WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: sales@wj.

ail: [email protected] Web site: www.wj.com

Tech-note
Author: Bert C. Henderson
Mixers in Microwave Systems (Part 1)
Mixers continue to play a critical role in RF
and microwave systems that employ frequen-
cy conversion. Although much has been writ-
ten concerning the theory and operation of
mixers, the purpose of this article is to present
some highlights of these details as they relate
to passive mixer design, theory, realization,
and usage.
MIXER THEORY
To achieve frequency conversion, a periodic
signal having frequency, f
R
, is modulated by a
periodic conductance (or resistance) wave-
form having frequency, f
L
. The current result-
ing from the RF voltage being modulated by
the fundamental component of the conduc-
tance waveform contains the sum and differ-
ence IF products: f
R
+ f
L
and f
R
- f
L
or f
L
- f
R
.
The other undesired currents generated are
intermodulation products having frequencies
of n f
L
f
R
, where n is an integer. Inter-mod-
ulation products have also been referred to as
idlers. In the case of an active mixer, one hav-
ing conversion gain, the conductance wave-
form is that of one or more transistors. In the
case of a passive mixer, which has conversion
loss, the conductance waveform is generally
that of one or more Schottky-barrier diodes.
Increasingly, however, MESFET devices have
been used instead of diodes to achieve wider
dynamic range in passive mixers.
Analysis of the frequency conversion proper-
ties of mixers is non-trivial. The simplest
model, which perhaps gives the best intuitive
understanding of the mixing process is the
linear phase-reversal mixer [1].
This model assumes the diode is nonreactive
and acts as a linear rectifier, having a square-
wave resistance waveform with zero forward
and infinite backward resistances. The mixer
is considered linear because the values of the
circuit elements, including diode conduc-
tance, are independent of RF and LO levels.
It has been shown that this model approxi-
mates the Schottky diode mixer closely
enough to formulate theoretical limits for
conversion-loss and intermodulation suppres-
sion [2, 3]. Figure 1 shows a conventional
double-balanced diode mixer. During positive
LO cycles, diodes D1 and D2 are turned on
while D3 and D4 are off. The opposite is the
case during negative LO cycles. This causes
the RF (signal) voltage as seen by the IF port
to change phase by 180 degrees every time
the LO signal changes polarity. This can be
represented mathematically by multiplying
the sinusoidal signal voltage with the Fourier
series for the square-wave switching function:
V
out
= V
RF
sin (!
R
t) [4/" # 1/n sin (n!
L
t)]
(1a)
= V
RF
4/" {(1/2) [sin(!
L
- !
R
)t
- sin(!
R
+ !
L
)t] (1b)
+ (1/6) [sin (3!
L
- !
R
)t
- sin (3!
L
+ !
R
)t] + ...}
Conversion loss is equal to the reciprocal of
conversion gain, and is defined as:
L = (2)
and RF-to-IF conversion loss is given by:
L = 20 log (3)
= -20 log 2/" = 3.92 dB
The 2/" term is the ratio of the signal voltage
to IF voltage. Equations 1 through 3 assume
that the RF and IF ports are conjugately
matched, all intermodulation (IM) products
are resistively terminated, and no parasitic
resistive or reactive losses exist. The above
analysis has been generalized to show that
when matched loads are presented to each IM
product, and the RF, IF, and image signals are
conjugately matched, the theoretical mini-
mum conversion loss is 3.92 dB. Also, when
all IM products and the sum (f
L
+ f
R
) product
are reactively terminated, the IF is conjugately
matched, and the RF and image signals are
identically terminated, then the theoretical
minimum conversion loss is 3 dB, with the
lost energy equally divided between conver-
sion to the image, and reflection-loss at the
signal frequency [3]. The image, in this con-
text, is a mixer-generated product having a
frequency of 2f
L
- f
R
. As discussed later in this
paper, in the context of image-rejection, the
image refers to noise or signal power having
frequency, 2f
L
- f
R
, that enters the mixer along
with the RF signal. Idlers are intermodulation
products that are associated with each mixer-
generated LO harmonic. Idlers of order n
comprise the two sidebands that are adjacent
to each LO harmonic, and have frequencies
of f
n
= nf
L
f
IF
, where n are integers greater
than one. Figure 2 shows that the spectrum of
signals present in a mixer includes the LO, IF,
image, LO harmonics, and idlers [4].
SUMMARY OF MIXER ANALYSIS
METHODS
The classical analysis of frequency conversion
is given by Torrey and Whitmer [5] for a sin-
gle exponential diode with small-signal RF
and large signal LO voltages applied. The
analysis considers the RF, IF, and image sig-
nals to be at low levels compared with the
LO. This allows these three signals to be con-
sidered as variations of the LO voltage and
current harmonics. The result is that their
voltage and current waveforms are linearly
Figure 1. Schematic diagram of a double-balanced diode
mixer.
LO
D1 D4
D3 D2
RF
IF
n=1,3,5...
Available RF Input Power
Available IF Output Power
V
RF
V
IF
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
related through an admittance matrix repre-
senting the mixer, with conversion loss being
given as a function of diode conductance.
The mixer can thus be regarded as a linear
network with separate terminals at the RF,
image, and IF frequencies. Theoretical mini-
mum conversion loss is shown to be about
3.9 dB for the case where signal and image
frequencies are terminated in the same resis-
tance, and about 2.5 dB for the case where
the image is short or open circuited. These
theoretical values vary depending on the cur-
rent-voltage characteristics of the particular
diode used. Actual conversion loss values are
shown to be higher due to junction capaci-
tance and spreading resistance.
More recently, Saleh [4], extended the forego-
ing analysis to include what he termed the Z,
Y, G and H mixers. In the Y and Z mixers, all
the idlers are short circuited and open circuit-
ed, respectively. In the G-mixer, the odd-order
idlers (including the image) are short circuited
and the even-order ones (including the sum
product) are open circuited. The reverse is
true for the H mixer. It is shown that the
optimum conductance waveform for the Y
signal to be terminated independently. Maas
has given a detailed description of this
process [6].
Non-linear microwave CAD programs are
available from various companies. Three of
these programs were evaluated. One did not
successfully converge for a four-diode mixer;
however, convergence for a single-MESFET
mixer took only one minute, with conversion
loss being within 1 dB of measured values.
The other two programs were evaluated
using the four-diode double-balanced mixer
examples supplied with the software.
Convergence took about 11 minutes, which
resulted in reasonable conversion loss of values.
MIXER PARAMETERS AND
OPTIMIZATION
The major goals of mixer design are to mini-
mize conversion loss, noise figure, and inter-
modulation distortion. Other important
parameters to optimize include VSWR and
compression.
CONVERSION LOSS
Conversion loss has three major components:
RF and IF mismatch loss, loss in the diode
spreading resistance, and loss in the diode
junction due to junction resistance and gen-
eration of IM products. A theoretical exam-
ple has been given [7] showing that mis-
match loss is typically 1 dB or less, but can
range from infinite to 0 dB; loss in the
spreading resistance is about 1 dB, and loss in
the junction is about 4 dB, for a total of 5 to
8 dB in a well-designed mixer.
In double-balanced mixers, the RF input and
image signals share the RF port, while the IF
and sum products share the IF port. It has
been shown that conversion loss can vary up
to 2 dB by open or short circuiting the image
[8]. This method is used to reduce (enhance)
mixer conversion loss. Maas has shown that
presenting a short circuit or capacitive termi-
nation, to the image provides the best trade-
off among conversion loss, noise figure and
third-order intermodulation. An open circuit
and Z mixers is a series of pulses, with duty
cycle related to the ratio of on to off resis-
tances of the diode. The optimum conduc-
tance waveform for the G and H mixers is a
square wave that is independent of the on-
to-off resistance ratio. Saleh found the theo-
retical limit for conversion loss to be 0 dB for
all idlers reactively terminated and the image
short or open circuited. These four types of
mixers are theoretical and are not perfectly
realizable in practical circuits.
During the last two decades, computers have
increasingly been used to analyze mixers. This
approach has obviated the need for many pre-
viously required limiting assumptions such as
a sinusoidal LO voltage at the diode, constant
(linear) diode-junction capacitance, and ter-
mination of idlers and LO harmonics in open
or short circuits. The general method is to
determine the diode conductance waveform
resulting from the applied LO, expand this
waveform into a Fourier series and relate the
resulting harmonics to mixing products
through a conversion matrix. The mixer is
represented as a linear network with a sepa-
rate port for each frequency, allowing each
IF LO
Image Signal
2nd-Order
Idlers
and Second
LO Harmonic
3rd-Order
Idlers
and Third
LO Harmonic
4th-Order
Idlers
and Fourth
LO Harmonic
Frequency
(
f
R
-
f
L
)
(
2
f
L
-
f
R
)
(
f
L
)
(
f
R
)
(
3
f
L
-
f
R
)
(
2
f
L
)

