A 96 CHANNEL RECEIVER FOR THE ILCTA LLRF SYSTEM AT
FERMILAB
*
Uros Mavric
#
, Brian Chase, Julien Branlard, Ed Cullerton, Dan Klepec
Fermi National Accelerator Laboratory, Batavia, IL, USA.
Abstract
The present configuration of an ILC main LINAC RF
station has 26 nine cell cavities driven from one klystron.
With the addition of waveguide power coupler monitors,
96 RF signals will be down-converted and processed. A
down-converter chassis is being developed that contains
12 eight channel analog modules and a single up-
converter module. This chassis will first be deployed for
testing a cryomodule composed of eight cavities located
at New Muon Laboratory (NML) - Fermilab. Critical
parts of the design for LLRF applications are identified
and a detailed description of the circuit with various
characteristic measurements is presented. The board is
composed of an input band-pass filter centered at 1.3GHz,
followed by a mixer, which down-converts the cavity
probe signal to a proposed 13 MHz intermediate
frequency. Cables with 8 channels per connector and
good isolation between channels are being used to
interconnect each down-converter module with a digital
board. As mixers, amplifiers and power splitters are the
most sensitive parts for noise, nonlinearities and cross-
talk issues, special attention is given to these parts in the
design of the LO port multiplication and distribution.
INTRODUCTION
Identifying the main error contributors in the LLRF
electronics is important for error minimization. The final
performance of the receiver is defined by many
phenomena, which makes the error identification problem
more complex. Usually a combination of computer
simulations and prototype on-the-bench testing usually
helps to understand the major pitfalls. A Matlab based
model has been developed to study the impact of the
receiver parameters on the control of multiple cavities.
Functional Block Diagram of the Receiver
The 96 channels receiver will be organized in 8
channels per board, which requires 12 boards. The size of
the board and consequently the number of channels per
board was chosen to facilitate the mechanical design and
mounting of multiple boards in the rack. The space
between two adjacent RF inputs will be 1.5 inches.
Besides eight RF 1.3GHz input channels, the receiver
board will have 16 down converted IF (13MHz) output
ports, 1 input LO (1.313GHz) port and various others I/O
ports as shown in Figure 1. One receiver board will be
equipped with temperature sensors and various testing
points in order to facilitate the debugging of the bard or to
inject signals in the operational conditions (for instance
dithering). We also plan to mount fault cable detectors
and LO power monitors.
ANALYSIS OF MAIN PARAMETERS IN
THE RECEIVER
Broad-band Noise Figure
Broad-band noise is assumed to be dominated by
amplitude noise. The overall noise figure of one channel
Figure 1: Block diagram of input and output ports of the
receiver.
is calculated according to the noise figure values given in
datasheets. Figure 2 shows the schematic of one channel
and absolute noise power levels at some test points along
the processing chain.
Figure 2: Schematic of one channel. Noise power levels,
signal power levels and nonlinearities are shown.
___________________________________________
*Work supported by Fermi Research Alliance LLC. Under DE-AC02-
07CH11359 with the U.S. DOE
#
mavric@fnal.gov
FERMILAB-CONF-07-299-AD
According to these calculations we expect to achieve
app. -152dBm/Hz of noise power at the output, which is
equivalent to 36dB of input noise figure. The output
power level is app. +5dBm, to match the Multi-channel
Field Control module (MFC) [1] input power
requirements. The relative noise floor is consequently -
157dBc/Hz. This is much lower than a -148dBc/Hz noise
floor of the 12bit ADC on the MFC board and is about the
same as a 14bit part. It is supposed that the noise figure of
the mixer is app. 1dB higher than its insertion loss (see
literature [2]) and that we managed to suppress all the
mixing products by proper matching. Broad-band
amplitude noise introduced by the LO is supposed not to
affect the overall noise floor since the LO amplifiers are
driven in saturation and the LO signal is filtered. The
main limiting component for the schematic in Figure 2, in
terms of broad-band noise floor, is the IF amplifier. If
lower-broad band noise specifications are to be met, a
higher input signal and a more linear mixer can be used.
In this case the IF amplifier can have lower gain.
