Flicker Noise of Microwave Power Detectors

Serge Grop and Enrico Rubiola

FEMTO-ST Institute
UMR 6174, CNRS and Université de Franche Comté
e-mail: [email protected]
e-mail: [email protected]

1000
I. I NTRODUCTION
500 quadratic region
A microwave power detector is a diode used in the quadratic (power detector)
200

output voltage, mV
region. In this region the resistance is proportional to voltage,
100
and therefore the filtered output voltage is proportional to 100 kΩ
50
the mean square input voltage. Tunnel diodes [1], [2], [3]
20
100 Ω
and Schottky diodes are used in practice. Tunnel diodes are
preferable because of their high gain, up to 1400 V/W, and 10
1 kΩ
of because of their suitability to cryogenic environment. Yet, 5

they are easily damaged in case of to power overload. Schottky linear region
2 (envelope detector)
diodes work only at room temperature, with a gain of the 1
order of 300 V/W. If overloaded, the Schottky detector turns −30 −20 −10 0 10
input power, dBm
into a peak-voltage detector, which is still useful to measure
the power if the signal can be assumed to be sinusoidal. On Fig. 1. Two-diode Schottky power detector response
the other hand, the overloaded detector cannot be used as
a demodulator by exploiting the beat note between a pump rf in video out
carrier and a probe signal. Simple mathematics tells us that
1/f noise turns into a flat floor of the two-sample (Allan)
deviation. Hence, the 1/f noise of power detectors is related
to the power stability floor of the device, which is an important external
~60 10−200 50 Ω to
issue in a number of high-precision applications involving Ω pF 100 kΩ
the measurement or stabilization of a RF/microwave power,
the measurement of a beat note, and the Pound frequency- Fig. 2. Diode power detector principle
locking of an oscillator to a reference resonator. Nonetheless,
the literature on this subject is totally absent.
where α(t) and ϕ(t) are the fractional amplitude and phase
II. P OWER DETECTOR AND AM NOISE fluctuations respectaly. In low noise condition (α(t) ≪ 1), the
A microwave power detector uses the nonlinear response to power is
a diode to turn the input power P to a DC voltage vd . For V0
small input power, the transfer function is P ≈
(1 + 2α(t)) = P0 + δP (3)
2R
vd = kd P (1) The amplitude fluctuations are obtained from the measure-
ment of the power fluctuation δP
which defines the detector gain. The technical unit in the
datasheets is V /W , which corresponds to the physical dimen- 1 δP
α(t) = (4)
sion A−1 . For higher input power , the detector response turns 2 P0
smoothly from quadratic to linear like that of the old good In optics, the power fluctuations δP
P0 is called Relative
AM (envelope) demodulator. Figure 1 shows the response of Intensity Noise (RIN) [4], [5]. Taking the power spectral
a two-diode Schottky power detector. densities eq. 4 yield
Figure 2 shows the typical scheme of the diode detectors.
The input resistor matches the high input impedance of the 1 1
Sα (f ) = S δP = SδP (5)
diode network to the standard value R0 = 50Ω over the 4 P0 4P02
bandwidth and over the power range. The output capacitor In the case of amplitiude noise, generally the spectrum
filters the video signal, eliminating the carrier. contains only white noise h0 f 0 , flicker noise h−1 f −1 , and
Let us consider microwave sinusoidal random walk h−2 f −2 .

v(t) = V0 [1 + α(t)]cos[ω0 t + ϕ(t)] (2)

978-1-4244-3510-4/09/$25.00 ©2009 IEEE 40

 gain in the paths. This is achieved by adjusting the path A,
0
X after modulating the source amplitude and inspecting on the
Sα (f ) = hi f i (6) null with a lock-in amplifier. Similarly, the signal at the lower
i=−2 input of the FFT analyser is y = g(Pc − Pb ), independent
= h0 + h−1 f −1 + h−2 f −2 (7) of the AM noise of the source. The cross spectral density <
Syx >m convergest the noise of the detector C, after averaging
The spectrum density can be converted into Allan variance on a sufficient number m of samples. For this to be true, the
using two amplifiers at the output of the detector C must be fully
h0 4π 2 independent. Referring to Fig. 4 these amplifiers must have the
σ 2 (τ ) =
+ 2 ln(2)h−1 + h−2 τ (8) input stage based on JFET transistors because this technology
2τ 6
exhibits the lowest current noise, negligible for our purposes.
The power
p stability of the detector is limited by the flicker Otherwise, the current noise of the amplifier shows up as a
floor σ = 2 ln(2)h−1 . fully-correlated voltage fluctuation across the common load
Using detector law eq. 1, the ac component of the detected resistor. The load resistor is a part of the detector. Its noise
signal is vd = kd δP which is related to α(t) by cannot be rejected.
vd = 2kd P0 α(t) (9) v1 va
g
Turning voltages into spectra, the above becomes
C i1
rf in
Sv (f ) = 4kd2 P02 Sα (f ) (10)
Therefore, the spectrum of α(t) can be measured using R
v2 vb
g
1
Sα (f ) = Sv (f ) (11)
4kd2 P02 i2

