
AD600/AD602
Rev. E | Page 23 of 28 
2.0
–2.0
0.5
1.0
1.5
–1.5
–1.0
–0.5
G
–0.1
0.1
0
1μ
10μ
10
1
100m
10m
1m
100μ
INPUT SIGNAL (V rms)
0
Figure 49. Gain Error for Figure 41 Without the 2 dB Offset Modification 
2.0
–2.0
0.5
1.0
1.5
–1.5
–1.0
–0.5
G
–0.1
0.1
0
1μ
10μ
10
1
100m
10m
1m
100μ
INPUT SIGNAL (V rms)
0
Figure 50. Adding the 2 dB Offsets Improves the Linearization 
The maximum gain of this circuit is 120 dB. If no filtering was 
used, the noise spectral density of the AD600 (1.4 nV/√Hz) 
would amount to an input noise of 8.28 μV rms in the full 
bandwidth (35 MHz). At a gain of one million, the output noise 
would dominate. Consequently, some reduction of bandwidth is 
mandatory, and in the circuit of Figure 47, it is due mostly to a 
single-pole, low-pass filter R5/C3, which provides a 3 dB 
frequency of 458 kHz, which reduces the worst-case output 
noise (at V
AGC
) to about 100 mV rms at a gain of 100 dB. Of 
course, the bandwidth (and therefore the output noise) could be 
further reduced, for example, in audio applications, merely by 
increasing C3. The value chosen for this application is optimal 
in minimizing the error in the V
LOG
 output for small input signals. 
The AD600 is dc-coupled, but even miniscule offset voltages at 
the input would overload the output at high gains; thus, high-
pass filtering is also needed. To provide operation at low 
frequencies, two simple 0s at about 12 Hz are provided by 
R1/C1 and R4/C2; op amp sections U3A and U3B (
AD713
)  
are used to provide impedance buffering, because the input 
resistance of the AD600 is only 100 Ω. A further 0 at 12 Hz is 
provided by C4 and the 6.7 kΩ input resistance of the AD636 
rms converter. 
The rms value of V
LOG
 is generated at Pin 8 of the AD636; the 
averaging time for this process is determined by C5, and the 
value shown results in less than 1% rms error at 20 Hz. The 
slowly varying V rms is compared with a fixed reference of 
316 mV, derived from the positive supply by R10/R11. Any 
difference between these two voltages is integrated in C6, in 
conjunction with Op Amp U3C, the output of which is V
LOG
. A 
fraction of this voltage, determined by R12 and R13, is returned 
to the gain control inputs of all AD600 sections. An increase in 
V
LOG
 lowers the gain because this voltage is connected to the 
inverting polarity control inputs. 
In this case, the gains of all three VCA sections are being varied 
simultaneously, so the scaling is not 32 dB/V but 96 dB/V or 
10.42 mV/dB. The fraction of V
LOG
 required to set its scaling to 
50 mV/dB is therefore 10.42/50 or 0.208. The resulting full-
scale range of V
LOG
 is nominally ±2.5 V. This scaling allows the 
circuit to operate from ±5 V supplies.  
Optionally, the scaling can be altered to 100 mV/dB, which 
would be more easily interpreted when V
LOG
 is displayed on a 
DVM by increasing R12 to 25.5 kΩ. The full-scale output of 
±5 V then requires the use of supply voltages of at least ±7.5 V. 
A simple attenuator of 16.6 ± 1.25 dB is formed by R2/R3  
and the 100 Ω input resistance of the AD600. This allows the 
reference level of the decibel output to be precisely set to 0 for 
an input of 3.16 mV rms, and thus center the 100 dB range 
between 10 μV and 1 V. In many applications, R2/R3 can be 
replaced by a fixed resistor of 590 Ω. For example, in AGC 
applications, neither the slope nor the intercept of the 
logarithmic output is important. 
A few additional components (R14 to R16 and Q1) improve the 
accuracy of V
LOG
 at the top end of the signal range (that is, for 
small gains). The gain starts rolling off when the input to the 
first amplifier, U1A, reaches 0 dB. To compensate for this 
nonlinearity, Q1 turns on at V
LOG
 ~ 1.5 V and increases the 
feedback to the control inputs of the AD600s, thereby needing a 
smaller voltage at V
LOG
 to maintain the input to the AD636 to 
the setpoint of 316 mV rms. 
A 120 dB RMS/AGC SYSTEM WITH OPTIMAL SNR 
(SEQUENTIAL GAIN) 
In the last case, all gains were adjusted simultaneously, resulting 
in an output SNR that is always less than optimal. The use of 
sequential gain control results in a major improvement in SNR, 
with only a slight penalty in the accuracy of V
LOG
, and no 
penalty in the stabilization accuracy of V
AGC
. The idea is to 
increase the gain of the earlier stages first (as the signal level 
decreases) and maintain the highest SNR throughout the 
amplifier chain. This can be easily achieved with the AD600 
because its gain is accurate even when the control input is 
overdriven. That is, each gain control window of 1.25 V is  
used fully before moving to the next amplifier to the right.