However, higher mixer linearity is usually associated with
higher LO port power, which means that the LO amplifier
has to be higher gain. This contributes to a higher close in
phase noise as presented in the following paragraph. The
design presented in Figure 2 is a compromise between
linearity, broad-band noise, close in phase noise and LO
input power.
Table 1: Equivalent time rms jitter and phase rms
deviation for some mixers presented in Figure 3.
(Integration bandwidth 100Hz to 500kHz)
Mixer
t
jrms
[fs] φ
jrms
[°]
HMC483MS8G (Active)
2.4 1.1e-3
SYM25DLHW (L10)
1.6 7.8e-4
ZFM-2000 (L7)
1.8 8.6e-4
SYM25DHW(L17)
1.4 6.7e-4
Narrow-band Noise Figure
Narrow-band noise figure of the receiver is assumed to
be defined only by phase noise. Phase noise introduced in
the LO distribution of the receiver is translated to the IF
port and it is defined by the slope of the mixer (V/rad)
(for more details see [3]). The major phase noise
contributor in the LO distribution part of the receiver is
the LO amplifier. The phase noise issues of a microwave
amplifier are thoroughly studied in literature [4]. Input
power to the LO amplifier and the nonlinearity of the
amplifier define the amount of the induced flicker noise.
The flicker noise is up-converted to the carrier, which is
transmitted through mixing on the IF port and therefore to
the whole system. Figure 3 shows an example of the
amount of phase noise generated by an MMIC amplifier
(HMC481MP86, www.hittie.com). The measurement
method used for measurements presented in Figure 3 is
known as single oscillator phase noise measurement
technique according to literature [5]. The results in
Figure 3 show that in the case of these particular mixers it
is better to use a passive mixer with an external LO
amplifier. It also shows that we do not need to drive a
higher level mixer to get the best phase noise
performances. However, lower LO power decreases
linearity, which is investigated in the following section.
The integrated phase noise for various mixers is given in
Table 1. The measurement method used for
measurements presented in Figure 3 is known as single
oscillator phase noise measurement technique according
to literature [5]. The integrated phase noise for various
mixers is given in Table 1.
10
1
10
2
10
3
10
4
10
5
-165
-160
-155
-150
-145
-140
-135
-130
-125
-120
-115
Freque ncy [Hz]
Noise Level [dBc/Hz]
HMC483MS8G (LO=0dBm, RF=+7dBm)
ZFM-2000 (LO=+7dBm, RF=+7dBm)
SYM-25DLHW (LO=+10dBm RF=+7dBm) Amp In Front of the Splitter
SYM-25DLHW (LO=+10dBm RF=+7dBm)
SYM-25DHW (LO=+17dBm RF=+7dBm)
SYM-25DMHW (LO=+13dBm RF=+7dBm)
SYM-2000 (LO=+7dBm RF=+7dBm)
LNA Noise Floor
(-171dBm/Hz)
Amplified AM Noise Floor From the Source
-137dBm/Hz(L7)
-145dBm/Hz (L17)
-148dBm/Hz (L10)
-124dBm/Hz (Active)
-155dBm/Hz
-144dBm/Hz(L13)
Figure 3: Close in phase noise measurements for mixer
and LO amplifier combination.
Nonlinearities
A lower level mixer is usually less linear than a higher
level mixer. Measurements of mixers presented in Figure
3 would show that there is approximately 20dB difference
between the power in the second and the third harmonic
for the level 7 and the level 17 mixers. Measurements
show that level 13 and level 10 mixers exhibit very
similar linearity performance, which favorizes the level
10 mixer. Comparison of a level 10 mixer and level 17
mixer shows that the second and the third harmonic differ
for approximately 13dB between the two mixers. For the
receiver design we have chosen the level 10 mixer due to
lower LO amplifier gain requirements.
A perturbation analysis with the 8 channel receiver and
8 superconducting RF cavities was carried out. Dispersion
in peak gradients between cavities causes the calibrated
channels in the receiver to work at different points of the
nonlinear curve. Analysis shows that the receiver should
operate in the compression point region between 0.01dB
and 0.1dB if less than 0.1% of gradient deviation wants to
be achieved.
Cross-Talk between RF Channels
In order to minimize cross-talk between channels
special attention has to be paid to cables, layout, chassis
design and matching of components. The IF cables
connecting the receiver chassis and the MFC digital board
are manufactured by Harting and measurements show that
isolation between adjacent channels is better than 80dB.
