slyt204.pdf
(
283 KB
)
Pobierz
Texas Instruments Incorporated
Amplifiers: Op Amps
Auto-zero amplifiers ease the design of
high-precision circuits
By Thomas Kugelstadt
(Email: tk@ti.com)
Senior Systems Engineer, Industrial Systems
A wide variety of electronic applications deal with the
conditioning of small input signals. These systems require
signal paths with very low offset voltage and low offset
voltage drift over time and temperature. With standard
linear components, the only way to achieve this is to use
system-level auto-calibration. However, adding auto-
calibration requires more complicated hardware and soft-
ware and can slow down time to market for new products.
The alternative is to use components with low offset and
low drift. The amplifiers with by far the lowest offset and
drift available are the auto-zero amplifiers (AZAs). These
amplifiers achieve high dc precision through a continuously
running calibration mechanism that is implemented on-chip.
With a typical input offset of 1 µV, a temperature-related
drift of 20 nV/ºC, and a long-term drift of 20 nV/month,
these amplifiers satisfy even the highest requirements of
dc accuracy.
Today’s AZAs differ neither in form nor in the application
from standard operational amplifiers. There is, however,
some hesitation when it comes to using AZAs, as most
engineers associate them with the older chopper ampli-
fiers and chopper-stabilized amplifier designs. This stigma
has been perpetuated either by engineers who worked
with the older chopper amplifiers and remember the diffi-
culties they had with them, or younger engineers who
learned about chopper amplifiers in school but probably
did not understand them very well.
The original chopper amplifier heralded the beginning of
the new era of self-calibrating amplifiers more than 50 years
ago. This amplifier provided extreme low values for offset
and drift, but its design was complicated and expensive. In
addition, ac performance was limited to a few hertz of input
bandwidth accompanied by a high level of output noise.
Over the years, unfortunately, the term “chopper amplifier”
became a synonym for any amplifier with internal calibra-
tion capability. Therefore, AZAs, often wrongly designated
as chopper or chopper-stabilized amplifiers, are associated
with the stigma of the older chopper technique.
This article shows that the auto-zero calibration tech-
nique is very different from the chopper technique and is
one that, when implemented through modern process
technology, allows the economical manufacturing of wide-
band, high-precision amplifiers with low output noise. The
following discussion presents the functional principles of
the chopper amplifier, the
chopper-stabilized amplifier,
and the AZA. It then com-
pares the efficiencies of
low-frequency filtering
when applied to AZAs
and standard operational
amplifiers. Finally, three
application examples
demonstrate the use of an
AZA as a signal amplifier
and as a calibrating ampli-
fier in dc—and wideband
ac—applications.
The chopper amplifier
Figure 1 shows a simpli-
fied block diagram of a
chopper amplifier.
A dc input signal is
chopped into an ac voltage
and amplified by an ac-
coupled amplifier. A phase-
sensitive demodulator
converts the output of A
1
Figure 1. Chopping principle in the time domain
Oscillator
V
IN
V
1
V
2
V
3
V
4
Low-Pass Filter
A
1
V
5
S
1
–
S
2
A
2
V
OUT
Wideband
Amplifier
+
Integrator
R1
R2
V
IN
,V
OUT
β
V
1
V
2
V
4
V
5
V
OUT
V
3
with Offset
V
OUT
β
0
t
0
t
0
t
V
IN
0
0
t
t
0
t
0
t
Note: Labels for amplifiers such as A
IN
, G
B
, and A
M
are used to identify amplifiers in figures and to represent
amplifier gain in equations.
19
Analog Applications Journal
2Q 2005
www.ti.com/aaj
Analog and Mixed-Signal Products
Amplifiers: Op Amps
Texas Instruments Incorporated
back to dc. The demodulator consists of a switch S
2
that is
synchronously driven to S
1
. An integrator then smooths
the switch output and presents the final dc output.
The circuit benefits from high overall dc gain and low
baseband noise. The dc gain, being the product of the ac
stage and the dc gain of the integrator, easily reaches an
open-loop gain of 160 dB and reduces the gain error,
both sides of the odd harmonics of the chopper frequency.
The amplitudes of the harmonics and their sidebands
decrease following a 1/n function, with n indicating the
order of the harmonic.