(
f
L
+
f
R
)
(
4
f
L
-
f
R
)
(
3
f
L
)
(
2
f
L
+
f
R
)
(
5
f
L
-
f
R
)
(
4
f
L
)
(
3
f
L
+
f
R
)
Figure 2. Frequency spectrum of mixing products.
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
mized by using a mixer with a low-pass IF
port that reflects the sum product energy
back into the diodes. Also, filter networks
with constant impedance as a function of fre-
quency can be used to control ripple.
The non-cyclical results in Figures 4A and 4B
have been reported previously [12] and are
attributed to reflections of the second har-
monic of the LO, in addition to the image.
To test this, the RF and LO frequencies were
changed, and the BPF at the RF port was
replaced with one that passes the image, but
rejects the 2f
L
product. This resulted in con-
version-loss variation as a function of L2
being reduced from 5 dB to 2 dB, and corre-
sponding to variations in the level of the 2f
L
product exiting the RF port. These results
show that the 2f
L
product termination can be
as significant as the image when considering
POWER
METER

LPF L2
WJ-M50
L1
LO
BPF
SPECTRUM
ANALYZER
RF R
L
I
Figure 3. Test setup used to measure conversion loss as a formation of Image and sum product terminations.
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Line Length L1 (in)
20
15
10
5
0
R
e
t
u
r
n

L
o
s
s

-

d
B
A
B
B
9
8
7
6
5
4
A
c
o
n
v
e
r
s
i
o
n

L
o
s
s

-

d
B
Figure 4A. Conversion loss and R-Port return loss for the test setup shown in Figure 3 as line length, L1, is varied, with
L2 set for best (A) and worst-case (B) conversion loss.
or inductive image termination can result in
significantly degraded noise figure and third-
order intermodulation performance [9].
Filtering and phase cancellation [10] have
been used to achieve image enhancement,
with the filtering method predominant.
Conversion loss can vary by up to 5 dB by
simultaneously reactively terminating the 2f
L
,
image and sum products [11]. To measure the
effect of varying these termination impedances,
a broadband mixer which covers 2 to 26 GHz
at the RF and LO ports, and 1 to 15 GHz at
the IF port, was tested as shown in Figure 3.
Figure 4A shows experimental conversion-loss
variation as a function of the line length (L1),
which is between the bandpass filter and the
RF port of the mixer. The BPF passes the RF
but rejects the image and 2F
L
frequencies.
Figure 4B shows conversion-loss variation as a
function of the line length (L2), which is
between the low-pass filter and the IF-port of
the mixer. The LPF passes the IF, but rejects
the sum frequency. For this experiment, f
L
=
4.5 GHz, f
R
= 3.0 GHz and f
IF
= 1.5 GHz. It
was found that for minimum and maximum
conversion loss, L1 and L2 are independent
of each other. Minimum and maximum con-
version loss values for this mixer were found
to be about 4.3 dB and 9.5 dB, respectively,
excluding filter and variable-line losses.
Results like these show that mixers having
conversion loss of 4 to 5 dB or less, must
employ enhancement techniques. They also
indicate that for swept frequencies, serious
conversion-loss ripple can result when filters
are placed adjacent to broadband mixer RF or
IF ports. Placing attenuators adjacent to the
mixer ports will reduce conversion-loss ripple.
Further experimental data shows that ripple
can be reduced from 5 dB to approximately
2.5 dB peak-to-peak, at the expense of
increasing conversion loss, by placing a 3-dB
attenuator at the RF or IF ports of the mixer.
Conversion-loss ripple caused by a varying
sum-frequency termination, can also be mini-
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
conversion-loss enhancement.
NOISE FIGURE
The single sideband (SSB) noise figure is
defined as:
NF = 10 log [P
n out
/( P
n in
x G)]
= 10 log F (4)
where,
P
n out
= Available noise output power at
IF frequency
P
n in
= Available noise input power at
RF frequency
G = Available power gain (algebraic
ratio)
F = Noise factor
The SSB noise factor is often described in
terms of equivalent input noise temperature,
T
mSSB
[7]:
F = 1+ (5)
where, T
o
= 290K
Noise figure for passive devices is equal to the
reciprocal of available power gain, as long as
both the noise source and the passive device
are at the same temperature. This implies that
mixer noise figure equals conversion loss; but,
Kerr and others have shown that this is not
strictly true: theoretically, mixer noise figure is
equivalent to that of an attenuator having
effective noise temperature T
M
equal to nT/2,
where n is the diode ideality factor and T is
the diode physical temperature. This results in
noise factor for a SSB mixer, one having infi-
nite image conversion-loss, being given as [6]:
T
mSSB
= (nT/2) (L-1) (6)
where L is algebraic RF-to-IF conversion-loss.
This is true for an ideal mixer in which all
idler frequencies are reactively terminated.
However, actual mixer noise factor values
tend to be higher, in part due to partial corre-
lation of down-converted shot-noise power,
which is generated by the time varying diode
series resistance.
Figure 4B. Conversion loss and R-Port return loss for the test setup shown in Figure 3 as line length, L2, Is varied, with
L1 set for best (A) and worst-case (B) conversion loss.
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Line Length L1 (in)
20
15
10
5
0
R
e
t
u
r
n

