AD603
indicated by the “slider” in Figure 1, thus providing continuous
attenuation from 0 dB to 42.14 dB. It will help, in understanding
the AD603, to think in terms of a mechanical means for moving
this slider from left to right; in fact, its “position” is controlled
by the voltage between Pins 1 and 2. The details of the gain-
control interface are discussed later.
THEORY OF OPERATION
The AD603 comprises a fixed-gain amplifier, preceded by a
broadband passive attenuator of 0 dB to 42.14 dB, having a
gain-control scaling factor of 40 dB per volt. The fixed gain is
laser-trimmed in two ranges, to either 31.07 dB (×35.8) or
50 dB (×358), or may be set to any range in between using one
external resistor between Pins 5 and 7. Somewhat higher gain
can be obtained by connecting the resistor from Pin 5 to com-
mon, but the increase in output offset voltage limits the
maximum gain to about 60 dB. For any given range, the band-
width is independent of the voltage-controlled gain. This system
provides an under- and overrange of 1.07 dB in all cases;
for example, the overall gain is –11.07 dB to 31.07 dB in the
maximum-bandwidth mode (Pin 5 and Pin 7 strapped).
The gain is at all times very exactly determined, and a linear-in-
dB relationship is automatically guaranteed by the exponential
nature of the attenuation in the ladder network (the X-AMP
principle). In practice, the gain deviates slightly from the ideal
law, by about ±0.2 dB peak (see, for example, Figure 16).
Noise Performance
An important advantage of the X-AMP is its superior noise per-
formance. The nominal resistance seen at inner tap points is
41.7 Ω (one third of 125 Ω), which exhibits a Johnson noise-
spectral density (NSD) of 0.83 nV/√Hz (that is, √4kTR) at 27°C,
which is a large fraction of the total input noise. The first stage
of the amplifier contributes a further 1 nV/√Hz, for a total input
noise of 1.3 nV/√Hz. It will be apparent that it is essential to use
a low resistance in the ladder network to achieve the very low
specified noise level. The signal’s source impedance forms a
voltage divider with the AD603’s 100 Ω input resistance. In
some applications, the resulting attenuation may be unaccept-
able, requiring the use of an external buffer or preamplifier to
match a high impedance source to the low impedance AD603.
This X-AMP structure has many advantages over former methods
of gain-control based on nonlinear elements. Most importantly,
the fixed-gain amplifier can use negative feedback to increase its
accuracy. Since large inputs are first attenuated, the amplifier
input is always small. For example, to deliver a ±1 V output in
the –1 dB/+41 dB mode (that is, using a fixed amplifier gain of
41.07 dB) its input is only 8.84 mV; thus the distortion can be
very low. Equally important, the small-signal gain and phase
response, and thus the pulse response, are essentially indepen-
dent of gain.
Figure 1 is a simplified schematic. The input attenuator is a
seven-section R-2R ladder network, using untrimmed resistors
of nominally R = 62.5 Ω, which results in a characteristic resis-
tance of 125 Ω ± 20%. A shunt resistor is included at the input
and laser trimmed to establish a more exact input resistance of
100 Ω ± 3%, which ensures accurate operation (gain and HP
corner frequency) when used in conjunction with external resistors
or capacitors.
The noise at maximum gain (that is, at the 0 dB tap) depends
on whether the input is short-circuited or open-circuited: when
shorted, the minimum NSD of slightly over 1 nV/√Hz is achieved;
when open, the resistance of 100 Ω looking into the first tap
generates 1.29 nV/√Hz, so the noise increases to a total of
1.63 nV/√Hz. (This last calculation would be important if the
AD603 were preceded by, for example, a 900 Ω resistor to allow
operation from inputs up to 10 V rms.) As the selected tap
moves away from the input, the dependence of the noise on
source impedance quickly diminishes.
The nominal maximum signal at input VINP is 1 V rms (±1.4 V
peak) when using the recommended ±5 V supplies, although
operation to ±2 V peak is permissible with some increase in HF
distortion and feedthrough. Pin 4 (SIGNAL COMMON) must
be connected directly to the input ground; significant impedance in
this connection will reduce the gain accuracy.
Apart from the small variations just discussed, the signal-to-
noise (S/N) ratio at the output is essentially independent of the
attenuator setting. For example, on the –11 dB/+31 dB range
the fixed gain of ×35.8 raises the output NSD to 46.5 nV/√Hz.
Thus, for the maximum undistorted output of 1 V rms and a
1 MHz bandwidth, the output S/N ratio would be 86.6 dB, that
is, 20 log (1 V/46.5 µV).
The signal applied at the input of the ladder network is attenu-
ated by 6.02 dB by each section; thus, the attenuation to each of
the taps is progressively 0 dB, 6.02 dB, 12.04 dB, 18.06 dB,
24.08 dB, 30.1 dB, 36.12 dB and 42.14 dB. A unique circuit
technique is employed to interpolate between these tap-points,
VPOS
VNEG
SCALING
REFERENCE
PRECISION PASSIVE
INPUT ATTENUATOR
FIXED GAIN
AMPLIFIER
GPOS
GNEG
V
OUT
V
G
6.44k⍀*
GAIN
CONTROL
INTERFACE
AD603
FDBK
694⍀*
20⍀*
0dB
–6.02dB –12.04dB–18.06dB –24.08dB –30.1dB –36.12dB –42.14dB
VINP
R
R
R
R
R
R
R
2R
2R
2R
2R
2R
2R
R
COMM
R = 2R LADDER NETWORK
*NORMAL VALUES
Figure 1. Simplified Block Diagram of the AD603
–4–
REV. C
ADI [ ADI ]