One of the most critical parts in terms of matching and
isolation is the mixer. There are various paths through the
mixer that cause the RF or IF signal to couple in the
adjacent channel. Measurements show that there is a
difference of more than 10dB between a poorly matched
mixer and a mixer with an improved match. Besides
matching and cables, proper layout and chassis
construction (shielding) is also an important way of
coupling reduction.
Temperature Stability of the Receiver
A receiver has been built on an aluminium plate with
heaters and temperature sensors regulated by a
temperature controller. Temperature excursions of 2°C on
the aluminium plate were induced and a phase fluctuation
that equals to 0.25° was measured. Using a more
sophisticated temperature controller (PID) would decrease
large temperature excursions and consequently decrease
phase deviation swings.
IMPLEMENTATION OF THE 8
CHANNEL RECEIVER
One Channel Layout
The specific parts that are planned to be used for the 8
channel receiver board are shown in Figure 2. The final
choice of IF amplifier has not yet benn made. Analog
Devices and other manufacturers offer a broad spectrum
of ultra low noise variable gain amplifiers. However, they
are designed to drive high impedance loads (> 500Ω),
which causes additional power losses in impedance
transformers. At higher gains the nonlinearity becomes a
serious issue. Monolithic microwave integrated circuit
amplifier modules are easier to use but tend to be more
nonlinear and are usually meant to be used at higher
frequencies.
Eight Channel Layout
Figure 4 shows a design of the bottom and top layer of
the board. Each board will be 8U (14 inches) high. The
input down-converter modules will be enclosed in a
aluminium cage. Isolation between LO branches will be
reinforced by ground fences and proper line routing.
Figure 4: A top and bottom view of the eight channel
receiver board.
Each of 12 8-channel boards will be placed on the top
of the rack in a crate. The crate will be slightly tilted so
that the rigid cables will not need to be additionally
banded before coming to the receiver board.
MEASUREMENT RESULTS OF THE
PROTOTYPE
According to the schematic in Figure 1, a two channel
prototype was constructed. Before acquisition, the two
13MHz signals were phase shifted in order to be in
quadrature. The acquired digital data was multiplied and
filtered with a 500 kHz low-pass FIR filter. This gives the
residual phase noise of the receiver and MFC board. Any
common phase deviation of the source generator is
subtracted by the method. Standard deviation of the
filtered signal is 3e-3° at 500 kHz SSB bandwidth.
On the IF port of the prototype we measured the second
and the third harmonics produced by the mixer, which
were 40dB high and the isolation between these two
channels was approximately 88dB with proper matching
on the mixer ports.
CONCLUSIONS
In the paper we discussed the main parameters that
have to be considered for proper operation of the receiver.
A target value for broad-band noise at the output of the
receiver is -157dBc/Hz, the integrated close in phase
noise over at 500 kHz bandwidth should be better than
3e-3°, the receiver should operate between 0.01dB and
0.1dB compression point and cross-talk measurements
show that we can achieve better than 80dB of isolation. A
per se part of the project is the implementation where
mechanics of the chassis, ease of module access and
diagnostics have to be considered. At present we are in
the phase of 8 channel board layout and two channel
prototype testing.
REFERENCES
[1] Varghese P., Barnes B., Branlard J., Chase B., W
Joireman P., Klepec D., Mavric U., Tupikov V.,
“Multichannel Vector Field Control Module for
LLRF Control of Superconducting Cavities”, This
Conference – WEPMN112
[2] A. Mass S. , “Microwave Mixers – Second Edition”,
Artech House, Inc. 1993
[3] Walls F.L. and Stein S.R., “Accurate Measurements
of Spectral Density of Phase Noise in Devices”,
National Bureau of Standards
[4] Hati A., Howe D.A., Walls L.F., Walker
D.K.,“Merits of PM Noise Measurements over Noise
Figure: A Study at Microwave Frequencies”, IEEE
Transactions on Ultrasonics, Ferroelectrics and
Frequency Control, vol.53, No.10, October 2006
[5] “Infrared and Millimeter Waves”, Vol.11, pp.239-
289, 1984