10-1. LLRF system From the view points of rf system, S1-Global can be divided into three periods. In Stage 1, the performance of each cavity, such as the quench limit and Lorentz force detuning, was examined. To examine the cavity efficiently, two RF sources were used. Each 5-MW klystron (1.3 GHz, 5 Hz, 1.6 msec.) drove four cavities. Figure 1-1 shows the schematic of the rf system. A conventional digital low-level RF (LLRF) system was adopted in this Stage 1 and next Stage2. An FPGA board on a commercial DSP board (Barcelona) was used to control the rf output from each klystron. The FPGA board has 10 16-bit ADCs and 2 14-bit DACs with an FPGA and istalled to cPCI [1] as shown in Fig.

spiritualblurtedAI and Robotics

Nov 24, 2013 (3 years and 9 months ago)

118 views

1
0
-
1.

LLRF system

From the view points of rf system, S1
-
Global

can be divided into three periods. In Stage 1, the
performance of each cavity, such as the quench limit and Lorentz force detuning, was examined. To
examine the cavity efficiently, two RF sources were used. Each 5
-
MW klystr
on (1.3 GHz, 5 Hz, 1.6
msec.) dr
o
ve four cavities.
Figure 1
-
1

shows the schematic of the rf system.

A conventional digital
low
-
level RF (LLRF) system was adopted

in this Stage 1 and next Stage2
. An FPGA board on a
commercial DSP board (Barcelona) was used to control the rf output from ea
ch klystron. The FPGA
board has 10 16
-
bit ADCs and 2 14
-
bit DACs with an FPGA

and istalled to cPCI

[
1
]

as shown in
Fig.
1
-
2
.

In Stage 2, the vector
-
sum performance was evaluated.
Figure 1
-
3

shows a schematic of the RF
configuration in Stage 2. To maximize
each cavity’s gradient, the ratios of the RF input powers to the
c
avities are optimized using tun
able
hybrids and variable tap
-
offs
[
2
]
.

One cPCI becam
e the FB
controller and operated

a vector
-
sum control.

The in
-
phase (I
-
component) and quadrature phase
(Q
-
component) of cavity pick
-
up signals are summed
(vector
-
sum)
and compared with the set
-
points.
The errors are multiplied by the
numerical gain and finally feed
-
forward tables are added. The data
acquisition

during rf

operation is every 1us although the feedback is carried out every 40 MHz
clocks.

Total llrf system in Stage
-
2 is summarized in
Fig. 1
-
4
.

The rf pick
-
up, forward and reflection signals
are divided into four ports. First cable is connected to the convention
al cPCIs via a downconverter.
Second

is connec
ted to the IF
-
mix rf monitor
. Third is conn
ected to the rf power meters (
Gigatronics)
(Fig.1
-
5
)

or 32 ch ADC board
and fourth is used by the cavity
-
group. Some output of the klystron and
reflection to the klyst
ron are connected to the VSWR meter and used as interlo
cks to protect a
klystron
. The rf po
wer
-
meters and fast interlock a
re EPICS IOCs.

Local oscillator (LO; 1310.16 MHz) is generated by clock dividers (1300/32/4) using AD95
10 and IQ
modulator
. Clock sig
nals (CLK; 40.625 MHz) is generated by clock dividers (1300/32). In order to
define the I and Q components, timing clock (timing; 10.156 MHz) is also used.

In IF
-
mix, th
ree IFs are mixed by an RF combiner and the mixture is input to each ADC.
These I
f
s
are

9.02 MHz (IF1, 1300/16/9 MHz), 13.54 MHz (IF2, 1300/32/3 MHz) and 18.06 MHz (IF3, 1300/8/9
MHz), respectively.
The signal is separated into three IQ components (cavity pick
-
up, cavity input
and RF reflection from the cavity) by digital signal processing.
This enables the use of a maximum of
30 RF signals by 10 ADCs.