Where Sv (f ) is the data obtained after measurement in

dBV2 /Hz. Fig. 4. AM noise measurement principle

III. E XPERIMENTAL METHOD The load resistor R is a parameter that can be chosen in
an interval from a few tens of Ohms to 100kΩ. High speed
Pa A va
adj. gain
diff. ampli applications require lower R while higher R yields higher kd .
x = g(Pc − Pa )
In the absence of information on the detector noise, we design
Ra
the system for the lowest flicker of the amplifier alone. This
low noise
source
Pc C vc
condition is met with a load resistor
JFET input dual channel
FFT analyzer
Rc en
AM
input R= (12)
monitor in
power B vb
meter Pb
y = g(Pc − Pb )
where en and in are the 1/f voltage-noise and current-
diff. ampli
Rb adj. gain noise of the amplifier from the populare Rothe-Dalke model
of amplifier [7], found in the datasheet. With the AD743
operational amplifier, the optimum resistor turns out to be of
osc. out input
lock−in out
Re
adjust the gain for the 3.2kΩ.
amplifier Im Re output to be zero

Fig. 3. Measurement scheme IV. R ESULTS

The preliminary results were obtained with some Schottky
Figure 3 shows the principle of the AM noise measurement. diodes Herotek DZR400KA. Figures 5 and 6 show the transfer
A simple detector can be measured directly only if a reference function and the gain of four power detectors.
source is available whose AM noise is lower than the detector 100 0

noise, and if the amplifier noise can be made negligible. These 334324
10 334323 -100

requirements are unrealistic. A more sophistical approach is to 334322

output voltage (−V)

334321
Gain (V/W)

1 -200
compare two detectors [6], a and b, by taking the differential
envelope
signal kd Pa −kd Pb . This method is still unsatisfactory because 0.1
detector
-300
334321
334322
the noise of the two detectors cannot be divided, so the result 0.01 power
detector
-400 334323
334324

relies upon the assumption that the two detectors have equal 0.001
-40 -35 -30 -25 -20 -15 -10 -5 0
-500
-30 -25 -20 -15 -10 -5 0
input power (dBm) Input power (dBm)
noise. Hence we got for the scheme of Fig. 3. From the circuit
topology, the signal at the upper input of the FFT analyser is Fig. 5. Transfer functions of 4 Fig. 6. Gain of 4 Schottky diode
x = g(Pc − (Pa ) where g is the system gain. For the AM Schottky diode power detectors power detectors
noise of the source to be rejected, we need A and C equal
41
 The AM noise was measured at −6, −10 (envelope detector an input power of −20dBm (σα = 2.1 × 10−6 ) and h−1 =
region) and −20dBm (power detector region) input powers −120dB at −10dBm (σα = 1.2 × 10−6). The measurement of
at 10GHz. Tab I summerizes the gain of a few detectors in the detector S/N 334321 is not fully trusted because envelope
different conditions. detector region is absent in Fig. 5 and 6. This could be the
signature of a loss not accounted for.
TABLE I
H EROTEK DZR400KA GAIN (V/W) V. C ONCLUSION
S/N Pin =-20dBm Pin =-10dBm Pin =-6dBm We presented the first results on the flicker noise of power
334321 -117 -133 -126
microwave detectors. Three channel measurement was used
334322 -342 -274 -206
in order to measure the noise of the center detector. With
334323 -333 -242 -184
this method, the noise contribution of the amplifier and the
334324 -356 -248 -184
source is reduced by measuring the cross-spectrum at the
output of two differential amplifiers. Two Shottky diodes
The preliminary measurements were done with the diodes detectors (Herotek DZR400KA) were characterized. The first
334321 and 334323. The measured spectra are to be validated (S/N 334321), showing a gain around 150V/W, present a
by checking on two parameters: 1) the residual of the master flicker noise h−1 = −115dB which corresponds of an Allan
oscillator AM noise must be lower than the observed cross- deviation σα = 2.1 × 10−6 at −6, −10 and −20dBm input
spectrum, and 2) the observed cross-spectrum must be higher power but these results are not fully trusted because envelope
than the statistical limit given by the single channel noise detector region is absent. The second (S/N 334323), showing
√ a gain around 300V/W, present a flicker noise h−1 = −115dB
divided by m. The master oscillator AM rejection K is
evaluated as (the same as previously) at −20dBm input power (power
detector) and a flicker noise h−1 = −120dB, corresponding
αmod of an Allan deviation σα = 1.2 × 10−6 at −10dBm (envelope
K= (13)
αC−A detector region). This measures will have an direct impact on
where αmod is a amplitude modulation implemented in the the power control/stabilization and Pound frequency/locking
synthesizer, and αC−A the readout of the lock-in amplifier control. With these values, we can access to the limit of the
converted into α at the output of the detectors. Of course, error signal. Tunnel diode characterization at room temperature
the same thing is done with the output C − B. In our and at some cryogenic temperature is in progress.
experiments, the rejection turns out to be between 80 to
100dB. The AM noise of the synthesizer (Wiltron 69137A R EFERENCES
at 10GHz) is measured through the cross-spectrum of the [1] C. A. Burrus, Backward diodes for low-level millimeter-wave detection,
channels A and B, by passing the differential amplifiers. The IEEE Trans. Microw. Theory Tech. 11 (1963), no. 9, 357362. 4, 13
flicker noise h−1 is −105dB which corresponds to a flicker [2] William F. Gabriel, Tunnel-diode low-level detection, IEEE Trans. Mi-
crow. Theory Tech. 15 (1967), no. 10, 538553. 4, 6
floor of σα = 6.6 × 10−6 of the Allan deviation (see Fig. 7). [3] R. N. Hall, Tunnel diodes, IRE Trans. Electron Dev. (?) (1960), no. 9,
19. 4
-100 [4] Irène Joindot, Measurement of relative intensity noise (RIN) in semicon-
ductor lasers, J. Phys. III France 2 (1992), no. 9, 15911603. 9
[5] Gregory E. Obarski and Jolene D. Splett, Transfer standard for the spectral
-105
density of relative intensity noise of optical fiber sources near 1550 nm,
J. Opt. Soc. Am. B - Opt. Phys. 18 (2000), no. 6, 750761. 9
-110 [6] Enrico Rubiola and Vincent Giordano, Advanced interferometric phase
Sα (f) (dB/Hz)

