Download Advanced Common-Mode Control techniques for Low

Advanced Common-Mode Control Techniques
for Low Voltage Analog Signal Processors*
Edgar Sánchez-Sinencio1, José Silva-Martínez1, and J. Francisco Duque-Carrillo2
[email protected]
1
Texas A&M University, Dept. Electrical Engineering,
College Station, TX 77840, USA.
2
University of Extremadura, Dept. Electronics and Elec. Eng.,
(06071) Badajoz (Spain).
Abstract
The control of the common-mode (CM) component of two voltage signals for low voltage (LV) analog
signal processors, is addressed in this tutorial. In particular it is shown that for large input and output
signal swings, the CM component must be well controlled in very LV Fully-Differential (FD) amplifiers.
Then, appropriate common-mode feedback (CMFB) loops are unavoidable needed. The general
requirements of such CMFB networks are derived and design solutions for input and output LV CMFBs
are revised. In addition, an efficient CMFB loop suitable for OTA-C based circuits is described.
1 Introduction
Scaling down of modern CMOS technologies is imposing strong limitations in the total supply voltage
that integrated circuits (IC) can sustain, and all the expectations indicate that this trend will continue in
future [1]. The impact of this scaling on the IC performance, is rather different as a function of the circuit
nature. Thus, in a digital IC, as devices dimensions shrink, more and more devices can be fabricated on
the same die, the parasitic capacitances also tend to decrease, and current density increases. The
combined effect is a reduction of the propagation delay, which allows higher throughput and clock rates
for digital circuits. In contrast, from a general point of view, technology scaling has a negative impact on
the performance of analog IC blocks. Thus, scaling down the transistor dimensions degrades the intrinsic
gain of the devices and, what is even more important, the very limited voltage room available reduces the
circuit dynamic range (DR) with respect to the DR counterparts using higher supply voltages. Therefore,
in such low supply voltage (LV) conditions, analog circuits require larger input/output voltage signal
swings, enhanced linearity, and larger rejection to undesired signals, to keep a similar performance with
respect to the operation at higher supply voltages. In other words, many analog design issues, which were
unimportant only a decade ago, are now of primordial importance and new circuit topologies and design
strategies must be investigated for supply voltages in the order of one MOS transistor threshold voltage
[2]. In this tutorial, the control of the common-mode (CM) component of two voltage (differential)
signals in LV analog signal processors, is addressed. Traditionally, common-mode feedback (CMFB)
techniques have been applied for the control of the CM output component in (FD) circuits. However,
most of these techniques are not valid for very LV applications and hence, new design solutions are
needed. On the other hand, as the total supply voltage is reduced, the control of the input CM voltage is
increasingly important.
2 Why and where CM control is needed?
Indeed, the most important building blocks in analog signal processors are the operational amplifier and the
Operational Transconductance Amplifier (OTA). Typically, the amplifier input stage is based on
differential input pairs (Fig. 1).
--------------------------------------------------------------*
This work was partially supported by Texas Instruments, Inc
The main constrain for LV operation is imposed by the very limited input common-mode range (CMR) of
the differential pair, which in the case of Fig. 1 is approximately given by VDD-2⋅|VDsat-VTp.
Traditionally, to extend the input CMR up to the rails, input stages based on parallel-connected
complementary input pairs [2] have been used. However, these solutions fail for total supply voltages in the
order of one MOS transistor threshold voltage (i.e., 1 V). This is due to the fact that for such LV operation
only in a very small voltage range close to one of the power supply sources (close to the negative supply
rail in the case of Fig. 1), the input voltage level is able to turn on the transistors drivers. Therefore, for railto-rail input CMR a front-end section must be used to restrict the CM signals to this region. An input
feedback loop (ICMFB) might be used for this purpose. Figure 2 represents a very low-voltage rail-to-rail
operational amplifier with ICMFB in unity-gain configuration. Vref,i accounts for an appropriate reference
CM voltage compatible with the CMR of the amplifier itself. On the other hand, one of the standard
techniques to extend the DR in analog blocks consists of using amplifiers with also differential output (FD
amplifiers). In such cases, the output swing is improved, as well as the rejection to all common-mode
interference signals. For very LV applications the input and output CM components of FD amplifiers must
be well controlled. As before, the input CM component must be set to any value compatible with the
amplifier CM range (i.e., Vref,i). The output CM component must be also controlled and set to the middle
value between supplies for maximum output signal swing. The reason of this control for the output CM
component arises from the fact that the external feedback around the amplifier only affects to the DM
component, while the information regarding to the CM component sent back is rejected by the input
differential circuitry of the amplifier. As consequence, the output CM component is operating in open-loop
mode and therefore, uncontrolled. Hence, an extra CM feedback (OCMFB, Output CMFB) is required to
fix the output CM component to a predetermined voltage value (Vref,o), which generally coincides with the
operating point where the amplifier differential gain reaches its maximum value. Figure 3 shows a LV FD
amplifier with input and output CMFB and external resistive feedback.