L
o
s
s

-

d
B
A
B
B
8
7
6
5
4
A
c
o
n
v
e
r
s
i
o
n

L
o
s
s

-

d
B
The predominant sources of noise in
Schottky diodes are thermal (Johnson) noise
in the series resistance, shot noise generated
by current flow across the barrier [7] and
flicker (1/f ) noise. Thermal noise is generated
by random current fluctuations in any resistor
with no external voltage present. Shot noise
results from a stream of electrons moving
through the diode barrier at random veloci-
ties, while producing an average dc current.
Thermal noise is a limiting form of shot noise
with zero bias applied.
Flicker noise is present in many devices,
including carbon resistors and silicon diodes
when current is flowing in them. For low fre-
quencies (below approximately 1 MHz), flick-
er-noise power is approximately proportional
to 1/f, where f is the operating frequency
[13]. Flicker noise in a Schottky diode is
related to surface-state density [14].
In addition to the above mechanisms of noise
generation, noise may become present at the
mixer output due to reciprocal mixing, cross
T
mSSB
T
o
modulation and imperfect LO-AM rejection.
Reciprocal mixing causes noise present on the
LO signal to be transferred to the IF output
when a second RF input at a high level
becomes present at the mixer RF input [15].
Rejection of AM noise on the LO is achieved
in balanced mixers in the same manner as L-
to-I isolation. Phase noise on the LO, howev-
er, is directly transferred to the IF signal. The
magnitude of the peak phase deviation is
multiplied
in harmonic mixers by the LO harmonic
number.
INTERMODULATION
Intermodulation (IM) distortion causes output
products to be generated at frequencies of:
f = nf
L
m
1
f
R1
m
2
f
R2
... (7)
where n, m1, m2,... are integers. The value, n,
is called the order of modulation, while the
sum (|m1| + |m2| +...) is referred to as the
order of intermodulation. As shown by equa-
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
Figure 5A. Downconverter spurious products chart.
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
2.0 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
f
R
/ f
L
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
f
I
F
/

f
L
1.0
L
-
R
L
-
2
R
L
-
3
R
L
-
4
R
5
R
-
L
2
R
4
R
-
L
2
L
-
2
R
3
L
-
4
R
2
L
-
3
R
3
R
-
L
4
L
-
5
R
4
L
-
4
R
3
L
-
3
R
R
3
R
2
R L
-
5
R
R 4
R
-
2
L
2
R
-
L
3
R
-
2
L
4
R
-
5
L
5
R
-
4
L
6
R
-
5
L
4
R
-
4
L
3
R
-
3
L
2
R
-
2
L
3
L
-
2
R
2
L
-
R
5
L
-
3
R
3
R
-
4
L
4
L
-
2
R
R
-
L
2
R
-
3
L
3
R
-
5
L
4
L
-
3
R
DOWNCONVERTER SPURIOUS PRODUCTS
tion 1, the linear square-wave phase-reversal
mixer generates IM products with frequencies
of f = nf
L
f
R
. IM products with intermodu-
lation orders greater than one are generated
by incremental diode nonlinearity and over-
loading, and can be considered nonlinear IM
products [16]. Overloading in diode mixers
occurs as the RF signal level approaches the
LO level, causing switching time to become a
function of RF as well as LO voltage.
It is important to identify the IM products
present in the IF output passband. This is eas-
ily done in a graphical manner for single-tone
products using charts such as those shown in
Figures 5A and 5B, which are for down- and
up-conversion, respectively. An example of
their usage is given in Appendix A.
Computer-generated IM search programs are
also very helpful in identifying the frequencies
of IM products [6,17].
IM suppression for single and two-tone prod-
ucts are generally of most interest. The order
of intermodulation is important because it
describes the behavior of the relative suppres-
sion between the IM and IF products as the
RF input power is varied. For example, the
two-tone, third-order IM product at f = f
L
+
2f
R1
- f
R2
, for P
RF1
= P
RF2
<<P
LO
, varies 3 dB
for every 1 dB of variation in the IF product
as P
RF1
and P
RF2
are varied. This behavior
generally applies to all orders of intermodula-
tion for any number of input tones. It gives
rise to the concept of input intercept point,
which equals the extrapolated input power to
the mixer (at each tone) that would cause the
output power levels of the IM and IF prod-
ucts to become equal. The benefit of using
the intercept method is that instead of having
to state both the input power level and rela-
tive level of suppression, only the intercept
point needs to be stated because suppression
is assumed to be 0 dB. Input intercept is
given by:
3IIP (dBm) = + P
RF in
where:
P
RF in
= Input RF Power for each tone; in dBm.
(8)
For example, the two-tone, third-order input
intercept point for a mixer with P
RF1
= P
RF2
=
-10 dBm, and relative suppression of 60 dBc, is:
3IIP = [60/(3-1)] - 10 = + 20 dBm (9)
Output intercept point equals input intercept
point plus device available power gain. It can
be shown that the theoretical third-order
input intercept point caused by overloading
in the linear phase-reversal mixer is equal to
P
LO
(dBm) + 9.0 dB [2]. In practice, the
third-order input intercept point for diode
mixers ranges from about 0 to 5 dB above the
LO power. It is higher for passive MESFET
mixers because the FET conductance wave-
form is more linear, and overloading is mini-
Suppression
(IM order-1)
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
mized by separating the RF and LO voltages
so that high RF levels are less able to phase
modulate the conductance waveform
[18,19,20]. Since overloading is caused by the
interference of the RF signal with the LO, its
Figure 5B. Upconverter spurious products chart.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
L
o
w
e
r

