WE3F-6
A HIGH SPEED FEED FORWARD PSEUDO AUTOMAllC GAIN CONTROL CIRCUIT
FOR AN AMPLIFIER CASCADE
Stewart Walker
Filtronic Components Ltd
ABSTRACT
A technique has been demonstrated that allows pseudo
automatic gain control to be performed on an RF
amplifier cascade in approximately 20ns . This gives a
limited RF output which has negligible harmonic content.
The limited RF output phase is relatively independent of
input power.
INTRODUCTION
Many EW systems use a combination of directional
antenna and interferometry techniques for measuring
the direction of arrival of an RF pulse. Typically the
amplitude difference between the signals detected by
several directional antennae yields coarse direction of
arrival information. While comparison of the phase of
the signals, by an interferometer yields fine direction
of arrival information. It can be seen that the RF
amplifier cascades that are used for amplifying the
incoming signals to an adequate level for the
interferomer must exhibit good AMlPM conversion
over the input dynamic range. This is to minimise
phase errors due the difference in amplitude of the
signals caused by the directional antenna. Typically
the phase shift of an of S21 for an amplifier does not
vary with input power, providing the input power is low
enough that the amplifier is operating linearly. As the
input is increased to the point at which an amplifier
stage stops to operate linearly, then the phase of S21
starts to change with increasing power level.
It would therefore be desirable to keep the individual
RF amplifier stages from saturating. In principle this
could be achieved using an amplifier cascade with
automatic gain control (A.G.C.). however in order to
operate effectively on incoming radar pulses it would
have to operate within approximately 20ns. It is
unlikely that a conventional closed loop feedback
A.G.C. system could be made in which the delay
between an RF pulse arriving and the pulse starting to
94 1
CH3389-4/94/~-0941$01.(XIO1994 lEEE
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on February 27, 2009 at 08:08 from IEEE Xplore. Restrictions apply.
W. Yorkshire, England, UK
become attenuated would be much under lOns, if
delays associated with the detector, feedback circuit,
attenuator, and RF gain cascade are taken into account.
Therefore intuitively it appears to be impossible to make
a stable conventionalA.G.C. system which would attack
and give a constant amplitude output within
approximately 20ns. For this reason an open loop
solution to the problem is presented in this paper. The
presented in this paper. The construction and results
from a 2.5Ghz to 5.3Ghz prototype circuit that was
constructed in order to investigate the circuits
performance, are also presented.
PRINCIPLES OF OPERATION OF THE FEED
FORWARD PSEUDO A.G.C. CIRCUIT
A block diagram of the circuit is shown in Fig 1. The n*
comparitor reference level is set such that each
comparitor changes state when the power at the
detector reaches a level which is slightly lower than the
level at which the n+lmamplifer starts to saturate. The
comparitor output is fed forward to switch in all
subsequent attenuators down the cascade, which
exactly cancel the gain of each stage. As the
feedfoward signal is applied to each OR gate
simultaneously, the time taken for the AGC to cut in is
governed by the propagation delay of one comparitor
and one OR gate, and the time it takes for the
attenuators to change state. The outputs of the
comparitors further down the cascade, which may
corrupted by switching transients from prior attenuators
do not effect the attenuation state, during an RF pulse.
The attenuation for the whole cascade is controlled by
one comparitor transition. The detector that drives each
comparitor is separated from the first attenuator that it
controls by an amplifier. This amplifier's reverse isolation
stops the detector being effected by switching transients
from the first attenuator, that it controls.
DESIGN OF A PROTOTYPE CIRCUIT
A prototype was constructed to operate across the 2.5
to 5.3 GHz band.. The intention was to keep the
ouput power constant to within a 5dB wide window,
over approximately 30dB wide dynamic range. AM to
1994 IEEE MlT-S Digest
 PM conversion of up to 2"ldB has been seen from
conventional amplifier cascades at similar frequencies.
The intention was to demonstrate that this could be
reduced by an order of magnitude, by using the
feedforward A.G.C. to ensure that the amplifiers never
saturate. The aim was also to show also that the
feedforward A.G.C. circuit could cut in approximately
20ns, and to find out if any unforeseen practical
problems that would effect its operation.
Fig 2 shows a diagram of the a single stage of the
prototype circuit. The low phase shift attenuator was
made by switching the RF through two different paths
with differing attenuation, but equal electrical length. The
attenuation state is switched between 3dB attenuation,
and 14dB attenuation. The phase through this circuit
was found to be more constant than a circuit with OdB
and 1IdB attenuation, when the circuit was terminated
by amplifiers at either end. This is because reflected
signals between the amplifers have less effect on the
phase with the 3dB pad in place in the low attenuation
state.
The gain of this entire stage was switched between
approximately 10dB and OdB. This gain was acheived
using 2 Avantek MGA64135 packed MMlC amplifiers
with nominally 10 dB gain each. As can be seen there
was 10 dB of loss in the through path, caused by a
combination of the Wilkenson power splitter for coupling
off the signal to the detector, the switched attenuator,
and some gain equalisation to improve the frequency
flatness. The detector was protected from detecting
switching transients of the attenuators further down the
cascade, which may have caused false triggering, by
the reverse isolation of one amplifier, and the isolation
of the Wilkenson Power Splitter.
It is not quite as important that a detector could detect
transients from an attenuator closer to the input of the
cascade, when an input pulse is applied. This is
because, by definition, that detector will not be effecting
the operation of any attenuators, as they will all be
FIG I BLOCK D I A G R A M CF
PSEUDO A G C C I R C U I T
rm*r IlliOC , IT"0
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on February 27, 2009 at 08:08 from IEEE Xplore. Restrictions apply.
controlled by a detector closer to the input of the
cascade. However when a pulse finishes, then false
triggering may be caused by this, which effects the
length of time it takes before the A.G.C. is able to
operate properly on a new pulse.
Fig 3 shows how the individual stages were connected
together to make the prototype. This differs slightly from
the circuit shown in Fig 5.1 in two main respects:
1) In order to demonstrate that fine amplitude control
could be incorporated at the end of the cascade a
switched attenuator, with approximately half of the
attenuation range that each stage attenuator had, was
placed at the end of the cascade. As the input power is
ramped up this was altemately switched to tum to its
high attenuation state (9dB) by a second comparitor on
each detector, and then to its low attenuation state
(4.5dB) as the next main attenuator stage was switched
in. The signal that switched this fine attenuator to its
low state was derived from the same comparitor
transition that turned on the main attenuator stage,
hence the attenuators change state at exactly the same
power level.
2) The logic was inverted to allow the OR gates to be
replaced with 3 input NAND gates which were readily
available. It was decided to use AC type logic for the
NAND gates, as it has a large output transition going
from within 0.5V from the negative rail to within 0.5V of
the positive rail. This allowed the gates to drive the
attenuators directly without extra driving circuitry.To
make driving the attenuators straight forward the logic
was run from +/-2.5V supplies to provide a negative low
output and positive high. The comparitors were run from
+2.5V and -7SV to allow the NAND gates to be driven
directly from the comparitors. Hysteresis approximately
equivalent to 0.5dB of RF power was put on each
comparitor circuit to avoid spurious transitions.
HIGH SPEED FEED F O P W A R D
ITKL ~ 1Hli3 Illzr
,
I
 FIG3
BLOCK DIAGRAM OF PROTOTYPE FEED FORWARD PSEUDO AGC CIRCUIT

