[ Pobierz całość w formacie PDF ]
low-cost
microwave spectrum analyzer
guidance and encouragement to anyone attempting
a similar project.
performance requirements
The operation of any spectrum analyzer can be
characterized in terms of its frequency coverage,
dispersion, dynamic range, sensitivity, and resolu-
tion. To display frequency components well into the
microwave region, I designed my spectrum analyzer
to cover dc to at least 2 GHz . The same design
strategy could be easily applied to other frequency
bands . In fact, the upper frequency limit of th is
analyzer was later extended to
2.5
GHz, as discussed
later.
Dispersion describes the ability of a spectrum
analyzer to display a broad slice of the frequency
spectrum in a single sweep. Many of the low-cost
analyzers on the surplus market display only a few
MHz at a time. Such narrow-dispersion spectrum
analyzers are useful as panadaptors, which display all
signals within several hundred kHz of a specified
operating frequency, but when tuning a microwave
local-oscillator chain , monitoring mixer image
response, or measuring transmitter harmonic con-
tent, it is often desirable to display a band several
hundred (or even thousand) MHz wide. The spec-
trum analyzer shown here can display the spectrum
from dc to
2
GHz in a single sweep. Since it's often
desirable to narrow this sweep for a closer look at a
particular signal, variable dispersion capabilities are
included in the design.
Sens it ivity and dynamic range define the minimum
and max imum signal amplitudes which an analyzer
can display without distortion. In accordance with
good engineering practice,
I
try to suppr-ss all
transmitted spurious products by
50
dB or more. To
accurately measure this performance, the spectrum
analyzer requires at least
50
dB of dynamic range . As
How to put together
a microwave
spectrum analyzer
from su rplus
odds and ends -
the completed unit
covers dc to 2 GHz
My
ongoing efforts to develop low-cost modules
for the 1296-MHz band have often required the use
of a spectrum analyzer for monitoring (and mini-
mizing) harmonic and spurious frequency com -
ponents. Numerous excursions through the local
surplus test equipment emporiums revealed that an
acceptable instrument could cost several thousand
dollars - well beyond the budget of the most
dedicated experimenter. While searching for the
unbeatable surplus buy which
never materialized,
I
noticed the ready availability and comparatively low
cost of a wide variety of S-band (2-4 GHz) test in-
struments and components. It occurred to me that a
microwave spectrum analyzer could be put together
from these available parts at a cons iderable savings.
This article documents the design, construction,
operation, and performance limitations of the
resulting microwave spectrum
analyzer.
While I
doubt that any reader will want to duplicate my
design in its entirety,
I
hope this article w ill provide
By H.
Paul Shuch,
WA6UAM,
Microcomm,
14908 Sandy Lane, San Jose, California 95124
54
r.I
august 1977
for maximum input level, I often want to display a
+
10 dBm (10 mW) signal (this is the local -oscillator
injection level required of many balanced mixers).
Thus, 50 dB dynamic range with a
+
10 dBm max -
imum input level yields an ultimate sensit ivity require-
with this design, I can resolve frequency components
to within about 2 MHz.
As most amateurs know, a general -coverage com -
munications receive r can be used as a rudimentary
high-frequency
spectrum
analyzer.
With
an
input
UP CONVERTER A SSEMBL Y
CAVI T y BANDPASS
r u,
TfR '
'0
r
2 GHI
o
r 2 0 0 0
DOliBLE
BALANCED
M,x ER
RElCOM MI - f
MICROLAB/~XR
l A- .}ON
l PF'
-6(18
F---------->
---f
I\~--;
AA-Qf.
N
F IX ED
I' AO
~ICROLAtJ/fXR
RF
IN PU T
OC- 2GHI
(-r-
-+-
--/
IOm W "' AX
3 STAGES
1N21
IN
OETECTOR
MOUNT
SWEP T
LOCA L
OSCILLATOR
ASS EMBL Y
2 - 4GHz
-+'OdBm
RF
COA X
TE E
7.~1l
VI DEO
LOAD
fig. 1. Block d iagram of the microwave spectrum analyzer. Most components were purchased on the surplus market. The swept local oscillator,
the key to the analyzer. is a surplus 2-4 GHz backward-wave oscillator. The display is an ordinary oscilloscope with dc coupling and provisions
for external sweep.
ment of - 40 dBm, or 0.1 p.W. Greater sensitivity (a
lower minimum discernible signal) could have been
obtained, but only at the expense of dynamic range.
The objective of any spectrum analyzer is to
display the various components of a complex
waveform in the frequency domain. The closer the
frequencies of any two components, the more dif-
ficult it is to separate them on the spectral display.
