmonic
Distortion

I 204

Das könnte Ihnen auch gefallen