I1 L

The unit was set up at 4GHZ with comparitor transitions
set so as to give the output power versus input power
characteristics shown in Fig.4.
PERFORMANCE OF THE PROTOTYPE CIRCUIT
A single stage was found to have approximately 1dB
gain ripple, and +/-2" phase shift between attenuation
states. The phase shift is thought to result mainly from
the.mismatach effects between the switching junctions
and amplifiers being suppressed more in the high
attenuation state.
The output power of the unit at 4 GHz is shown in Fig
4 as the input power is increased in real time over a
range of 36 dB. This demonstrates that over a 28 dB
range the output power stays within +/-MI3 of a fixed
output level. It also shows that the transitions appear
clean, without false triggering.
&er the pulse has finished appears more significant due
to the transients coming from attenuators at the
beginning of the cascade and being amplifed by the
following amplification. Because of the length of time
that these transients carry on for after the pulse has
finished it appears that they are large enough to
retrigger the attenuators further down the cascade,
hence causing more transients.
The phase of S21 at an input power of -1dBm refered to
the phase shift when the input power was -4ldBm was
measured. The worst case phase shift over 4OdB
dynamic range was approximately 8'. This results in a
approximately 0.2"ldB average AMlPM conversion. This
compares with figures of 2'/dB that have been seen for
conventional limiting amplifier cascades.
The harmonic output of the unit was also measured for