Resolution relates to the minimum frequency separa-
tion between two signals of equal amplitude which
will still permit the operator to discern two separate
frequency components on the display.
Resolution can be approximated as twice the i-f
bandwidth of the analyzer system. Generally, the ob -
jectives of wide dispersion and narrow resolution are
mutually exclusive. When measuring transmitter
audio intermodulation products with a two-tone test,
for example, a resolution of a few hundred Hz is re-
quired, and dispersion is likely to be several tens of
kHz . When viewing harmonics of a 100 -MHz
oscillator, on the other hand, 2-GHz dispersion may
be required, but a resolution of several tens of MHz is
acceptable.
Resolution is primarily a function of i-f bandwidth,
which for my analyzer is fixed at 1 MHz. Therefore,
signal applied to the antenna terminals, the receiver
is manually tuned through its frequency range
(dispersion). Frequency components are detected
and displayed (perhaps w ith the receiver 's S-meter).
Resolution is a function of the i-f bandwidth, which is
probably a few kHz. Sensitivity is a function of the
receiver's noise floor, and dynamic range is limited
by the receiver's agc and overload characteristics .
Obviously, w ide dispersion measurements require
considerable operator intervention, in the form of
tuning. Gain variations of the receiver from band to
band will limit the accuracy of its amplitude indica-
tion. Additionally, any nonlinearity in the receiver's
age circuit may prevent accurate amplitude measure-
ment across the receiver's entire dynamic range .
Also, the receiver's image and spurious rejection may
be insufficient to eliminate false indications.
Ideally, a workable spectrum analyzer should be a
superheterodyne receiver in which these shortcom-
ings are minimized. Frequency tuning should be both
automatic and rapid . Instead of an S-meter,
amplitude is displayed on an oscilloscope. If the
scope's horizontal deflection is slaved to the
receiver's tuning mechanism , the result is a display in
the frequency domain. Dynamic range must be max-
august
1971
rml
55
Alfred 622BK sw eep os cilla to r. bu t any similar
generator sho uld work satis fa ctorily. These sw eep
ge nerators co nsist o f a vo lt age -co ntrolled oscillator ,
typica lly a bac kward wave oscillator or BW O Ie
microwave oscill a to r bu ilt arou nd a device similar to a
traveli ng wave tu bel, power supplies, a sawtooth
generat or for dev eloping a co nstan tly varyi ng vco
control volt age, and leve ling circuitry to mainta in
constant ou tpu t across the band .
S tart
and
s top
fr e-
que ncy adjustm en ts permit the osci llator to sw eep
all , or any porti on of , the 2 to 4 GHz b and . l eveled
outpu t power is ty pically 10 t o 30 mill iwatts.
M any companies are cur rently retiring their BW O
sw eep generators in favor of wioeba nd. solid -sta te
units. so quite a few BWO generators have recently
appeared on the surplus ma rket at prices rang ing
fr om $200 to $400 or so . Sinc e thi s is the m ost c ostly
component o f the microwave spectrum analyze r,
ma ke sure th e un it you buy is in good ope rating con -
dition . Reputab le elec tro ni cs sur pl us dealers will
often le t yo u po wer up an instrumen t and make a few
measurements prior to purchase. A prac tical test re-
qu i re s the u se o f a mic rowave power meter
(bol ometer bridge or equivalent ) t o observe output
power in th e leveled CW m od e as the generator is
m anu ally tu ned ac ross th e band . A lthou gh
a
few dB
var iat io n is acceptable, dead spots or severe power
drop-off a t the high end o f th e band indic ates a fa il-
ing BWO. A good, used a W Q should provide veers
of reliabl e life in inte rm itt ent amateur servi ce.
•
,
,TIO
'J ••
f
•
•
•
HUNit
u
", nl'
1>(1; '"0"
;:).,,"10
IHT,O .
•
F.onl p"n,,1o l l h"
mic,ow"~,,
s p "ctr u m a nalv,e. s how in g 1h .. o p " ' -
atinq con l .ol. Th ..
~e,iabl"
I-I '"l""ualo' fu p p " . ' "I ll . although c,, 1i
b , al"d I" 6 dB SloPS. p ..,m". " m p lilu d " c o m p" ' i"o" o f "i g " " ,,, d is
p l" V"d 0 "
Ih " o .cillo .co p"
Th " i·f b"ndpau lil1", te " " I"
IS ' h "
f ' ..q u " n c v
co~,,'''g''
0 1 . h.. " na ' v ''''. u
di "cuu ..d in ' h
'
•
•
•
im ized, and spurious / image responses eliminated. to
the greatest possible exten t .