Remote
-
attenuators
(EPICS IOCs)
are installed after the 4 port rf
dividers and these signals are connect
ed to ninety six downconverters
or 10ch downconverters.

The rf cables are carefully calib
rated since the beam
-
based calibration is not available at S1
-
Global
.
The attenuation of each cable is

measured by network
-
analyser by measuring the S11 with changing
the reflection position (line
-
strecher). This method enables us to cancel the VSWR of the

cable itself.
Figure 1
-
6

shows the typical

measured signals obtained by this procedure. Offset of the circle
corresponds to the VSWR of
the
cable itself and

the cable loss is calculated from

the radius of the
circle.

10
-
2. LLRF system for DRFS (Stage 3)

In Stage 3 of S1
-
Global, a

distributed RF system (DRFS)
[
3
]

was evaluated. Two 800
-
kW klystrons
were connected to one RF modulator; each klystron drove two cavities. The di
gital feedback system
based on
u
TCA
[
4
]

(
Fig.
2
-
1
)

was located near
the cryomodule t
unnel
.
Figure
2
-
2

shows the
photograph of the LLRF racks located in the tunnel
. Not only digital FB system but also LO signals
for uTCAs and downconverters are generated

as shown in
Fig.
2
-
3
.

Figure
2
-
4

shows the schematic of
LLRF system in the tunnel.

The
total operation time of DRFS was about 66 hours (
Fig.
2
-
5
) and total doze at the LLRF racks was
140 mSv

(~2 mSv/h)
. Since the
total ionizing dose effects for semiconductors appears from ~8 Gy

[5]
,
the DRFS condition (without beam) will allow 4,000 hours operation. Radiation level (with beam)
measured at FLASH is 15mGy/h
[5]

and it allows 500 hours operation without shield. In

Quantum
beam


and STF
-
2, some radiation shield will be required to prot
ect the LLRF
equipment

installed in
the beam tunnel.


1
0
-
3
.

Interlock sy
s
tem

Quench introduces a serious cryogenic heat load. If quench occurs, the RF input to the cavity should
be shut down as soon as possible; otherwise, the system cannot operate until r
ecovery of the
cryogenic system. Thus we developed a rapid quench detection and interlock system. The loaded Q
(Ql) of each cavity (~2.4e6) is calculated using the RF decay time at the end of the RF pulse; if the
value is lower than the threshold for Ql (
e.g., 2e6), RF operation stops at the next pulse as shown in
Fig.
3
-
1
. This quench
-
interlock system works well and helps lower the heat load to the cryogeincs.

DRFS operation makes it difficult to detect quench with this procedure because of the cross
-
talk

between cavities. However, we continued to use the Ql value obtained from simple RF decay. Quench
detection sometimes requires an additional pulse, but the cryogenic heat load is reduced even in this
case.
Arc detectors

attached near the cavity couplers a
nd RF power interlocks are also introduced to
protect the couplers and klystrons, respectively, from breakdown.


1
0
-
4
.

Cavity diagnostics

A special feature of the DRFS is that it was operat
ed without circulators
[
3
]
. If the cavities are paired,
the RF reflection can be cancelled when the paired cavities are operated with exactly the same
parameters (dynamic detuning , Ql and so on). If the cancellation is not perfect (for example, during
commissioning of the cavities
), reflection to the klystron occurs, and cross
-
talk between the cavities
subsequently appears. The former may results in the damage of the klystron and the latter makes
cavity diagnostics difficult. Correct on
-
line cavity diagnostics are essential for DRF
S operation.
Therefore, we have developed these diagnostics.