S
i
d
e
b
a
n
d

O
u
t
p
u
t


(
f
o
u
t
/
f
L
)
U
p
p
e
r

S
i
d
e
b
a
n
d

O
u
t
p
u
t


(
f
o
u
t
/
f
L
)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
IF Input (f
In
/f
L
)
2
L
-
3
I
2
L
-
2
I
L
+
3
I
2
L
-
1
2
L
-
4
I
4
I
3
L
-
3
I
L
+
2
I
5
I
-
L
3
I
4
I
-
L
4
L
-
4
I
3
L
-
2
I
4
L
-
3
I
3
I
-
L
2
I
L
+
I
5
L
-
4
I
4
I
-
2
L
4
I
3
I
2
I
L
-
5
I
L
-
4
I
5
I
-
L
4
I
-
L
L
-
3
I
L
-
2
I
2
L
-
4
I
2
L
-
3
I
3
I
-
L
I
4
I
-
2
L
2
I
-
L
3
I
-
2
L
4
L
-
4
I
3
L
-
3
I
2
L
-
2
I
L
-
I
UPCONVERTER SPURIOUS PRODUCTS
effects can be reduced by using a square-wave
LO, assuming the RF voltage level remains
below that of the LO. IM suppression caused
by overloading in a double-balanced mixer
has been given for various products as a func-
tion of diode and circuit imbalance [21].
Non-linear intermodulation can be reduced
by placing a resistor in series with each mixer
diode, thus, reducing its overall non-linearity
[22]. Also, placing two diodes in series or in
parallel reduces intermodulation. Various
classes of mixers with these configurations
have been described by Cheadle, and are
given in Figure 6.
CROSS MODULATION
Cross modulation is the process whereby
modulation or noise that is present on an
adjacent strong RF input signal is made to
appear on the IF output signal. This is similar
to reciprocal mixing, in which the noise origi-
nates from the LO signal. A method of com-
puting cross modulation levels has been given
by Gretsch [23].
Part 2 of Mixers in Microwave Systems will
discuss such topics as impedance matching,
diode-mixer design, mixer realization, and use
of mixers.
APPENDIX A: IM CHART
EXAMPLES
Use of Figures 5A and SB is straightforward.
These charts comprise the family of lines:
f
OUT
= n + m f
IN
Where f
OUT
and f
IN
are the output and
input frequencies, respectively, normalized by
the LO frequency. In Figure 5A the L-R and
R-L lines represent the transfer functions for
input-to-output frequency for the IF product
when f
R
< f
L
and f
L
< f
R
, respectively. The
goal is to determine which IM products will
appear within the IF passband for given val-
ues of f
IF
, f
R
and f
L
. For example, when f
R
= 6
to 8 GHz, f
L
= 10 GHz and f
IF
= 2 to 4
GHz, a square is drawn on the L-R line with
corners corresponding to the points f
IN
= 0.6
and 0.8. The 2R-L and 3R-2L lines cut
through this box, so that when f
IN
= 0.6, we
see that f
2R-L OUT
= 0.2, and when f
IN
0.8,
f
3R-2L
= 0.4, corresponding to output fre-
quencies of 2 and 4 GHz, respectively.
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
As f
R
increases to 8 GHz, the L-R IF product
decreases in frequency while the 2R-L and
3R-2L products increase in frequency, travers-
ing the 2 to 4 GHz IF passband at two and
three times, respectively, the rate of the IF fre-
quency shift. The up-conversion chart in
Figure 5B is used in the same manner.
ACKNOWLEDGEMENT
The author wishes to thank S.E. Avery, Dr.
R.K. Froehlich, and Dr. R.G. Ranson for
their thoughtful review of this paper, M.A.
OMahoney, and T.G. Skala for their non-lin-
ear CAD evaluations, R.W. Bruce for the
usage of his extensive bibliography, and R.Y.S.
Parsons for preparing the draft.
REFERENCES
1. Tucker, D.G. The Input Impedance of
Rectifier Modulators, lEE Proc., Vol.
107B, No. 1, pp. 273-284, January 1960.
2. Gardiner, J.G., et al. Distortion
Performance of SingleBalanced Diode
Modulators, lEE Proc., Vol. 117,
No. 8, August 1970.
3. Kelly, A.J. Fundamental Limits on
Conversion Loss of Double Sideband
Resistive Mixers, IEEE Trans.
Microwave Theory Tech., Vol. MTT-25,
No. 11, November 1977, pp. 867-869.
4. Saleh, A.A.M. Theory of Resistive
Mixers, Cambridge, Massachusetts: MIT
Press, 1971.
5. Torrey, H.C. and CA. Whitmer. Crystal
Rectifiers, New York: McGraw-Hill, 1948
6. Maas, S.A. Microwave Mixers, Artech
House, Dedham MA, 1986.
7. Held, D.N. and AR. Kerr. Conversion
Loss and Noise of Microwave and
Millimeter-Wave Mixers: Part 1-Theory
and Part 2-Experiment, IEEE Trans.
Microwave Theory Tech., Vol. MTT-26,
No.2, February 1978.
8. Burkley, C.J. and R.S. OBrien.
Optimization of an 11 GHz Mixer
Circuit Using Image Recovery, Int. J.
Electron, Vol. 38, pp. 777-787,
June 1975.
9. Mass, S. Two-Tone Intermodulation in
Diode Mixers, IEEE Trans. Microwave
Theory and Tech., Vol. MTT-35, No. 3,
March 1987, pp. 307-314.
10. Oxley, T.H. Phasing Type Image
Recovery Mixers, IEEE MTT-S mt.
Microwave Symposium Digest, pp. 270-
273,1980.
11. Dickens, L.E. and D.W. Main. An
Integrated-Circuit Balanced Mixer, Image
and Sum Enhanced, IEEE Trans.
Microwave Theory Tech., Vol. MTT-23,
No. 3, March 1975.
12. Pound, R.V. Microwave Mixers. Usington,
Massachusetts: Boston Technical
Publishers, Inc., pp. 81-87, 1964.
13. Shurmer, H.V. Microwave
Semiconductor Devices. New York:
Wiley Interscience, 1971.
14. Sze, SM. Physics of Semiconductor
Devices. New York: Wiley Inter-science,
1969, p. 459.
15. Sabin, WE. and E.O. Schoenike. Single-
Sideband Systems and Circuits,
New York: McGraw-Hill, 1987.
16. Steinbrecher, D.H. Mixer
Fundamentals, Notes presented at 1989
RF Technology Expo, Santa Clara, CA.
17. Barn, R. A Mixer Spurious Plotting
Program, RF Design, May 1989,
pp. 32-43.
18. Weiner, S., D. Neuf and S. Spohrer.
2 to 8 GHz Double Balanced MESFET
Mixer with +30 dBm Input 3rd Order
Intercept, IEEE MTT-S mt. Microwave
Symposium Digest, pp. 1097-1100, 1988.
19. Maas, SA. A GaAs MESFET Mixer
With Very Low Intermodulation, IEEE
Trans. Microwave Theory Tech., Vol.
Mfl-35, No. 4, April 1987.
20. Oxner, E. A Commutation Double-
Balanced MOSFET Mixer of High
Dynamic Range, Proceedings 1986 RF
Expo East, pp. 73-87.
21. Henderson, B.C. Reliably Predict Mixer
IM Suppression, Microwaves and RF,
Vol. 22, No. 12, Nov. 1983.
LO POWER FOR DB MIXERS
(dBm) CIRCUIT MIXER CLASS
Class 1
Class 2, type 1
Class 2, Type 2
Class 3, Type 1
Class 3, Type 2
Class 3, Type 3
+7 to +13
+13 to +24
+13 to -24
-20 to +30
+20 to +30
+20 to +30
Figure 6. The various classes of mixer-diode configurations.
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
Copyright 1990 Watkins-Johnson Company
Vol. 17 No. 1 January/February 1990
Revised and reprinted 2001 WJ Communications, Inc.
22. Cheadle, DL. Consider a Single Diode
to Study Mixer Intermod, Microwaves,
Dec. 1977.
23. Gretsch, W.R. The Spectrum of
Intermodulation Generated in a
Semiconductor Diode Junction, Proc.
