AD8362
loop to settle. Because the scaling parameters of the two
squarers are accurately matched, it follows that Equation 4
is satisfied only when
At high frequencies, signal levels are commonly specified in
power terms. In these circumstances, the source and termina-
tion impedances are an essential part of the overall scaling. For
this condition, the output voltage can be expressed as
MEAN(VSIG2) = VATG
(5)
2
VOUT = SLOPE × (PIN − PZ)
(11)
In a formal solution, extract the square root of both sides to
provide an explicit value for the root-mean-square (rms) value.
However, it is apparent that by forcing this identity through
varying the VGA gain and extracting the mean value by the
filter provided by the capacitor(s), the system inherently
establishes the relationship
where PIN and the intercept PZ are expressed in dBm.
In practice, the response deviates slightly from the ideal straight
line suggested by Equation 11. This deviation is called the law
conformance error. In defining the performance of high accuracy
measurement devices, it is customary to provide plots of this
error. In general terms, it is computed by extracting the best
straight line to the measured data using linear regression over
a substantial region of the dynamic range and under clearly
specified conditions.
rms(VSIG) = VATG
(6)
Substituting the value of VSIG from Equation 3,
rms[GOVIN exp(−VSET/VGNS)] = VATG
(7)
As a measurement device, VIN is the unknown quantity and all
other parameters can be fixed by design. To solve Equation 7,
3.0
3.8
3.5
2.5
2.0
rms[GOVIN/VATG] = exp(VSET/VGNS
therefore,
VSET = VGNS log[rms(VIN)/VZ]
)
(8)
–40°C
3.2
2.9
2.6
2.3
2.0
1.7
1.4
1.1
0.8
0.5
0.2
1.5
1.0
0.5
(9)
0
The quantity VZ = VATG/GO is defined as the intercept voltage
because VSET must be 0 when rms (VIN) = VZ.
–0.5
–1.0
–1.5
–2.0
–2.5
–3.0
+25°C
+85°C
When connected as a measurement device, the output of the
buffer is tied directly to VSET, which closes the AGC loop.
Making the substitution VOUT = VSET and changing the
log base to 10, as needed in a decibel conversion,
–40°C
+25°C
+85°C
–60 –55 –50 –45 –40 –35 –30 –25 –20 –15 –10 –5
INPUT AMPLITUDE (dBm)
0
5
10 15
VOUT = VSLP log10[rms(VIN)/VZ]
(10)
Figure 45. Output Voltage and Law Conformance Error
@ TA = −40°C, +25°C, and +85°C
where VSLP is the slope voltage, that is, the change in output
voltage for each decade of change in the input amplitude.
Figure 45 shows the output of the circuit of Figure 47 over the
full input range. The agreement with the ideal function (law
conformance) is also shown. This was determined by linear
regression on the data points over the central portion of the
transfer function for the +25°C data.
Note that VSLP = VGNS log (10) = 2.303 VGNS
.
In the AD8362, VSLP is laser-trimmed to 1 V using a 100 MHz
test signal. Because a decade corresponds to 20 dB, this slope
can also be stated as 50 mV/dB. The Altering the Slope section
explains how the effective value of VSLP can be altered by the
user. The intercept, VZ, is also laser-trimmed to 224 μV (−60 dBm
relative to 50 ꢀ). In an ideal system, VOUT would cross zero
for an rms input of that value. In a single-supply realization of
the function, VOUT cannot run fully down to ground; here, VZ
is the extrapolated value.
The error at −40°C, +25°C, and +85°C was then calculated by
subtracting the ideal output voltage at each input signal level
from the actual output and dividing this quantity by the mean
slope of the regression equation to provide a measurement of
the error in decibels (scaled on the right-hand axis of Figure 45).
The error curves generated in this way reveal not only the devia-
tions from the ideal transfer function at a nominal temperature,
but also the additional errors caused by temperature changes.
Notice that there is a small temperature dependence in the
intercept (the vertical position of the error plots).
VOLTAGE VS. POWER CALIBRATION
The AD8362 can be used as an accurate rms voltmeter from
arbitrarily low frequencies to microwave frequencies. For low
frequency operation, the input is usually specified either in
volts rms or in dBV (decibels relative to 1 V rms).
Figure 45 further reveals a periodic ripple in the conformance
curves. This is due to the interpolation technique used to select
the signals from the attenuator, not only at discrete tap points,
but anywhere in between, thus providing continuous attenua-
tion values. The selected signal is then applied to the 3.5 GHz,
40 dB fixed gain amplifier in the remaining stages of the VGA
of the AD8362.
Rev. D | Page 17 of 32