The 1/f noise of A
1
present in the baseband adds to the
modulated input signal after the first modulation stage,
M
1
. The combined signal is amplified by A
1
and fed into
the demodulator, M
2
. The 1/f noise, experiencing its first
modulation through M
2
, introduces sidebands on both
sides of the odd harmonics of f
CH
. For the modulated
input signal, however, M
2
represents the second modulat-
ing stage. V
M1
is now demodulated, causing sidebands of
the input signal to occur around the even harmonics of
f
CH
. The input signal reappears in the baseband, and the
roll-off of the subsequent low-pass filter limits the base-
band to frequencies far below the chopper frequency.
Note that the AM does not change the spectral density
of the white noise. The residual baseband noise is there-
fore limited, low-frequency white noise.
Despite the small values for offset, drift and baseband
noise, this approach has some drawbacks. First, the ampli-
fier has a single-ended, noninverting input and cannot
accept differential signals without additional circuitry at the
front end. Second, the carrier-based approach constitutes
a sampled data system, and overall amplifier bandwidth is
limited to a small fraction of the chopper frequency. The
chopper frequency, in turn, is restricted by ac amplifier
gain-phase limitations and errors induced by switch
response time. Maintaining good dc performance involves
keeping the effects of these considerations small. Chopper
frequencies are therefore in the low-kilohertz range, dictat-
ing low overall bandwidth.
V
AA
OUT
12
,
to almost zero.
The baseband is defined as the maximum usable input
bandwidth. Baseband noise consists of the input offset
voltage (also known as dc noise), the 1/f noise, and low-
frequency white noise. The reduction of baseband noise
happens in several steps:
• Offset and drift in the output integrator stage are nulled
by the dc gain of the preceding ac stage.
• dc drifts in the ac stage are also irrelevant because
they are isolated from the rest of the amplifier by the
coupling capacitors.
• The 1/f noise of the ac amplifier is modulated to higher
frequencies via the demodulator.
Figure 2 clarifies the process of noise reduction by demon-
strating the effects of chopping in the frequency domain.
The chopping of the input signal constitutes an ampli-
tude modulation (AM), with the chopping frequency, f
CH
,
being the carrier, and the input voltage representing the
modulating signal. Both switches, S
1
and S
2
, are replaced
by the modulators, M
1
and M
2
.
V
M1
(f) in Figure 2 shows that the modulation of a square
wave causes sidebands of the input signal to appear on
Figure 2. Chopping principle in the frequency domain
f
CH
f
CH
V
M1
V
M2
V
IN
A
1
V
OUT
M
1
M
2
N(f)
V
Noise
f
V
IN
(
f
)
V
M1
(
f
)
V
M2
(
f
)
V
OUT
(
f
)
f/f
CH
f/f
CH
f/f
CH
f/f
CH
0
0
1
3
5
0
1
23
4
5
6
0
20
Analog Applications Journal
Analog and Mixed-Signal Products
www.ti.com/aaj
2Q 2005
Texas Instruments Incorporated
Amplifiers: Op Amps
The chopper-stabilized amplifier
The classic chopper-stabilized amplifier
solves the chopper amplifier’s low-bandwidth
problem. It uses a parallel path approach
(Figure 3) to provide wider bandwidth while
maintaining good dc characteristics. The
stabilizing amplifier, a chopper type, biases
the fast amplifier’s positive terminal to force
the summing point to zero.
Fast signals directly drive the ac amplifier,
while slow ones are handled by the stabilizing
chopper amplifier. The low-frequency cutoff
of the fast amplifier must coincide with the
high-frequency roll-off of the stabilizing
amplifier to achieve smooth overall gain-
frequency characteristics. With proper design,
the chopper-stabilized approach yields band-
widths of several megahertz with the low-drift
characteristic of the chopper amplifier.
Unfortunately, because the stabilizing amplifier
controls the fast amplifier’s positive terminal,
the classic chopper-stabilized approach is
restricted to inverting operation only.
In addition, the high residual output noise
of the chopper amplifier is amplified by the
fast amplifier’s noise gain. Keeping output
noise small dictates additional filter effort,
thus increasing complexity and cost of the
chopper-stabilized design.
The auto-zero amplifier (AZA)
Similar to the chopper-stabilized approach,
the AZA uses a main amplifier for wideband
signal amplification and a nulling amplifier for
offset correction. Figure 4 shows a block
diagram of the TLC2654, an AZA developed
by Texas Instruments in the mid-80s.
With the calibration path lying in parallel
with the signal path, both inputs of the main
amplifier are available for differential input
operation.