IEEE, Vol. 54, No. 11, November 1966.
24. Faber, MT. and W.K. Gwarek.
Nonlinear-Linear Analysis of Microwave
Mixer With Any Number of Diodes,
IEEE Trans. Microwave Theory Tech.,
Vol. MTI-28, No. 11, Nov. 1980.
25. Mouw, RB. A Broad-Band Hybrid
Junction and Application to the Star
Modulator, IEEE Trans. Microwave
Theory Tech., Vol. MTT-16,
pp. 911-918, Nov. 1968.
26. Henderson, B. Orthogonal Mixers:
Punching Up Earth/Space Payload
Performance, MSN~ January 1982,
Vol. 12, No. 1.
27. Will, P. Termination Insensitive Mixers,
Professional Program Session Record 24,
WESCON, San Francisco, 1981.
28. Norton, D. Three Decade Bandwidth
Hybrid Circuits, Microwave Journal,
Vol. 31, No. 11, Nov. 1988, pp. 117-126.
29. Henderson, B. and J. Cook. Image
Reject and Single Sideband Mixers,
MSN, Vol. 17, No. 9, August 1987.
30. Aikawa, M. and H. Ogawa. Double
Sided MICs and Their Applications,
IEEE Trans. Microwave Theory Tech.,
Vol. 37, No. 2, February 1989.
31. Hirota, T., Y. Tarusawa and H. Ogawa.
Uniplanar MMIC Hybrids A Proposed
New MMIC Structure, IEEE Trans.
Microwave Theory Tech., Vol. 37, No. 2,
February 1989.
32. Izaclian, J., et al. A Uni-Planar Double-
Balanced Mixer Using A New Miniature
Beam Lead Crossover Quad, IEEE mt.
Microwave Symposium Digest,
pp. 691-694, 1988.
33. Pavio, AM., et al. Double Balanced
Mixers Using Active and Passive
Techniques, IEEE Trans. Microwave
Theory Tech., Vol. 36, No. 12, December
1988.
34. Titus, W., et al. Distributed Monolothic
Image Rejection Mixer, IEEE Int. GaAs
IC Symposium, pp. 191-194, 1986.
35. Ali, F., S. Moghe and R. Ramachandran.
A Highly Integrated X-Ku Band
Upconverter, IEEE GaAs IC
Symposium Digest, pp. 157-160, 1988.
36. Private communication with AF. Podell.
37. Fotowat, A. and E. Murthi. Gilbert-
Type Mixers vs. Diode Mixers,
Proceedings RF Technology Expo 1989,
pp. 409-413.
38. Wilson, SE. Evaluate The Distortion of
Modular Cascades, Microwaves, Vol. 20,
March 1981.
39. Sorger, G.U. The 1 dB Gain
Compression Point for Cascaded Two-
Port Networks, Microwave Journal,
July 1988.
40. Avery, SE. Dual Mixers, Watkins-
Johnson Company Tech-notes, Vol. 13,
No. 4, July/August 1986.
41. Schindler, SA. MIL-Specification
Mixers, Watkins Johnson Company
Tech-notes, Vol. 15, No. 2, March/April
1988.
Note: Many of the referenced articles appear
in the following volume of IEEE Press
Selected Reprint Series: EL. Kollberg,
Microwave and Millimeter-Wave
Mixers, IEEE Press, New York, N.Y.,
1984.
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
Mixers in Microwave Systems (Part 2)
The frequency-conversion function of a
mixer plays a critical role in RF and
microwave systems. Part 1 of this article
deals with mixer theory, analysis of frequen-
cy conversions, conversion loss, noise figure,
and intermodulation. Part 2 will discuss
impedance matching, diode-mixer design,
mixer realization, and use of mixers in
microwave system environments.
IMPEDANCE MATCHING
RF and IF port mismatch is a major contrib-
utor to conversion loss. Three main cases
exist: RF and image frequencies having the
same termination, image short-circuited, and
image open-circuited. For the RF and image
equally terminated, theoretical minimum
conversion loss is 3.0 dB, with IF VSWR
equal to 1:1, but with RF VSWR equal to
3:1. This means that minimum conversion
loss is obtained at the expense of poor RF
port impedance match [3]. A single-bal-
anced (two-diode) mixer design example
using computer numerical analysis shows the
real part of RF impedance, Rsig, to be about
150 ohms for signal and image equally ter-
minated. For short-circuited image, R
sig
=
100 ohms, and for open-circuited image,
R
sig
= 120 ohms [24]. Tucker has tabulated
input resistance for various modulator con-
figurations [1], and Maas has given specific
impedance values for a diode operated at 10
GHz [6]. Also, Saleh has given RF and IF
impedances for the above three cases [4].
IF impedance is real only when the image is
terminated in an open or short circuit [5].
For a single diode with the three cases given
above, the real part of IF impedance, RIF, is
approximately 200 ohms, 150 to 350 ohms
and 200 to 2000 ohms, respectively.
The real part of LO impedance has been
approximated as [11]:
R
LO
= R
s
/t (10)
where R
s
is the diode series resistance and t
is the conductance-pulse duty ratio.
DIODE MIXER DESIGN AND
REALIZATION
Many types of mixer circuits and realizations
exist. A given mixer circuit may be realized
in various ways to cover different frequency
ranges; for example, different mixers can
employ the same basic balun circuit, but can
be realized in bifilar-core, semi-rigid coax or
balanced microstrip. Many types of mixer
circuits exist: single-ended, single-balanced,
double-balanced, triple-balanced, Class IV,
and image-reject. RF, LO, and IF ports may
be interchanged in any passive mixer due to
the linear relationship between small-signal
RF, IF, and image signals.
A single-ended mixer, shown in Figure 1,
comprises a single diode with triplexed RF,
LO, and IF ports. This circuit is rarely used
because it does not provide the extra inter-
uni-planar single-balanced, image-reject
mixer designed at Watkins-Johnson
Company.
Double-balanced mixers, shown in Figures 4
and 5 as ring and star circuits comprise four
Figure 1. Single-ended mixer.
BPF BPF
R L
I L
P
F
Figure 2. Single-balanced mixer.
BPF
LPF
I
IF
I
R
I
L
B
D1
D2
C
L
A
I
L
D
I
R
i
1
i
2
Figure 3. Thin-film single-balanced mixer.
modulation suppression
given by balanced mixers.
Figure 2 shows a single-bal-
anced mixer composed of
two single-ended mixers and
a balun. A balun interfaces a
single-ended input port with
two output ports having
voltages that are equal in
magnitude but opposite in
phase. The balun isolates the
LO from the IF port, and
suppresses even-order inter-
modulation products. (Even
though the balun is at the
LO port, it causes IM prod-
ucts with even RF harmon-
ics to be suppressed.) Single-
balanced mixers are most
often found in image-reject
mixers. Figure 3 shows a
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
diodes and two baluns [25]. RF, LO, and IF
ports are all isolated by means of circuit bal-
ance. In Figure 4B, LO voltages at J3 and J4
are equal and opposite, assuming a perfect
LO balun and identical diode impedances,
so that J1 and J2 are virtual grounds with
respect to LO voltage. RF and IF output