3 General requirements of a CMFB loop
Regardless of a CMFB is intended for controlling the input or output CM component in FD amplifiers,
the basic structure is the same in both cases. It consists of a CM sense circuit, which ideally provides an
output signal proportional to the average of the signals applied at the inputs. The output of the CM sense
block is compared with an appropriate reference voltage in an error amplifier. This error signal (voltage
or current) is adequately injected to close the negative feedback loop. Strictly speaking, the following
requirements must be satisfied by both ICMFB and OCMFB loops:
(i) The dc gain of the CMFB loop must be large enough to keep an accurate control of the CM
component. If not, an asymmetrical swing occurs which entails a loss in the DR.; (ii) The gain-bandwidth
product of the CMFB loop (LGBWCM) should be at least equal to the gain-bandwidth product of the DM
loop (LGBWDM) counterpart. If LGBWCM,o < LGBWDM, the system does not perform as a fully-balanced
system in the range of frequencies comprised between LGBWCM,o and LGBWDM; while if LGBWCM,i <
LGBWDM occurs, the amplifier does not operate properly since the input CM voltage can be out of the
amplifier input CMR in the frequency band comprised between LGBWCM,i and LGBWDM.; (iii) The
CMFB loop must only act over CM voltage signal, and should not affect the DM signals, otherwise
harmonic distortion components due to the CMFB loop are induced in the signal. Although this is a
common requirement for both ICMFB and OCMFB loops, it is much more difficult to achieve in the
latter case, since the input DM range extends almost from rail-to-rail, whereas for ICMFB the input DM
signal is very small, due to the external negative feedback around amplifier. In practice, the DM linearity
is one of the main constraints specially in LV OCMFB. The opposite situation generally occurs respect to
the CM linear range. In ICMFB a large input CMR is required compared with the OCMFB counterpart.
4 LV input CMFB loops
As stated above and regardless if an LV amplifier topology is either singled-ended or double-ended, the
input CM voltage must be set to an appropriate dc reference voltage value, Vref,i, compatible with the
reduced amplifier CMR. A negative CMFB loop can be used for such purpose. Figure 4 illustrates the
circuit implementation of an ICMFB. As observed, the amplifier input CM voltage is sensed by an error
amplifier. This CM sense technique [2,3] presents a very reduced DM linear range, however, unlike
OCMFB loops, the DM voltage swing in ICMFB is small due to the external feedback around amplifier.
Next, the CM voltage is compared with Vref,i. The output of the error amplifier controls four identical
current sources IS, in such way that the product IS⋅RS coincides with Vi,cm - Vref,i, where Vi,cm accounts for
the input CM voltage signal. Notice that each couple of current sources IS along with a passive resistor
RS implement a floating voltage source [7], which value depends on the existing CM component at the
input circuit terminals. On the other hand, it can be easily demonstrated that in Fig. 4, Vi,cm is equal to
CM signal at the input terminals, divided by the gain of the ICMFB loop. Needless to say that, in order to
guarantee a rail-to-rail input swing, the ICMFB must be designed to provide a minimum dc loop gain
equal to the full range voltage swing (i.e., VDD) divided by the amplifier input voltage range. With this
technique rail-to-rail operation with a 1-V total supply in a standard CMOS technology, can be designed
[6,7].
5 LV output CMFB loops
A) OCMFB loops for buffered LV FD amplifiers: CMFB networks for the control of the output CM
voltage component in FD amplifiers, have been addressed [3]. However, most of the traditional CM
sense circuits cannot be used in very LV applications. The key problem is due to the FD amplifier
voltage signals, which are the input of the CM sense block, are centered around VDD/2 and for very LV
supply this value is lower than VT. Therefore, no CM detector circuits, where the amplifier output signals
have to be connected to MOS gate terminals, can be used. A possible design solution to overcome this
limitation is drawn in Fig. 5 [5]. As observed, a couple of matched passive resistors are used to sense the
output CM signal and generate a current i1. This current is compared with i2, which is generated by the
desired output reference voltage (i.e., Vref,o = VDD/2). Both currents are compared and converted in a
control voltage, which can be used to control an appropriate bias currents of the FD amplifier, closing the
CM loop to provide the required stabilization of the CM voltage signal. This very LV CM sensing
technique presents outstanding advantages, such as linearity and rail-to-rail capability, however, the FD
amplifier needs to be buffered to avoid the negative impact of the passive resistors on the DM gain.
