Coaxial Magnetron
Coaxial Magnetron
Project Report
ATC-74
M. Labitt
24 June 1977
Lincoln Laboratory
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
LEXINGTON, MASSACHUSETTS
FAA-RD-77-3l
M. Labitt ATC-74
9. Performing Organization Nome and Address 10. Work Unit No. (TI1AIS)
This work was performed at Lincoln Laboratory, a center for research operated by Massachusetts
Institute of Technology under Air Force Contract F19628-76-C-0002.
16. Abstract
Coaxial Magnetron
Stability Document is available to the public through
Spurious Response the National Technical Information Service,
Airport Surveillance Radars Springfield, VA 22151
MOVing Target Indicator
19. Security Clonil. <01 this report) 20. Security Clossil. Col this poge) 21. No. 01 Poges
iii
ILLUSTRATIONS
13. through 29. Coaxial magnetron (QKH-1739LL) operating at 2.7 GHz. 21-25
iv
ILLUSTRATIONS (continued)
~ Page
v
COAXIAL MAGNETRON SPECTRA AND INSTABILITIES
I. INTRODUCTION
A. Coaxial Magnetron
It was found that the present method of increasing rise time, i.e. adding
inductance in series with the pulse forming network (PFN), was inadequate. This
method did slow the rise time, but it also increased the fall time to such an
extent that the RF envelope was severely distorted (Figure 1), and the spectrum
degraded. Also, the additional inductance mismatched the modulator and essen-
tially transformed the PFN into a one-section network.
B. Conventional Magnetron
Initially it was believed that the diode network concept would be useful in
controlling the spectrum of the DX276. In the normal configuration a two-section
PFN is used together with an RC despiking network. This results in a long RF
2
l ATC-74(1) L
WI
3
l ATC-74(2) 1__
Operation of Diode Network of Fig. 2: After the first pulse, the 250 pf
capacitor charges to the magnetron pulse voltage and then decays to a value
below the conduction voltage of the magnetron. This voltage is set by the
Zener diode stack. On the second and subsequent pulses, the fast recovery
diodes are normally biased off. When the magnetron pulse is applied the
capacitor loads the pulse transformer increasing the rise time after the
pulse voltage exceeds the Zener stack voltage. This loading only lasts
approximately as long as the time constant RC = 250 nsec, where R = 1000
ohms is the source impedance of the modulator.
Magnetron
Fast
Recovery
Diode
Output Stack diode stack
Pulse
Transformer
Diode network
4
l ATC-74(3a) l-
l ATC-74(3b) ~
5
l ATC-74(4) L
l ATC-74(5) L
6
tail which causes the high side lobes on the low frequency side of the spectrum.
To control the long tail the old PFN was replaced with a lO-section network.
This network reduced the fall-time but also reduced the rise-time. The diode
network was then used to control the fast rise time (as with the coax magnetron)
in order to prevent the tube from moding.
Figure 6 shows the spectrum using the lO-section network and the diode
network. Figure 7 shows the standard configuration spectrum. An improvement
is apparent but it is not deemed worth the additional complexity. DX276 measure-
ments described in the rest of this report are all made using the standard con-
figuration of the ASR-7 radar transmitter/modulator.
Modern MTI radar processors require that the transmitter have a high
degree of frequency stability. The effect of frequency fluctuation can be demon-
strated by the following situation. Consider the radar return from two ranges
separated by a half a pulse length at the same azimuth. Upon arrival at the
radar the two returns will partially overlap. If the radar frequency varies from
pulse to pulse, then the resultant signal level at the overlap point will fluctuate.
Most modern MTI devices process the signal after the steady component has been
removed. Thus, in a conventional MTI radat frequency instability will increase
I
the number of false alarms, while in a CFAR MTI the frequency instability will
result in a loss of sensitivity.
Consider a perfectly stable radar (other than the magnetron frequency) look-
ing at stationary clutter and define the ratio of the fluctuation to steady com-
ponents as ¢, T the pulse length; then the relationship to the frequency filter
is:
f
13¢ (1) *
rms TIT
------------
*Equations (1) and (2) are derived in the Appendix.
7
-I ATC-74(6) I~
Fig. 6. DX-276 spectrum using diode network and 10-section PFN (V-axis:
10 dB/div; H-axis: 10 MHz/div).