voltages, respectively, are proportional to the
difference and sum of the residual LO volt-
ages present at J1 and J2. The RF and IF
ports are isolated through the R-port balun
alone. IF bandwidth in microwave double-
D1-D2 and D3-D4. The center-tapped LO
balun provides the IF ground return path.
RF current returns through the time-aver-
aged conductances of D1-D4 in parallel
with D2-D3. This seems paradoxical since
D1, D4, and D2, D3 are never on at the
same time during a given LO cycle.
However, the periodic conductance wave-
form for each diode has a non-varying
Fourier component, approximately equal to
1/2 of the peak conductance. This average
conductance provides the RF current path.
Triple-balanced (also known as double-dou-
ble-balanced) mixers are shown in Figure 6
as ring and star circuits, which comprise
two diode ring-quads and three baluns. The
major benefit of using a triple-balanced
mixer is very broadband IF port response.
Triple-balanced mixers with RF and LO
ports operating over 2 to 26 GHz, and the
IF port operating over 1 to 15 GHz, have
been constructed [26].
Class IV mixers, commonly known as ter-
mination (or load) insensitive [27], com-
prise two diode bridge quads and two 100-
ohm chip resistors that are embedded in a
network of 100-ohm transmission line
baluns [28]. Diode currents for IM prod-
ucts with frequency f = nf
L
mf
R
, where
m and n are both even integers, are dissipat-
ed in the two resistors. In double-balanced
mixers using ring quads, these currents cir-
culate around the diode ring, causing fur-
ther intermodulation of out-of-band signals
reflected back into the mixer. Class IV mix-
ers suppress the even-by-even products, and
so tend to have more constant conversion
loss and IM suppression as RF and IF load
impedances are varied.
Image-reject mixers (IRM) are used to sup-
press unwanted image noise and signals.
They are also commonly used as SSB
upconverters [29]. Image rejection is
achieved through phase cancellation or fil-
tering, and is defined as the ratio of avail-
able IF power to available downconverted
image power at the IF output port. As
balanced mixers are generally limited to fre-
quencies below approximately 3 GHz, due
to inductance present in the physical realiza-
tion of the single-ended IF-port.
Theoretically, assuming identical diodes and
perfect baluns, all IM products are balanced
out at the IF-port, except those having odd
RF and LO harmonic coefficients, m and n.
In all passive mixer designs, current return
paths for RF, LO, and IF circuits must exist.
For example, in the mixer of Figure 4B, the
LO current path alternates between diodes
J
1
I
IF
I
R
J
2
J
3
J
4
I
R
L L
I
L
I
L
L
I
I
R
i
3
i
4
i
2
i
1
D1 D4
D3 D2
R
Figure 4. The ring double-balanced mixer is formed by combining two single-balanced mixers.
L
I
L I
I
R
R
(A)
(B)
I
L
I
R
I
R
I
L
L L
D3
D2
i
3
i
4
i
2
i
1
D1
D4
Figure 5. The star double-balanced mixer is formed by combining two single-balanced mixers.
(A)
(B)
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
image frequency is, f
IM
= 2f
L
- f
R
, regardless
of whether f
IM
equals f
R1
or f
R2
.
The image signal is normally thought of as a
mixer-generated IM product that exits the
mixer, and which is related to image
enhancement. However, for an IRM, the
image refers to signals and noise power at
the image frequency that enter the mixer
along with the desired RF signal. Image
noise that is higher than the thermal noise
floor level (such as that generated by a
broadband amplifier placed ahead of the
mixer) will increase the system noise figure
by up to 3 dB above the expected SSB noise
figure level, because it downconverts to the
IF along with the noise associated with the
desired RF signal. The contribution of image
noise to the overall noise figure can be
reduced by using an image-reject mixer.
Equation 10 and Table 1 show that with
only 10 dB of image rejection, the image
noise contribution is reduced from 3.0 to
0.41 dB. Equation 10 is based on the defini-
tion of noise figure as being the input S/N
ratio divided by the output S/N ratio.
Downconverted image noise causes the out-
put noise power to be multiplied by the fac-
tor, (1 + IR), where IR = 10
[-IR(dB)/10]
thus
increasing overall noise figure.
Change in NF = 10 log (1 + IR) (10)
Figure 8 shows that phase-cancellation IRMs
consist of two mixers, two quadrature
hybrids and one in-phase power divider.
Mixers M1 and M2 are identical and have
IF output currents I
1
and I
2
, which are
equal in magnitude but are in phase quadra-
ture. The presence of the RF harmonic coef-
i
7
i
8
i
6
i
5
D5 D8
D7 D6
i
7
i
8
i
6
i
5
D5 D8
D7 D6
L
i
3
i
4
i
2
i
1
D1 D4
D3 D2
i
3 i
4 i
2
i
1
D1
D4 D3
D2
I
IF
I
IF
I
IF
I
IF
I
R
I
R
I
L
I
L
R
I
R
I
R
R
L
L
(B)
(A)
I
I
Figure 6. A triple-balanced mixer is formed by combining two double-balanced ring or star mixer.
Figure 7. Image-reject mixer configuration, and corresponding frequencies.
RF
INPUT
R
LO
INPUT
I
1 I
1
I
2
I
2
f
I
OUTPUT
f
R1
f
L
f
R2
INPUTS
L
shown in Figure 7, the RF and image frequencies, referenced to the LO frequency, are the
mirror images of each other. If the RF frequency is defined as f
R1
, then the image is f
R2
. The
Image Rejection Change in NF
(dB) (dB)
0 3.0103
10 0.4140
20 0.0430
30 0.0043
Table 1. Image-noise contribution to noise figure as a
function of image rejection.
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
ficient, m, in the phase angle of the mixing
products, I
1
and I
2
, is a result of the power
series expansion for the diode current-volt-
age characteristic.
Since I
2
= j
m
I
1
and setting:
|I
1
| = |I
2
| = I,
the currents exiting the IF quadrature cou-
pler are:
I
1
= (I/2) (1 + j
(m+1)
)
= I for m = -1 (f
L
- f
R1
)
= 0 for m = +1 (f
R2
- f
L
) (11a)
I
2
= (I/2) (j + j
m
)
= 0 for m = -1 (f
L
- f
R1
)
= jI for m = +1 (f
R2
- f
L
) (11b)
I
1
and I
2
combine in the output quadrature
coupler so as to channelize the (f
L
- f
R1
)
product into port I
1
, and the (f
R2
- f
L
) prod-
uct into port I
2
.
Image rejection is a function of the cumula-
tive amplitude and phase imbalance of the
hybrids and mixers, and is given as,
IR [dB] = -10 log (12)
Figure 8. Block diagram of an image-reject mixer. A
single sideband mixer is formed by reversing RF and IF
ports.
0 1 2 3 4 5 6
AMPLITUDE IMBALANCE (dB)
7 8 9 10
0
10
20
30
40
50
I
M
A
G
E