The pulse performance of the unit is shown in Figs 5

to 7. They show the output from a negative going
detector connected up to the RF output, as the power FIG . 2 C I A G R A M OF R F CIRCUITRY OF SINGLE
STAGE OF PROTOTYPE FEED FORWARD A.GC C I R C U I l
level of an applied 250ns pulse is increased. Fig 5
shows the output of the unit with the input power just
below the power level at which the first transition 2 SECTION

occurs. In this state the unit operates identically to a VIKENSON

POWER DIVIrIER
(3 dB LOSS)
conventional amplifier cascade. It is noticeable that the RF
INPUT
RF pulse appears to take approximately ions to get
close to its ultimate level. This is thought to be caused
by the signal source, and is thought to be the reason
why the unit appeared to be slightly slower than
origanally anticipated. Fig 6 shows the ouput from a an
identical input pulse to that in Fig 5 except with its power
increasedto a level that was 3dB above the power level
where the 3"1transition occurred. The time that it took for
the output to reach its final level after the start of the
rising edge of the pulse was 22.5ns. Fig 7 shows the
same 25017sinput pulse, except with its power level 5 dB
above where the final transition occurred. This shows
the attenuation to "cutin" in approximately the same time
as Fig 6. On this plot the effect of switching transients

943

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on February 27, 2009 at 08:08 from IEEE Xplore. Restrictions apply.
 a 2 GHz fundemental at power levels between
-40dBm and OdBm. It was found that the second and third F1G.L OUTPUT W W E R OF THE PROTOTYPE UNIT AS THE
INPUT POWER IS SWEPT OVER A RRNGE OF 36 dB
harmonic levels were always greater than 5OdB below
the fundamental at the output. This illustrates that an
amplifier cascade using this type of A.G.C. would -3dBM
provide a very clean amplified signal for frequency
measurement, or other uses, over a wide input dynamic
range.
FURTHER WORK

In order to make a unit of this type into a practical unit,

there are several areas which would require further I I l-i I I
work:

I ) In order not to cause spurious outputs after a pulse

ends, the switching transients resulting from the
attenuators being switched to a their low attenuation
state should be preventedfrom being amplifed by the RF
amplifiers. This could be acheived by combination of FIG
~~ 5 _._
.PP
PULSE DETECTED
filtering the RF lines, slowing down this switching pulse, AT OUTPUT OF
PSEUDO A.G.C.
or a different design of attenuator PROTOTYPE
(INPUT POWER
BELOW FIRST
2) The time taken for the unit to "cut in" should be TRANSITION)
reduced to the region of ions over the full dynamic (NEGATIVE GOING
DETECTOR)
range. With the current unit significant degradation in
"cut in" time was observed close to the transitions,
where the comparitors were not being significantly
overdriven.
FIG6 RF 125 mV/div -375 mV 50.0 n s / d i v 0.000 3
PULSE DETECTED
3) In order to keep the RF output within a narrow AT OUTPUT OF
wind? the gain of the amplifier cascade would require PSEUDO
PROTOTYPE
A.G.C. G v.

to be flat with frequency and temperature compensated.

(INPUT POWER
Altematively, one conventional limiting amplifer stage 3dB ABOVE THIRD
TRANSITION)
could be placed at the output. This would, however,
(NEGATIVE GOING
cause additional phase shifl in S21, with input power DETECTOR)
OSV

level. Cursor 0-X

-145 mv
22 5 ns
It is the authors opinion that the feedforward techniques
described in this chapter may lend themselves well to a FIG7 RF
PULSE DETECTED
GaAs monolithic approach, with an amplifer, switched AT OUTPUT OF
attenuator, detector, high speed comparitor and high PSEUDO A . G .C .
PROTOTYPE
speed logic circuitry on a GaAs MMIC. This should (INPUT POWER
011

improve the speed and reduce the size. 5dB ABOVE SIXTH
TRANSITION)

(NEGATIVE GOING
DETECTOR]
CONCLUSIONS
Cursor 0-x
-169 "J
A novel feed forward pseudo A.G.C. circuit has been 26.5 ns

presented. This is intended to reduce the AM to PM

conversion in the RF output of a limiting RF amplifier
cascade. There may also be applications for this
technique in amplifying RF pulses over a wide input
dynamic range to give an RF output that is free from
harmonics. The design and construction of a prototype
circuit are described. The key points of the
performance of the prototype are approximately
0.2"ldB AM to PM conversion, and approximately 20ns
"cut in" time.

944

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL. Downloaded on February 27, 2009 at 08:08 from IEEE Xplore. Restrictions apply.