The main amplifier, A
M
, and the nulling amplifier, A
N
,
each have an associated input offset voltage (V
OSM
and
V
OSN
, respectively) modeled as a dc offset voltage in series
with the noninverting input. The open-loop gain of the
signal inputs is given as A
M
and A
N
. Both amplifiers also
have additional voltage inputs with the associated open-
loop gains of +B
M
and –B
N
.
Offset correction of the overall amplifier occurs within
one cycle, f
AZ
, of the auto-zero clock and is split into two
modes of operation: an auto-zero phase and an amplification
phase. The oscillator, generating f
AZ
, initiates the auto-zero
phase by driving both switches into position 1. The inputs
of the nulling amplifier are shorted together, while its out-
put is connected to capacitor C1. In this configuration A
N
Figure 3. Chopper-stabilized amplifier
Summing
Point
R1
R2
V
IN
–
Wideband
Amplifier
+
V
OUT
Stabilizing
Amplifier*
*Similar to the chopper amplifier in Figure 1
Figure 4. Simplified TLC2654 block diagram
V
OSM
+
V
IN
A
M
V
OUT
–
+B
M
Oscillator
2
V
OSN
2
V
C2
+
1
A
N
S
1
1
–
–B
N
S
2
V
C1
C2
C1
Note: Labels for amplifiers such as A
IN
, G
B
, and A
M
are used to identify amplifiers in figures
and to represent amplifier gain in equations.
measures its input offset voltage and stores it via C1.
Mathematically we can express the voltage at C1 as
VAV BV
C
=
−
,
1
N
OSN
N
C
1
which, by simple rearrangement, is
A
SN
N
VV
=
.
(1)
C
1
1
+
B
N
This shows that the offset voltage of the nulling amplifier
times a gain factor appears at the output of A
N
and thus
on the C1 capacitor.
In the amplification phase, when both switches are in
position 2, this offset voltage remains on C1 and essentially
corrects any error from the nulling amplifier. A
N
amplifies
21
Analog Applications Journal
2Q 2005
www.ti.com/aaj
Analog and Mixed-Signal Products
Amplifiers: Op Amps
Texas Instruments Incorporated
V
C1
by the factor B
N
and subtracts it from the amplified
input signal,
input offset voltage of the complete amplifier, we should
set up the equation for the generic amplifier in Figure 5:
(5)
V
=
k V
(
+
V
),
AV V
NIN
(
+
).
OUT
IN
OS
_
Eff
SN
where k is the open-loop gain of the amplifier and V
OS_Eff
is its effective offset voltage.
Putting Equation 4 into the form of Equation 5 gives us
At the same time, the output of A
N
charges capacitor C2 to
VVAVV
==
(
+
)
−
BV
.
ON
C
2
N
IN
OSN
N
C
1
Replacing V
C1
with Equation 1 results in
V
+
V
OSM
OSN
V
=
A
B
V
+
.
OUT
N
N
IN
B
VAV
V
N
OSN
N
=
+
.
(2)
C
2
N
IN
1
+
B
From here it is easy to see that k = A
N
B
N
and
V
+
V
Equation 2 shows that V
OSN
has been reduced by a factor
1 + B
N
, indicating how the nulling amplifier reduces its
own offset voltage error even before correcting the main
amplifier. The potential, V
C2
, now serves the main amplifier
as an offset correcting voltage, forcing its output, and thus
the output of the complete AZA, to
OSM
OSN
V
=
.
OS
_
Eff
B
N
Thus, the offset voltages of both the main and the
nulling amplifiers are reduced by the gain factor B
N
. If we
consider the open-loop gains of the local amplifiers, A
N
and A
M
, to be in the region of 10,000 or higher, it quickly
becomes evident that even an inherent offset voltage of
millivolts is reduced to an effective input offset voltage of
microvolts for the complete AZA.
The AZA constitutes a sampled data system. The process
of sampling therefore generates frequencies consisting of
the sum and difference of the input signal frequency, f
S
,
and the auto-zero clock frequency, f
AZ
. The summing
frequency, f
AZ
+ f
S
, can be filtered easily and is therefore
of little importance. However, the difference frequency,
f
AZ
– f
S
, can alias into the baseband if f
S
≥
V
=
A
(
V
+
V
)
+
B
V
.
OUT
M
IN
OSM
M
C
2
Replacing V
C2
with Equation 2 and combining terms gives us
AB
B
NM
N
V
=
VAAB VAV
(
+
)
+
+
.