R
E
J
E
C
T
I
O
N

(
d
B
)
0.1
1
5
10
20
40
2
Figure 9. Image-rejection versus amplitude and phase imbalance.
RF
0 0
90 90
50!
M1
I'
1
I
1
I
2
I'
2
M2
RF
QUADRATURE
HYBRID
IF
QUADRATURE
HYBRID
LO
(1 + A
2
- 2A cos ")
(1 + A
2
+ 2A cos ")
passive balun structures composed of various
types of transmission lines. Recently, howev-
er, active-balun mixers have been designed
for MMIC circuits, with the goal of reduc-
ing the amount of GaAs surface area
required. Diode and FET mixers also have
been built using passive and active baluns.
Balanced mixers operating in the frequency
range of 1 to 3000 MHz generally use trans-
mission line baluns composed of bi-, tri-, or
quadfilar wire wrapped on ferrite cores.
These structures are multioctave, employing
magnetic coupling up to about 200 MHz,
and electric coupling up to frequencies of
about 3000 MHz.
Microwave mixers operating above approxi-
mately 2 GHz are realized using various
combinations of micro-strip, coplanar wave-
guide (CPW) and slotline. Most microwave
mixers are built using soft dielectric balanced
microstrip with soldered-in diode ring-quad
packages. The major challenges to these
designs include the crossover of the RF and
LO lines shown schematically in Figure 4B,
and the dual-sided nature of most broad-
band microwave balun structures. A number
of catalog mixer designs exist that provide
various tradeoffs among conversion-loss,
bandwidth and intermodulation perfor-
mance. Mixers realized using CPW and slot-
line [30,31] have increased in popularity due
to small size, ease of fabrication and low
conversion-loss. [32].
A number of MMIC mixer circuits have
been described recently that use multioctave
distributed active baluns [33, 34]. Passive
baluns are also used in MMIC mixers. One
approach, requiring a minimum of GaAs
space, consists of printed spiral transformers
[35]. These have bandwidths of 4:1 with
about 1 dB of insertion loss, compared with
the 6 dB of added noise figure typically
found in distributed active baluns [36].
Various bipolar Gilbert-cell mixers have been
described, which offer conversion gain and
small size, but at the expense of higher noise
figure [37].
MIXERS AND SYSTEM SPECS
Various tradeoffs exist between gain, noise
figure, compression, and intercept point
when cascading mixers with other devices in
systems. For example, if an amplifier and
Figure 9 gives image rejection as a function
of total phase and amplitude imbalance. It
shows, for example, that to achieve 20 dB of
image rejection, amplitude imbalance must
be less than about 1.6 dB, and phase imbal-
ance must be less than about 12 degrees.
MIXER REALIZATION
Balanced mixers are generally realized with
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
nals. When this is impractical, group delay
can be approximated by placing two mixers
in an up-down configuration and halving
the resulting delay to get group delay for one
mixer alone.
MAKE OR BUY DECISION
Whether to build or buy a mixer involves
trading off such factors as cost, performance,
availability, packaging, testing and screening.
The cost of designing and building a high-
performance mixer may indeed be higher
than that of buying an existing catalog
model. When catalog mixers do not meet
the required performance, they can generally
be modified by the manufacturer. For exam-
ple, to achieve higher levels of compression
and IM suppression, the diode can generally
be replaced with one having a higher barrier
level. Testing for phase and gain, match and
track, is routinely done by mixer manufac-
turers who have large quantities of mixers to
select from and automated test stations set-
up specifically for this purpose [40]. Also,
QPL (Qualified Products List) mixers are
increasingly being used to reduce cost and
delivery times [41].
CONCLUSION
This article has summarized the topics of
mixer theory, design, realization, and usage.
It was shown that image enhancement tech-
niques must be used to minimize conversion
loss, and that the image termination should
be a short circuit, rather than an open cir-
cuit, in order to minimize noise figure and
third-order intermodulation. Conversion-
loss ripple of up to 5 dB peak-to-peak can
result when filters are placed adjacently to
broadband mixer ports. Theoretical limits
for conversion-loss are 3.92 dB for conju-
gately matched broadband mixers, 3.0 dB
for mixers with conjugately matched IF and
equally matched signal and image, but with
reactively terminated idlers; 0 dB optimally
matched signal and IF, and for reactively ter-
minated image and idlers. A broad array of
mixer circuits exist, which are commonly
mixer are cascaded, the amplifier should pre-
cede the mixer to minimize overall noise fig-
ure, but the opposite arrangement would be
required to maximize overall intercept point.
Cascaded third-order output intercept point
for two stages has been given as [38]:
OIP3 (dBm) =
-10 log +
where OIP3
n
and G
n
are the algebraic third-
order output intercept and gain of the nth
stage. This formula assumes linear IM sup-
pression relative to the IF product, and cas-
caded voltages all adding at (worst-case)
phase maximums. The cascaded third-order
output intercept point is maximum when g
2
is large, indicating the amplifier should fol-
low the mixer to optimize intercept point.
Cascaded 1-dB power compression can be
approximated for amplifiers using the same
formula as for cascaded intercept point [39].
This relationship is based on the fact that for
amplifiers, output power at 1-dB compres-
sion is generally 10 dB below the two-tone
third-order output intercept point. The out-
put 1-dB compression and third-order out-
put intercept points in mixers are generally
less than 10 dB apart and are less pre-
dictable, so this relationship should be used
carefully for mixers cascaded with amplifiers
and other devices.
GROUP DELAY
Group delay for RF and microwave mixers is
in the range of 0.350 to 0.500 ns. There is
no inherent group delay increase in a passive
mixer, except that which is caused by the
transmission line lengths and reactive ele-
ments that are present in the mixer circuit.
Group delay of broadband mixers can be
measured by pulsing the RF input signal and
measuring delay using a fast oscilloscope
with and without the mixer present. The dif-
ference in delay equals the group delay. This
method requires the oscilloscope to be fast
enough to display both the IF and RF sig-
1
OIP3
1
G
2
1
OIP3
2
realized using soft dielectric balanced
microstrip and other transmission-line struc-
tures. Present areas of design include using
uni-planar thin-film balun structures to
minimize device cost and size, usage of
MESFETs to achieve wider dynamic range
than possible with Schottky diodes, and
designing compact broadband balun struc-
tures for MMIC mixers.
ACKNOWLEDGEMENT
The author wishes to thank SE. Avery, Dr.
R.K. Froehlich, and Dr. R.G. Ranson for
their thoughtful review of this paper, MA.
OMahoney, and T.G. Skala for their non-
linear CAD evaluations, R.W. Bruce for the
usage of his extensive bibliography, and
R.Y.S. Parsons for preparing the draft.
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
REFERENCES
1. Tucker, D.G. The Input Impedance of
Rectifier Modulators, lEE Proc.,
Vol. 107B, No. 1, pp. 273-284,
January 1960.
2. Gardner, J.G., et al. Distortion
Performance of Single-Balanced Diode
Modulators, lEE Proc., Vol. 117,
No. 8, August 1970.
3. Kelly, A.J. Fundamental Limits on
Conversion Loss of Double Sideband
Resistive Mixers, IEEE Trans. Microwave
Theory Tech., Vol. M71-25, No. 11,
November 1977, pp. 867-869.
4. Saleh, A.A.M. Theory of Resistive
Mixers, Cambridge, Massachusetts:
MIT Press, 1971.
5. Torrey, H.C. and C.A. Whitmer.
Crystal Rectifiers, New York: McGraw-
Hill, 1948.
6. Maas, S.A. Microwave Mixers, Artech
House, Dedham MA, 1986.
7. Held, D.N. and KR. Kerr. Conversion
Loss and Noise of Microwave and
Millimeter-Wave Mixers: Part 1-Theory
and Part 2-Experiment, IEEE Trans.
Microwave Theory Tech., Vol. MWI-
26, No.2, February 1978.
8. Burkley, C.J. and R.S. OBrien.
Optimization of an 11 GHz Mixer
Circuit Using Image Recovery, mt. J.
Electron, Vol. 38, pp. 777-787,
June 1975.
9. Mass, S. Two-Tone Intermodulation in
Diode Mixers, IEEE Trans. Microwave
Theory and Tech., Vol. MTT-35,
No. 3, March 1987, pp. 307-314.
10. Oxley, T.H. Phasing Type Image
Recovery Mixers, IEEE MTT-S mt.
Microwave Symposium Digest,
pp. 270-273, 1980.
11. Dickens, LE. and D.W. Maki. An
Integrated-Circuit Balanced Mixer,
Dec. 1977.
23. Gretsch, W.R. The Spectrum of
Intermodulation Generated in a
Semiconductor Diode Junction, Proc.
IEEE, Vol. 54, No. 11, November 1966.
24. Faber, MT. and W.K. Gwarek.
Nonlinear-Linear Analysis of
Microwave Mixer With Any Number of
Diodes, IEEE Trans. Microwave
Theory Tech., Vol. MYI-28, No. 11,
Nov. 1980.
25. Mouw, RB. A Broad-Band Hybrid
Junction and Application to the Star
Modulator, IEEE Trans. Microwave
Theory Tech., Vol. MTI-16, pp. 911-
918, Nov. 1968.
26. Henderson, B. Orthogonal Mixers:
Punching Up Earth/Space Payload
Performance, MSN, January 1982,
Vol. 12, No. 1.
27. Will, P. Termination Insensitive
Mixers, Professional Program Session
Record 24, WESCON, San Francisco,
1981.
28. Norton, D. Three Decade Bandwidth
Hybrid Circuits, Microwave Journal,
Vol. 31, No. 11, Nov. 1988, pp. 117-126.
29. Henderson, B. and J. Ceok. Image
Reject and Single Sideband Mixers,
MSN, Vol. 17, No. 9, August 1987.
30. Aikawa, M. and H. Ogawa Double
Sided MICs and Their Applications,
IEEE Trans. Microwave Theory Tech.,
Vol. 37, No. 2, February 1989.
31. Hirota, T., Y. Tarusawa and H. Ogawa.
Uni-planar MMIC Hybrids A
Proposed New MMIC Structure,
IEEE Trans. Microwave Theory Tech.,
Vol. 37, No. 2, February 1989.
32. Izadian, J., et al. A Uni-Planar Double-
Balanced Mixer Using A New
Miniature Beam Lead Crossover
Image and Sum Enhanced, IEEE
Trans. Microwave Theory Tech., Vol.
MTT-23, No. 3, March 1975.
12. Pound, R.V. Microwave Mixers.
Usington, Massachusetts: Boston
Technical Publishers, Inc., pp. 81-87,
1964.
13. Shurmer, H.V. Microwave Semiconductor
Devices. New York: Wiley Interscience,
1971.
14. Sze SM. Physics of Semiconductor
Devices. New York: Wiley Inter-science,
1969, p. 459.
15. Sabin, WE. and E.O. Schoenike.
Single-Sideband Systems and Circuits,
New York: McGraw-Hill, 1987.
16. Steinbrecher, D.H. Mixer
Fundamentals, Notes presented at
1989 RF Technology Expo,
Santa Clara, CA.
17. Bain, R. A Mixer Spurious Plotting
Program, RF Design, May 1989,
pp. 32-43.
18. Weiner, S., D. Neuf and S. Spohrer.
2 to 8 GHz Double Balanced
MES-FET Mixer with +30 dBm Input
3rd Order Interecept, IEEE MTT-S
Int. Microwave Symposium Digest,
pp. 1097-1100, 1988.
19. Maas, SA. A GaAs MESFET Mixer
With Very Low Intermodulation,
IEEE Trans. Microwave Theory Tech.,
Vol. MIT-35, No. 4, April 1987.
20. Oxner, E. A Commutation Double-
Balanced MOSFET Mixer of High
Dynamic Range, Proceedings 1986 RF
Expo East, pp. 73-87.
21. Henderson, B.C. Reliably Predict
Mixer IM Suppression, Microwaves
and RF, Vol. 22, No. 12, Nov. 1983.
22. Cheadle, DL. Consider a Single Diode
to Study Mixer Intermod, Microwaves,
WJ Communications, Inc. 401 River Oaks Parkway San Jose, CA 95134-1918 Phone: 1-800-WJ1-4401 Fax: 408-577-6620 e-mail: [email protected] Web site: www.wj.com
The Communications Edge