M any of the obj ectiv es di scuss ed prev iously are
m et in the de sign show n in f ig . 1. a wi de dynamic
range m icr owave rec eiver w ith an electron ic all y
tun ed local osci lla to r and ample image rejec tion . It in -
cludes a 2-GHz
i-t
amplifier w ith variable ga in and fix-
ed bandw id t h, and a sensit ive detector for dri ving an
osc illosco pe. Unlike co nven tio nal rece iv ers , t his
desig n up-co nve rts the in coming signal to an i-f in
the m icr owave regi on . A lthoug h thi s app roach com -
plicat es i-f design . it permits wide dispersio n tuning.
It also improves separa tion o f the rf and im age
signals so a simple lowpass f ilte r can b e used
to
eliminat e image responses.
The mic rowave spec trum analyzer is divided in to
three sepa ra te sub-systems : th e loc al os cillator ,
u ni ty -gain upconverter. and display sec tions .
lo c al oscillator
Cen tral t o the design of th is spec trum analyzer w as
the availability, on th e sur plu s ma rket , of a leveled .
sw ep t signal sourc e co vering 2 t o 4 GHz. I used an
In,.. . lo . of 'h" s p "ctrum """Iv, ..,
con~
..r, .. , " ..c . io n s hOw ing ' h ..
doubl.. ·b " ' " ,,c ed mi . " , ."d an."u.,io " p "d". inp u' l if'", . I ' f ilt. , .
" n d ".,i.bl.. l-' " n ,, " u " lo ' .
56
~~
august 1977
.
.
-I
COOO Vf"- <0 1<'
.
..
.-
.
.
_
_
. 0 ... .
I ~
_ ~
I·..
·.-I·~
__
- -
,.
-.7-
- 0
-
- +-
-
_
1-
2 0
, 0
- 10 ..... /
,
,
:"9111 -4,- 1 1_,,0_,, _1
1--"' -..
- 1
1··..,0-64 -1
,
,
,-
-
.
-1 ..
- 1)
1-
0
•0
.0.0 _1
1°.
0
- '10 ....
1
f ill , 2.
P o ......
1.... .
1. In Ih. upI;:O nV. ' TIl. u .ed ....ll h Ih. mk. O....IlY.
' P IlI;:T. u m " n.lY.Il. (....il h Ihll v" .lllb lll ll llll n UIlIO' . 11 1 liTO dB . ....h lch , .. . u l" in u n; l y
conYllrilo n 1I" lnl,
As can be seen in the block diagram, fig . 1. th e
spectr um analyzer co nverter assemb ly consists of an
input lowpass filter. an $ -band double-balanced
mix -
er. so me attenuator pads , a high·Q i -f f ilter . a
variable
i-f
a rte nua to r.
and sufficien t i-f am plification
to bring maximum co nversion tJain to un ity . The i-f
det ect or and its video load. thoug h installed in th e
co nverter assemb ly,
siderable con fusion on the display . Thus an input
low pass filter w as installed in the system to block all
sig nals above 3 GHz fr om enterin g the
me er .
I
used a
M icrolab /FXA LA -JON filt er w hich I salvaged f rom
ano ther piece of eq uipm e n t. Al t hough t he f ilte r
o rig inally co st $40, simil ar devices are availa ble
th rou gh su rplus outle ts for $5 to $10.
are discussed
later
with
th e
display .
Th e charac te ristics of the balanced miller . more
than any o ther co mpo nent . establish th e linearity and
dyn amic range of the analyzer. I used a Relcom M1F
mixer w hich
I
fou nd on the surplus market for $35
{t he mi xer retails fo r abo ut $2001. The rated fr eq uen -
cy respon se of t his m ixer is dc-2 GHz at t he i-t port,
and 2-4 GHz at the rf and LO ports . Not e that the in -
coming sig nal is applied to the
i-t
p ort ;
t he
rf port
dr ives th e i·f system . Thu s. all ports are ope rated
withinn t heir specif ied trequencv ranges.
W ith th e 10 mW of local -oscillato r inject ion applied
to t he mixer from th e sweep generator, the mixer' s
conversion eHiciency is co mpressed by 1 dB at an in -
put signal level of
1
mW. Since
I
w anted to analyze a
10 mW sig nal on the spec trum analyzer withou t ex-
ceeding
1
d B co mpression, it wa s necessary t o place
a 10 d B attenuation pad ahead of the mixe r's input
Ii -f
po rtl .