-I ATC-74(7) L
8
if the coho is locked to the center of the magnetron pulse. If the coho is
locked to the tail of the pulse, then the requirement is more stringent,
f
__13¢ (2) *
rms 2nT
Normally the coho is locked to the tail, but it does not prove to be much of
4
a problem to do mid-pulse locking. If a residue-to-c1utter ratio of 0.16 x 10-
(-48 dB) and a pulse length of 0.7 ~sec are assumed, then for a center-locking
system, the magnetron jitter should not exceed
f 3150 Hz (3)
rms
2. Measurement Technique
E kAB sin e
where E is the output voltage, A and B are the amplitudes of the two signals,
k is a constant and e the phase difference. Notice that E can only be greater
than zero when both A and B are greater than zero. Consequently, the phase
detector output will appear on a scope as in Figure 9a in an idealized form.
A sample and hold (S/H) is set to sample the phase detector in the center of the
overlap. The line stretcher is adjusted so that 8 is near zero and, as a result,
the output of the sample and hold is proportional to the phase difference, e,
as long as e is small.
9
l ATC-74(8) L
Sample from Line Stretcher
!-!agnetron r-------, 30 MHz
DC volt meter ~v
RMS
S/H A/D 3 Pulse
canceller D/A o
Voltmeter
Trigger Clocking
trigger
10
TA!C-74(9a)L
~~ I-~-I- -_£-
; - . - - -: -I - : : •
1A I 1 liB
I
Phase
1___.
I ' K A B sin Q
Sampling Time
11
The relationship between the magnetron ,and e can be determined
jitte~ f
rms
by tracing the signal through the various components (Figure 8). The FM jitter
at RF is converted down to the same amount of jitter at the 30-MHz IF. The phase
at the phase detector is therefore,
e 21T f
IF
T
where f
IF
is the FM'd IF frequency;and T is the cable delay time. A change in e
is given by
or
erms 21T T f
rms
f
rms
r~E {(f
mag
- E {f
mag
})2t/2
1
where E{ } is the expectation (averaging) operator. Notice that f
is not really
rms
the rms value of the frequency, but the rms value of the difference from the mean.
However, this definition is common and will be used here.
Rn En - 2 En- 1 + En- 2
where E is the nth voltage sample. Thus, the rms voltmeter measures this
n
quantity,
12
Because E and E are not correlated.
n m
-2
E {E
n
E}
m
E E E
n m
R
rms [E {E
n
2
- 2 En E
n-l + n n-2
E E
2
-2 E E _ + 4 E - 2 E E
n n l n-l n-l n-2
1/2
2 }]
+ n E E
n-2
- 2 E
n-l n-2 + n-2
E E
2 -2 ] 1/2
[6 E {E n } - 6 E
or
R 16 E
rms rms
where A
IF is the IF wavelength and C is the velocity of light. Remembering
that
erms
f
rms 2n T
and
R
erms rms lie
6V
16
we can condense all of the above to
13
f
rms
3. Coaxial Magnetron
4. Conventional Magnetron
Measurements of the Amperex DX276 have been made at 2.7, 2.8 and 2.9
GHz. No statistically significant differences were noted. Unlike the coaxial
tube, the DX276 is not heater power sensitive. The standard deviation of fre-
quency was measured several times at each frequency and found to be
f = 2068 -+ 660 Hz
14
l ATC-74(9b)L
12 ~l~
-+
r\! J . +jt'1 :tt-8Jlmmlfft-HFrmmmBIEE1t±lUBfEfEmmma
. ++++h-H~'+ ~rb
r'vP-l'+'l'¥'I-P+f'+'lflh mag 11
tt:~::tt:tt:tt:tt~ttl:tJ~ztqJtztliTr-rTE,B=t+E'rn~=El-J-tij:- tEEEEtai=El1nnaftEE=E=EElE
10 1-+-I-+-l--H+++++++i'IlH--+--f-JH-i--H++++++H--1-+-H-i--H++++++++1-+-H-I--H+++++++-r-H--H-1-H-++++++++H--H
+
f"l
td-, '
tit-
t!· ,-1-+
8 oj. -;.-~ I
15
The frequency drift was measured by observing the spectrum drift across the
screen of a spectrum analyzer. Because the anode portion of the tube was in-
accessible, the temperature of the coaxial tube was measured by recording the
temperature of air coolant exhausting from the tube. The temperature of the
5586 was measured using thermo-couples directly attached to the tube anode.