(3)
OUT
IN
M
N
N
OSM
M
OSN
1
+
The auto-zero architecture is optimized in such a way
that A
M
= A
N
, B
M
= B
N
, and B
N
>> 1. This allows Equation 3
to be simplified to
f
AZ
/2. Older AZA
designs therefore required the limitation of the input
bandwidth to less than half of the auto-zero frequency.
Most of the amplifiers available in the mid-80s had typical
clock frequencies in the range of only 400 to 500 Hz, thus
narrowing the signal bandwidth down to 250 Hz. The
TLC2654 was one of the first amplifiers that allowed high-
frequency auto-zeroing at 10 kHz, thus extending the
input bandwidth up to 5 kHz.
The breakthrough to real wideband operation happened
only with the recent introduction of AZAs such as the
OPA335. Modern process technology, with gate structures
in the submicron region, made the economical integration
of complex anti-aliasing circuitry possible. The strong
attenuation of alias frequencies enabled wideband opera-
tion across the entire amplifier bandwidth.
V
=
V
A
B
+
A
(
V
+
V
).
(4)
OUT
IN
N
N
N
OSM
OSN
Most obvious is the gain product of both the main and
nulling amplifiers. The A
N
B
N
term in Equation 3 explains
why AZAs have extremely high open-loop gain. To under-
stand how V
OSM
and V
OSN
relate to the overall effective
Figure 5. Generic amplifier with effective offset
V
OS_Eff
+
V
IN
V
OUT
k
–
22
Analog Applications Journal
Analog and Mixed-Signal Products
www.ti.com/aaj
2Q 2005
Texas Instruments Incorporated
Amplifiers: Op Amps
Figure 6 shows the inner structure of the OPA335. The
two nulling amplifiers, A
N1
and A
N2
, operate in an alter-
nate mode in parallel with the main amplifier, A
M
. While
A
N1
nulls its offset during the auto-zero phase, A
N2
is in
the amplification phase, correcting the main amplifier’s
offset voltage and vice versa.
The alternating operation of the nulling amplifiers mini-
mizes output voltage ripple and intermodulation distortion
(IMD) by keeping the amplifier’s gain bandwidth constant
during operation. Proprietary circuit design has made
further improvements to the nulling amplifiers. Each
amplifier consists of a multistage composite amplifier. This
configuration drastically reduces the quiescent current
down to 300 µA (versus the 1.5 mA of the TLC2654) while
maintaining a high open-loop gain of 130 dB. In addition,
the previous external capacitors have been made redun-
dant by achieving the same effective capacity values
through Miller equivalence.
Let’s return to the process of auto-zeroing. The nulling
amplifier, whose switches are in position 1, is in the auto-
zero phase, thus charging its capacitor to
GA
GA
C SN
BIN
BZ
VGAV
=
(
−
AV V
)
=
.
(6)
CBIN
SNZ
1
+
During the amplification phase (with switches in posi-
tion 2), the output voltage of the nulling amplifier, V
N
,
adds to the output voltage of the main amplifier, V
M
. With
VGAV V
=
(
+
)
−
AV
,
N
B
IN
IN
OSN
Z
C
we can replace V
C
with Equation 6 to obtain
V
GA
OSN
BZ
VGAV
=
+
.
(7)
N
B
IN
IN
1
+
Figure 6. Simplified OPA335 block diagram
V
OSN
2
+
A
IN
1
+
2
–
G
B
–
1
OPA335
+
V
C
A
Z
–
A
N1
V
OSN
1
+
A
IN
+
–
G
B
Oscillator
–
1=2
+
V
C
A
Z
–
A
N2
V
N
V
OSM
+
A
M
G
O
V
OUT
V
IN
V
M
–
Note: Labels for amplifiers such as A
IN
, G
B
, and A
M
are used to identify amplifiers in figures and to represent amplifier gain in equations.
23
Analog Applications Journal
2Q 2005
www.ti.com/aaj
Analog and Mixed-Signal Products
Plik z chomika:
TirNaNog
Inne pliki z tego folderu:
slyt331.pdf
(570 KB)
slyt333.pdf
(771 KB)
slyt335.pdf
(746 KB)
slyt332.pdf
(706 KB)
slyt334.pdf
(626 KB)
Inne foldery tego chomika:
ACE
AcornUser
AmigaComputing
AmigaFormat
AmigaShopper
Zgłoś jeśli
naruszono regulamin