Data Sheet

ADA4051-1/ADA4051-2

Rev. C | Page 17 of 22

THEORY OF OPERATION

The ADA4051-1/ADA4051-2 micropower chopper operational

amplifiers feature a novel, patent-pending technique that sup-

presses offset-related ripple in a chopper amplifier. Instead of

filtering the ripple in the ac domain, this technique nulls the

initial offset of the amplifier in the dc domain, thus preventing

ripple at the overall output.

Auto-zeroing and chopping are two techniques widely used in

high precision CMOS amplifiers to achieve low offset, low offset

drift, and no 1/f noise. Each of these techniques has pros and

cons. Auto-zeroing results in more in-band noise due to aliasing

introduced by sampling. On the other hand, chopping produces

offset-related ripple because it modulates the initial offset

associated with the amplifier up to its chopping frequency.

To accomplish the best noise vs. power trade-off, the chopping

technique is the better approach when designing a low offset

amplifier because there is no increased in-band noise. It is

preferable to suppress the offset-related ripple inside a chopper

amplifier because the offset-related ripple would otherwise need

to be eliminated by an extra off-chip postfilter.

Figure 59 shows the block diagram design of the ADA4051-1/

ADA4051-2 chopper amplifiers employing a local feedback loop

called autocorrection feedback (ACFB). The main signal path

contains an input chopping switch network (CHOP1), a first

transconductance amplifier (Gm1), an output chopping switch

network (CHOP2), a second transconductance amplifier (Gm2),

and a third transconductance amplifier (Gm3). CHOP1 and

CHOP2 operate at 40 kHz of chopping frequency to modulate

the initial offset and 1/f noise from Gm1 up to the chopping

frequency. A fourth transconductance amplifier (Gm4) in the

ACFB senses the modulated ripple at the output of CHOP2,

caused by the initial offset voltage of Gm1. Then, the ripple is

demodulated down to a dc domain through a third chopping

switch network (CHOP3), operating with the same chopping

clock as CHOP1 and CHOP2. Finally, a null transconductance

amplifier (Gm5) tries to null any dc component at the output of

Gm1 that would otherwise appear in the overall output as ripple.

A switched-capacitor notch filter (NF) functions to selectively

suppress the undesired offset-related ripple without disturbing

the desired input signal from the overall input. The desired input

dc signal appears as a dc signal at the output of CHOP2. Then,

the initial offset is modulated up to the chopping frequency by

CHOP3 and filtered out by the NF. Therefore, initial offset does

not create any feedback and does not disturb the desired input

signal. The NF is synchronized with the chopping clock to filter

out the modulated component. In the same manner, the offset

of Gm5 is filtered out by the combination of CHOP3 and the

NF, enabling accurate ripple sensing at the output of CHOP2.

In parallel with the high dc gain path, a feedforward transcon-

ductance amplifier (Gm6) is added to bypass the phase shift

introduced by the ACFB at the chopping frequency. Gm6 is

designed to have the same transconductance as Gm1 to avoid

pole-zero doublets. This design prevents any instability introduced

by the ACFB in the overall feedback loop.

Gm6 (= Gm1)

C2

C1

Gm3

C3

NF

Gm1

Gm2

CHOP1

CHOP2

Gm4

Gm5

CHOP3

OUT

+IN

–IN

Figure 59. ADA4051-1/ADA4051-2 Chopper Amplifiers Block Diagram

The voltage noise density, which is equal to the thermal noise

floor dominated by the Gm1, is essentially flat from dc to the

chopping frequency because CHOP1 and CHOP2 eliminate the

1/f noise generated in Gm1 and the ACFB does not contribute

any additional noise. Although the ACFB suppresses the ripple

related to the chopping, there is a remaining voltage ripple. To

further suppress the remaining ripple down to a desired level, it

is recommended to have a postfilter at the output of the amplifier.

The remaining voltage ripple originates from two sources. The

first type of ripple is due to the residual ripple associated with

the initial offset of the Gm1. It is proportional to the magnitude

of the initial offset and creates a spectrum at the chopping

gain buffer, this ripple has a typical value of 4.9 μV rms and a

maximum of 34.7 μV rms. The second type of ripple is due to

the intermodulation between the high frequency input signal

and the chopping frequency. This ripple depends on the input

the difference between the chopping frequency and the input

summation of the chopping frequency and the input frequency

frequencies is shown in Figure 60.

0

100

200

300

400

500

0123456789

10

INPUT FREQUENCY (kHz)

Figure 60. ADA4051-1/ADA4051-2 Modulated Output Ripple vs. Input Frequency