1. Coaxial Magnetron
2. Conventional Magnetron
Figure 11 illustrates how the DX276 drifts during warmup. Note that
half of the total frequency excursion of 7 MHz occurs in 8 minutes. In contrast,
the coaxial tube takes about 18 minutes to drift to half its total excursion.
The frequency drift rate is greatest at turnon. The anode temperature reads
0
90 C, half of which occurs in the first four minutes.
C. Spurious Responses
1. Measurement Technique
16
l ATC
,, .
J,..cl'
• • ~ f ;-"
+,' +
t:::r
j-t
"i
.+~
-+ r-t:.
++,
I
1500
-20 -10 0 10 20 [) 40 50 60
Relative Time (minutes)
Fig. 10. Frequency drift and exhaust temperature rise of coaxial mag
tron (QKH-1739LL); ASR-7 normal operating conditions.
81
a 'J -" ~ Frequency Drift (MHz)
r~r.it
.•...•. IT'Tffi.X'. t T+:T0'.;f
i," t" '".'.'.1
++ t- j .. i. ',' ,
m.'~·.-.r., It!'j - f - i -
Ui~ +~.l J~l"-" ~ TF:.'..
'I I .' • t-. l_ ' , t
t , . ,,-
J .. .. ~!1
1 · , t " at:::
. . , , , t ; .c
~
o il jl -r: .: HI: f'"
§t+t"+IFT
l'iImt'i
+ ' '·1·
It'.. '.
t"+. -H ..t- . . I.:', .' "';'
S ""i .~".
';' I rr ' i i t 1'1 1 J . l j . +, ,.i- -i" T '
III r"'
OQOQ 1
;::l 0 :_.i ':
(0
+ . r'1;1 ~ _
Et ~ lS+ijJ!.+t+U,:L ~1=ci~~ _:~
i.ttf + j<,. r! . l.w.l... -t HI'
rt ....
~ .... :=; i++h It. +-ijrtH T,' .
tIt1,,1; [~trl'"+
o 0 tt-t
;::l 'f.. . ;~' 1\' [._ ,1": : Li i-, j'! "
~. .. . ' i ", ~-H-~'
........ ""i
t:I~ it\. +-+ f-H ~Jl+-,,':-; ,~ :
:><(1) t1 t :' : . : .'.. -+ - :( r 1 l :
1..0
N~
-....1(1) " HI! '!! ,"
O\::l ttJ:H .. +.'.ffi' H. .H'.. . ' ... ' T .. , 11, ", jT !.. j tit', it' j t-';',1 'I
+"1
'-/ n ,~-l
~W . iIII Itfl ' ~ niH
H
wo 0...:::
: !Ii11"'\", III I I " . ' ! ':;-j
~p..
tn~ w
:;d r"'
lH'l
~ tff'- -iiJ{ .iiHji.Jlt-ql,+ ._.~.11-- _t~ -
'" 0
if ~t H' i-fl Tn Ii
;-+-t;i t· L'i ' tj -- H '
-....Irt
;::l III !ll). i 'i · 'r ,: t 'rtf: :
o ::l c ' : l;jj
~' 1t+.=;
Hj-. .f-ttll
fll±... t[It1r1,
1 1.1 Ttt '"
~ p.. >-J ". c. '1 f - : I
j
1.TI'. t t..'. . : ~ . :.' rt-' :f
, '
III (1) ro -" ; If! j . j T' , f-,· ,
~' 11 +- t .', . , 1-.'·' i~ _ _-
f-lX ~ 0 ++ :T Ii. '1 1 " ,-j.. 1.::- '
::r
o III :J:
:; IIf-l\#t IT I - _l , ;
, I ', tI' I
'1j ~
H H~
(0 iCIl
~ rt III I, 'li1 I i" I ! 1+
III .8.\+ . 1 Ii. t j '.; ;.. J
rtrt
r"- ro
~~ it! st.