T his pad also assu res proper impedance
term inati on for the mi xer, as does the 6 d B at-
tenuatc r
at the output
(rf
oortl .
Fixed
atte nuetors
f or
dc to 2 GHz are available to the surplus ba rgain
hunter for as little as $5.00. or may be purchased new
for $15 to $20.
With 2 GHz
i-t .
and a swe pt LO cov ering 2 to 4
GHz, th e mixer will respond to signals in the dc-Z
GHz region . as we ll as in the 4·6 GHz image band .
A ny compo nents in the image band w ill cause con -
Spllc trum .n.IY' Il' I-I . 1 1;:1;On _Th. Ih," " Z-G Hr _m pl ll i" rl ll ' .
,, "
h ll
b Oll om
Clodll dll'Il<:TO. I. III uPPIl. I" "
augus t
1977 ~':; 57
..
filter to vary
i-f
gai n . The at t enuat or I used was also
salvaged from the T5 -406 n oise ge nerator, but any
continuously va riable or st ep attenuator ra ted to 2
GHz is accep table - 10 d B steps will allow coarse
system gai n con trol;
if
1 dB resolution is included.
The j·f bandwidth of this analyze r is estab lished by
a high -Q tunable c oaxial or cavity fi lter which is tun-
ed to 2 GHz . I used the filter from a surplus T5-406
n oise generator, but any cavity with a
Q
o f 1000 or
greater should be acceptable . It's also possible 10 use
Enmpl. 0 1
s p ec tr l l
impurity ,
I '
d il p ll ye d
On
\h" m ie'ow"". t p ",Cl t u rn
e n,, 'v " " ·
P ' .....
ne.
0 1 h .. .
monic . I"b h"rmonic. Ind .pu,io"t lign"lt
s h o wn in
A
i ,
lh ..
r n u lt
0 1 en overdrive" uhl " mphlie' T h..
urn"
"rnplili • •. with drl"" ' ''duc''d to t he , ,, le d le v" l. ,• • h own '"
8 : I h " On" .pu,; ou.
c o m p o ne n t i. d o wn b.,. mo,e th"n 20 dB . Oi" p l" .,. I, l , o m d e 10 2
GH~
a
trensmesion -mode
cavity wavemeter as an i-f f ilt er.
Th ese widely available de vices have a loaded
Q
of
several thousand . and exhibit only a
few
d B of inser-
tion loss at resonance. N o te that an eb sc eotlcn -type
wa vemeter is
not ac cep table
because the f ilte r mu st
p as s
the et te nuator can also be u sed for accu rate signal
level com parison. T his is ac com pl ished by view ing
one signal c omponen t. sett ing a convenien t refer-
ence level on the display, varying the arrenuaro r for a
like indication on the other signal component , and
noting t he change in attenua tcr settmqs .
Consi derable
i-t
ga in is req uired t o ach ieve the
desi red sensitivity. I cascaded th ree stages of the
Mic r oc o mm PA · 13 b u ffe r am p li f ier. " T h e se
microstripline amp lifier modules
o Her
10 dB of gain
pe r stage across the 2.0·2 .3 GHz band , and are
biased for 30 mW output at 1 dB gain co mpression.
Sin ce i·f noi se f igu re is not a limi ting fac t or so far as
system sensitivity is concerned , any available w id e
d ynamic range amplifier for 2 GHz may be used .
display
T he l oc a t-o sc ttta t or sig n al fo r t h e spe ctru m
analyze r is swep t by a saw tooth w avefo r m ;
th erefore, displayi ng a signal in the frequency do-
m ain is simply a matter o f dete cting the ou tpu t sig nal
fr om the i·f amplifiers, applying the recove red video
to the vertical defle ction amplif ier o f an osci llos co pe,
and applying th e saw tooth ou tput vol tage f rom th e
sw eep generat o r to the oscillosc ope's horizontal
axis . Since a relat ively slow sw eep rat e is used , the
i-f
maximum
signal
t o
the
am pl ifiers
at
reson ance.
A 50 dB variabl e at tenuat or was in stall ed after the
" Ava ila ble IOf $6495 pe r s tage (plu s postage and halldhn gl trom
M,(:rQ<: omm. 14908 Sandy Lane. Sa n Jose. Ca lifornia 95124 _
Mie , o . l, ; p lin e
s id"
0 1 t h e
2 ·G HI
"mphfie'
IM ie ,oeo m m
PA -lll.
Th, .... o lth ..." u n its p ' o vi d ,, ](1 dB ,,"in .. 1 2 G HI
58
~:;
a u g ust
1977
[ Pobierz całość w formacie PDF ]