Tech-note
Author: Bert C. Henderson
Copyright 1990 Watkins-Johnson Company
Vol. 17 No. 2 March/April 1990
Revised and reprinted 2001 WJ Communications, Inc.
Quad, IEEE Int. Microwave
Symposium Digest, pp. 691-694, 1988.
33. Pavio, AM., et al. Double Balanced
Mixers Using Active and Passive
Techniques, IEEE Trans. Microwave
Theory Tech., Vol. 36, No. 12,
December 1988.
34. Titus, W., et al. Distributed
Monolothic Image Rejection Mixer,
IEEE Int. GaAs IC Symposium,
pp. 191-194, 1986.
35. Mi, F., S. Moghe and R. Ramachandran.
A Highly Integrated X-Ku Band
Upconverter, IEEE GaAs IC
Symposium Digest, pp. 157-160, 1988.
36. Private communication with AF. Podell.
37. Fotowat, A. and E. Murthi. Gilbert
Type Mixers vs. Diode Mixers,
Proceedings RF Technology Expo 1989,
pp. 409-413.
38. Wilson, SE. Evaluate The Distortion
of Modular Cascades, Microwaves,
Vol. 20, March 1981.
39. Sorger, G.U. The 1 dB Gain
Compression Point for Cascaded Two
Port Networks, Microwave Journal,
July 1988.
40. Avery, SE. Dual Mixers, Watkins-
Johnson Company Tech-notes, Vol. 13,
No. 4, July/August 1986.
41. Sehindler, SA. MTL-Specification
Mixers, Watkins Johnson Company
Tech-notes, Vol. 15, No. 2, March/April
1988.
Note: Many of the referenced articles appear
in the following volume of IEEE Press
Selected Reprint Series: E.L Kollberg,
Microwave and Millimeter-Wave
Mixers, IEEE Press, New York, N.Y.,
1984.