::l S ~ Ll'., tttl - - , 'j 1 ,
OQ'1j Lt
IL -+--1.
1 ,-' j ;f '4, , ,
, ; . ': 1 1+ j
m j r r 1
(0
m' ' ~'
(j ~ i rttfll: ,I l ;:.:: 1- ,1,
III
o _~1f=£- Ij-f d1 If t
;::l rt d- J +H, r ' t 1= ' ' :
p..~ ,I rn j t:+. t t ' I ! i-I L
r"- ~
rt ro g ,I
r"-
o ~ -H '
tiJijj~! IJ It Ii rnr ~IItT I+f ' '.
l I; !inl
j II t-1t
;::l r"'
rn rn IT i i 't... 1.' '. I
I , ~t·. f
o ro !J C -l ' -
ill ~I L11' r- ..f I·1 t
o !"e lilt -'
H'l ~ j~ +1
11'1.•. ..,+- _: - '-Itt .
('l
o
Ii 'j. _ I · t I I .fi-h I. , : ' I 1'. ~ if [1
~
ro -f-. t. '. -t Ht- I l!t t~ : !
::l
rt 1 . I
~ I-HI. : iL - -j =r .,~ t t= -tt~tt ~~ If I,I.
! I :i ' ,,' I,
I~
r"'
o
;::l , r: tr-t-r- -+-i' H H·l r +j .. " -=f Ii' , '"
III r
f-l a~JLil l1t h t
~~r ~-, nITf 1.-l-l.. I.+.Jl. i~t1 ' ' .1;.; '.;',
transistion, then through a 3-ft section of WRl12, WR187 or WR90 waveguide,
then back to coax and, thence to the spectrum analyzer.
+
c = 25.6 10glO f - 11.5 - 5
where c is the additional sensitivity (in dB) one has as a function of frequency
f (in GHz). Thus, at 18 GHz the spectra are 20.6 dB more sensitive, making
the 60 dB below the main lobe level really 80.6 dB below the main lobe.
The table given below indexes sets of figures showing photographs of the
spurious responses obtained on each magnetron at low, mid and upper test fre-
quencies. The caption block on the face of each spectrum photograph identifies
the frequency at the center of the analyzer screen (fJ, the relative gain setting
(G 1) and the level of harmonic energy if present. Conditions under which each
re
spurious response spectrum was photographed are identical to those for the pre-
viously numbered spectrum figure except for the conditions noted in the individual
figure captions.
19
AT
Frequency - GHz
N
:.1' ,
o -50
I , '1T ". T~ "
+rt i
I ,
.
~Lu4"':::::' -'-'-t'+~ ,'-.:
-60
'c. ~r-
," cw··
-- .. ~ ....•. "':j
4 5 6 8 9 10 20 30 60
f (GHz)
-~TC-74(
N
.....
N
N
•
.,
lATC-74 (
N
Vol
to
25
~TC-74(30-
N
00
30
•
lATC-74(
W
N
35
--~TC-74(64
•
.'
-~TC-74(68-
Fig. 68.
f 5.5 GHz
C
G 1 37.5 dB
re
2nd harm
-57.5 dB
Fig. 73.
f 10.5 GHz
G
c 44.7 dB
re1
i-IR90 filter;
4th harm
-70.7 dB
w
00
40
•
~TC-74(81
Fig. 88.
f 8.5 GHz
c
G 1 42.3 dB
re
3rd harm
-68.3 dB
-~TC-74((8
45
~TC-74(98-1
•
--~TC-74
lATC-74
50
Figures Numbered
Magnetron 2.7 GHz 2.8 GHz 2.9 GHz
Coaxial (QKH-1739LL) 13 thru 29 30 thru 46 47 thru 63
Conventional (DX276) 64 thru 80 81 thru 97 98 thru 114
Figure 115 shows the RF envelope of the QK1739LL coaxial magnetron as viewed
on a sampling oscilloscope under the following conditions.
(1) The standard diode network controlling the rate of rise was used
(as described in Section II).
(2) A 2-watt RF pulsed signal was injected into the magnetron during
the turn-on time of the modulator pulse to the magnetron. The RF circuit
is shown in Figure 116.