B) OCMFB loops for unvuggered LV FD amplifiers: The typical fully-differential integrator based on a
differential pair circuit requires a common-mode feedback system, which consists of a common-mode
detector, and a very simple OTA that compares the common-mode voltage with the common-mode
ground. As mentioned earlier, to avoid additional harmonic distortions the common-mode detector must
be linear within the OTA linear range; sometimes the common-mode detector is even more complex than
the OTA input stage. Typically the common-mode loop presents several parasitic poles. Eventhough the
gain-bandwidth product of the CMFB loop could be as large as the GBWDM of the open loop differential
gain its phase margin is often worse due to the several parasitic poles. To alleviate this problem the
common-mode loop transconductance gain must be relaxed, resulting in large dc offsets and limited
rejection to common-mode noise signals. An important property of OTA-C based circuits is that the
outputs of the previous stage are usually directly connected to the inputs of the next stage [8]. This
property can be exploited to arrange the CMFB, because the common-mode voltage of the previous stage
is automatically sensed by the input stage of the next OTA, as shown in Fig. 6. The CMFB can easily be
arranged by comparing the common-mode voltage (VC) with a proper reference. The amplifier can be
implemented by using a PMOS differential pair and a current mirror or a class AB amplifier. The
advantage of the circuit shown in Fig. 6 is that the common-mode detector is embedded in the differential
circuitry, therefore the linearity of the CMFB is similar to the linearity of the OTA's differential
transconductance. Also, silicon and power are saved. For LV applications, the pseudo differential circuits
are very often used [7]. Typically, a drawback of these structures is its poor PSRR. A pseudo differential
OTA with high rejection to supply VSS supply noise and CMFB is shown in Figure 7. The commonmode error is converted into current by the resistors; this current is sensed by MC and mirrored to the
outputs of the previous OTA. The CMFB transconductance loop gain is nearly equal to the differential
transconductance gain, and determined mainly by the resistors.
6 Conclusions
The importance and applications of common-mode circuits have been presented. The LV constraints on
circuits have forced designers [2, 4-9] to look for new approaches since conventional approaches are not
longer valid for low supply voltages. Strategies for closed loop (for Op Amps) and open loop (for OTA)
have been outlined. Further research work is needed in this field to deal with very high frequency LV
circuit design.
References
[1] Y. Taur, “The incredible shrinking transistor,” IEEE Spectrum (Special Issue in The 100-Million
Transistor IC), vol. 36, pp. 25-29, July 1999.
[2] S. Yan and E. Sánchez-Sinencio, “ Low Voltage Analog Circuit Design Techniques: A Tutorial,”
IEICE Trans. Fundamentals, vol. E83-A, No. 2, pp 179-196,February 2000.
[3] J. F. Duque-Carrillo, “Control of the common-mode component in CMOS continuous-time fully
differential signal processing,” Analog Integrated Circuits and Signal Processing, vol. 4, pp. 131-140,
Sept. 1993.
[4] J. F. Duque-Carrillo, J. L. Ausín, G. Torelli, J. M. Valverde, and M. A. Domínguez, “1-V rail-to-rail
operational amplifiers in standard CMOS technology”, IEEE J. Solid-State Circuits, vol. 35, pp. 33-44,
January 2000.
[5] S. Karthikeyan, S. Mortezapour, A. Tammineedi, and E. K. F. Lee, “Low-voltage analog circuit
design based on biased inverting opamp configuration”, IEEE T. Circuits and Systems II, vol. 47, pp.
176-184, March 2000.
[6] J. Ramírez-Angulo, R.G. Carvajal, J. Tombs and A. Torralba, “ Simple technique for Opamp
continuous-time 1V supply operation,” Electron Lett., vol. 35, No. 4, pp 263-264, Feb. 1999.
[7] R. Castello, F. Montecchi, F. Rezzi, and A. Baschirotto, “Low-voltage analog filters”, IEEE T.
Circuits and Systems II, vol. 42, pp. 827-840, Nov. 1995.
[8] E. Sánchez-Sinencio, and J. Silva-Martínez, “ CMOS transconductance amplifiers, architectures and
active filters: a tutorial,” IEE Proc.-Circuits Devices Syst., vol. 147, No.1, pp 3-12, Feb. 2000.
[9] J. Ramírez-Angulo, A Torralba, R.G. Carvajal and J. Toms, “Low-Voltage CMOS Operational
Amplifiers with Wide Input-Output Swing Based on a Novel Scheme”, IEEE T. Circuits and Systems I,
vol. 47, pp. 772-774, May 2000.
VDD
VBIAS1
Vi −
Vi +
VBIAS2
Fig. 1. A simple PMOS input stage
−
ICMFB
Vo
+
Vi
Vref,i
Fig. 2. Rail-to-rail very LV amplifier in unity-gain configuration with ICMFB
R2
Vo+
R1
−
Vref,i
ICMFB
+
R1
+
−
OCMFB
Vref,o
Vo−
R2
Fig. 3. FD LV rail-to-rail amplifier with ICMFB and OCMFB
VDD
Vref,i
Rs
+
Is
Rs
−
Is
+
+
−
−
Is
+
−
Is
Fig. 4. ICMFB (shadowed area) for rail-to-rail LV amplifier operation
VDD
vbias
M11
CMF
M12
2R
vo+
i1
R
i2
VDD/2
vo2R
M13
Fig. 5. A LV continuous CMFB circuit [5]
VDD
VB1
IB
VB2
IB
IB
IB
- +
IB
IB
VC
IB
VREF
IB
VSS
Fig. 6. Typical OTA connection in fully differential OTA-C based circuits. The commonmode voltage is obtained from the input of the following stage.
VDD
IB
IB
IB
IB
GND
R1
MC
IB
IB
IB
M1
IB
R1
M1
IB
VSS
Fig. 7. Pseudo-differential OTAs including the CMFB for the first one.