and amplitude noise measurements, Rev. Sci. Instrum. 73 (2002), no. 6,

1/f 24452457, Also on arxiv.org, document arXiv:physics/0503015v1. 1
-115 [7] H. Rothe and W. Dahlke, Theory of noisy fourpoles, Proc. IRE 44 (1956),
811818. 7
-120
white

-125

-130
1 10 100 1000 10000 100000
frequency (Hz)

Fig. 7. AM noise Sα (f ) of the Wiltron 69137A synthesizer at 10GHz

Figure 8 shows the results of the AM noise measuremnt of

the Schottky diode power detectors.
For the Schottky diode power detector 334321, a flicker
noise h−1 = −115dB is obtained which corresponds at a
flicker floor σα = 2.1 × 10−6 of the Allan deviation. With
the other detector 334321, we measured h−1 = −115dB for
42
 334324 334323 334321 AM rejection 334324 334323 334321 AM rejection 334321 AM rejection
334324 334323

-110 -110
sin -110 sin
gle Pin =-10dBm g Pin =-6dBm
le
single channels ch S/N 334321 ch S/N 334321
an ann
-120 1/f -120 ne els
ls -120
white cross−spectrum sta sta
tis t
Sα (dBm)

-130 -130 tic ist

Sα (dB)

al -130 ica

Sα (dB)
lim l li
statistical limit it 1/f mi
t
-140 -140 1/f
Pin = −20dBm cross−spectrum -140
cross−spectrum
S/N 334321
-150 -150 -150

-160 -160
1 10 100 1000 10000 100000 1 10 100 1000 10000 100000 -160
1 10 100 1000 10000 100000
frequency (Hz) frequency (Hz) frequency (Hz)

334323 334322 3343224 AM rejection 334323 334322 3343224 AM rejection

-70 -90

-80
$P_{in}=−20dBm$ -100 Pin = −10dBm
-90 S/N 334323 S/N 334323
-110
-100 sin
g le c
Sα (dB/Hz)

Sα (dB/Hz)

-110 -120 han

cro nel
single channels ss− s
-120 -130 sp ect
1/f rum
white cross−spectrum
-130
statistical limit -140 1/f
-140 stat
isti white
-150 cal
-150 lim
it
-160 -160
1 10 100 1000 10000 100000 1 10 100 1000 10000 100000
frequency (Hz) frequency (Hz)

Fig. 8. AM noise Sα (f ) of the Schottky diodes (up S/N 334321 and down S/N 334323

43