In Figure 117 the front edge of the RF envelope is expanded to show the
starting jitter with the priming signal. The scope is triggered from the front
edge of the applied modulator voltage pulse. The jitter is estimated to have a
standard deviation of 3 nsecs. Without pJiming (Figure 118), the jitter deviation
increases to about 10 nsecs.
51
I ATC-74(115) L
l ATC-74(116~ L
Circulator Load
2 IJatt
Trigger - ....-.; s- Band
ulse Gen
52
-I ATG-74(117) I
53
Let V (t) represent the magnetron amplitude waveform of the nth pulse. V (t)
n n
is a random variable. V(t) is defined as the mean of V (t). The normalized
n
clutter return from a point target is just
(4)
where the integrals are taken over the pulse and E{ } is the expectation opera-
tor. Thus, the noise-to-c1utter ratio is
fo v 2(t). dt
(5)
f V2 (t) dt
where
2 -2
o 2 is by definition the variance of V. Since V (t) ~ V (t), the denomina-
v
tor is simply the peak power times the pulse length. The numerator is more in-
vo1ved. 0 2(t) is greatest at the beginning of the pulse (Figure 119) and
v
2
tapers to zero at t
o
= 125 nsec. Unfortunately, ov (t) cannot be estimated
directly from Figure 119. This is because these pictures are of the output of
a square-law detector and represent power.
Looking at Figure 117 and 119, one can represent the time jitter (standard
deviation) as 1i~ear1y decreasing
t
o (t) =
t
0
ot
(1-
t
) o < t < t
o
(6)
o
t < t
o
where 0 is the initial time jitter and t is the time when the jitter vanishes.
ot 0
-
p = V-2
1/2
then lip 2 V II V 2 p II V
and
54
Fig. 119. Front edge of RF envelope - no
priming (V-axis: linear power - not cali-
brated; H-axis: 50 nsec/div) .
••
55
1/2 1/2
a (t) 2 P a (t) a «p (7)
p v v
p = m t o< t < t
o
(8)
a (t)
p
= m at (t) o< t < t
o
(9)
2 m 2 t )2
a
v 4t
a
ot
(1 -
t
o< t < t
0
0
<Ptiming jitter
-36 dB (no priming)
This value can be compared to the value of <P caused by the coaxial magnetron
frequency jitter. The value of <P obtained when the heater is optimally adjusted
is
56
-52 dB (f - 2 kc)
<P freq • jitter rms
, to the high voltage pulse to the magnetron. The resultant peak-to-peak fre-
quency current deviation constitutes the pushing figure. The frequency shifts
were observed both on a spectrum analyzer and on a phase detector as in the
frequency jitter tester. The current shifts were monitored on a scope by the
use of a current transformer in the anode line to the magnetron. The pushing
figures measured were:
Because priming the coaxial magnetron greatly reduces the timing jitter, it
was thought that the priming signal could be used to phase lock the magnetron.
In essence the magnetron would become an "amplifier" and one would have a co-
herent transmitter. Measurements of the rms phase error of the magnetron rela-
tive to the priming signal were made at four different priming powers (up to
6 kW). A plot of the rms phase error erms (standard deviation) is shown in
Figure 120. At 6 kW the erms is 2.7 degrees. This corresponds to
erms 2
10 log ( 360 X 2n) -26.5 dB
57
,
0
0
0
~
I-l
Q)
~
0
0-
ClO
t::
.~
Ul
''""'' .~
'"
~ I-l
0-
CIl
::-
,...
CIl
0
0
~
e,
'oJ
I-l
0
I-l
I-l
Q)
<1l
CIl
III
..c::
p.,
0
N
..-l
0 ClO
~
.~
f%.<
o suu
o (sacl.Illap) 6
~
58
It appears that it would require an excessive amount of priming (in the
order of the magnetron output power itself) in order to develop a reasonable
~ of -40 dB.
G. OTP Compliance
IV. CONCLUSIONS
Coaxial Standard
Pulling +
Pushing +
Frequency Jitter +
Time Jitter +
Spurious Spectrum +
Long-term Stability +
Lifetime +
Break-in Time +
Cost +
*OTP Manual of Regulations and Procedures for Radio Frequency Management, Chapter 5,
Section 5.3 (January 1973).
59
lATC-74(l21)L
60
Pulling; can be alleviated by the use of an isolator or circulator. The
ASR-7 uses a circulator; therefore, it is not a problem for either tube.
Time Jitter; is intrinsic and the most serious problem of coaxial tubes.
It can be partially corrected by priming and rise time control. The ASR's already
have a circulator that can be used to inject the priming signal.
The coaxial tube Costs about a factor of 10 more than the standard tube.
The cost may be reduced by mass production and/or a competitive market.
61
APPENDIX
INTRODUCTION
THEORY
1
C(t) C (t -
(prf) ) (1)
u(t) := e i8(t)
.,.,. ,.01
The squiggle above 8(t) is to remind one that 8(t) is a random variable (rv). The
signal returning to the radar is then
62
-.
S/(t) = c'<t)@ u(t) eie(t) (2)
'''1
~
where
-
represents the convolution operator.
interpulse period, while Set) can fluctuate between periods.
1\0''''
C(t) fluctuates only during an
The average value
of Set) at the time t is then
",--"
Set) = C(t) ~ u(t) (3)
~'I
Notice that Set) is still a random variable but only during the interpulse
period. Most radar processors (MTD, MTI cancellers) remove Set), consequently
".,.. -
~~ 1'\
the residue Set) - Set) is of prime importance. The variance of Set) is the
average power remaining after the mean has been removed. The variance is given
by
(4)
where E{ } is the expectation operator (or average) and the asterisk represents
the complex conjugate. Thus, the quantity
2
cr
-~ {~2}
2
cr = E {['e'(t) ® u(t) i8 (t) ] [2*(t) ® u(t) (-i) 8(t)] } (6)
63
T +T T +T
0 0
a
2
= E
f ~~
C(t- T)
~
e(t) dt
f """*
C
.M
(t" - T) 8(t") dt" (7)
T T
0 0
or
T +T
JO ~,11 "",,* ,w\
E C(t" - T) C (t - T) 8(t) 8(t") dt dt" (8)
T
o
. . '"
Since C(t - T) C (t -T) and
~*" ~ 4'\A"
8(t) 8(t ) are independent random variables, we
can apply the expectation operator separately,
TfO+T
a
2
T
R (t - t" ) E
c
te(t) 8'( t ") } dt dt" (9)
o
where
R
c
(t - t") = E {~
C(t - T) 4"*"
C (t - T) }
(10)
C
0
~
f
_00
R (t)
c
dt
64
.-
The function G(t) is not known. however it cannot be a constant otherwise no
FM could take place. The simplest time dependence would be of the linear form
...-<1
;vIA
G(t) = wt (11)
Any higher order dependence cannot be measured with our present equipment and it
is not clear that it exists. L0 is a random variable representing the random
frequency error of the magnetron. Taking (11) as our model we find
T +T
o
o
2
f T
o
t
2
dt
co
E {~2} (12)
3
T
J o
R
c
(t-t~) dt dt~ C T
o
(13)
(T + T)3 - T
E l~2} 0 3T 0 (14)
(15)
This corresponds to locking the coho to the center of the magnetron pulse. If
the coho is locked to the tail-end of the magnetron pulse. T = -T. then:
o
(16)
65
EXAMPLE OF THE RAYTHEON QK1739 COAXIAL MAGNETRON IN THE ASR-7
4
If we assume a residue-to-c1utter ratio ¢ of 0.16 x 10- and a pulse length
of 0.7 microseconds, then for a center-locking system
1 1
f
rms ~ 2TI
2
3150 Hz
TIT
where T is the delay line delay of 0.35 microseconds and ~ is the gain of the
d
three-pulse canceller to a white random signal. Thus, the output (in degrees) is:
If we were to allow the residue ¢ to be equal to the thermal noise N, the im-
provement ratio'wou1d deteriorate by 3 dB. ¢ would be equal to 10- 4 and the
output of the stability tester would be
2
erms TIT
(18)
This happens to be equal to the value measured for the coaxial magnetron at
2700 MHz at its nominal heater power. At higher heater power (which may reduce
tube life) erms can drop to 0.5 degrees.
66
CONCLUSION
67
ACKNOWLEDGEMENT
68