Dynamic detuning caused by the Lorentz force in the cavity affects the RF performance at
operation above 20 MV/m. Piezo
-
actuators are installed in all the cavities, so the cavity tuning can be
co
ntrolled dynamically. Cavity detuning should be maintained below 50 Hz during the RF flat top.
Optimization of the piezo requires a precise detuning monitor (e.g., < 5 Hz). Dynamic detuning is
typically measured by the pulse
-
short
ening method [
6
], in which

the RF pulse length is changed and
the detuning at the end of the pulse is calculated from the phase change. These values appear
reliable but microphonics (fluctuation in the pulse
-
to
-
pulse detuning) is not considered. In addition,
this method is not suit
able for circulator
-
less systems such as the DRFS because the release of stored
energy in one cavity becomes the RF input to the other cavity (and vice versa). Thus, we cannot
define a clear “RF
-
off” status, which is essential for the pulse
-
shortening meth
od. To solve these
problems, we developed a real
-
time detuning monitor based on the cavity differential equation. The
cavity voltage (
V
cav
) and cavity input voltage (
V
for
) satisfy the following cavity equation,

d
dt
V
cav
=

(
ω
1
/
2

j

ω
(
t
)
)
V
cav
+
2
ω
1
/
2
V
for

(1)

where

1/2

(
=

𝑓
0
2
𝑄
𝑙
)

and

(t) are the bandwidth and dynamic detuning of the cavity,
respectively. By using eq. (1), we can obtain Ql and detuning.

Eq.(1) can be modified as

1
2
d
dt
|
V
cav
|
2
=
ω


(

)
(
|
V
for
|
2

|
V
r f
|
2
)

(2)

where
V
ref

is the reflection voltage from the cavity and satisfies
V
cav
=V
for
+V
ref
.

We can calculate the time
-
dependent Ql by using eq.(2) except in the RF flat
-
top region where the
cavity voltage (V
cav
) is constant.

The precise cavity input power is needed for the
detuning calculation. Because the directional RF
co
uplers have a directivity of 20~30 dB, cross
-
talk occurs between V
for

and V
ref
, so it is difficult to
determine the precise cavity input power. To obtain better resolution for detuning and Ql, both V
for

and V
ref

are corrected by using the measured directivities.
Figure
4
-
1

shows the calculated detunings
at Stage 2. Upper
eight

plots correspond to the detunings and lower eight show the normalized V
cav
,
V
for

and V
ref
. Since total 24 ch signals (8 cav.
x

V
f
or
, V
ref
, and V
cav
) are required for the
measurements
,
signals obtained by IF
-
mix are utilized for the real
-
time detuning calculation. These real
-
time
detuning monitor is quite useful to adjusting the piezo compensation.

Figure
4
-
2

shows the results of the

cavity diagnostics (dynamic detuning and Ql values) during
DRFS operation

with unbalanced condition
.

Under the
such conditions (caused by difference in
detuning etc.), cavity input signals exist even after rf off and the
Ql values obtained from a simple
d
ecay constant
does not work. However,
those calculated from eq. (2) (blue line) were constant under
various detuning condition. The cavity diagnostics works well for detuning control, especially for
optimization of piezo control.




References:

[1
] S.Michi
zono

et al., “
Vector
-
sum control of superconducting rf cavities at STF
”, PAC’09, Vancouver, May 2009,
pp.2204
-
2206
.

[2
]
Christopher Nantista and Chris Adolphsen,

“Klystron Cluster Scheme for ILC High Power RF Distribution”
,
PAC’09,
Vancouver, May 2009,
pp.2036
-
2038
.

[3
]

S. Fukuda et al.,
“D
istributed

RF S
cheme

(DRFS)
-

N
ewly

P
roposed

HLRF S
cheme

for

ILC”
, LINAC

10, Tsukuba,
2010, pp.112
-
114.

[4]

T. Miura et al.,
“Performance of the Micro
-
TCA Digital Feedback Board for DRFS Test at
KEK
-
STF”
, These proceed
ings.

[5
]

E. Negodin: Dose measurements and radiation protection measures for electronics in the XFEL
tunnels

http://www.xfel.eu/project/meetings/project_meetings/2009/

[6
] Y. Yamamoto et al.,
“Test Results of the International S1
-
Global Cryomodule”
, SRF2011, Chicago, 2011, THIOA01.





GS1andDRFS_1st.tif

Fig.1
-
1

Schematic of the LLRF

system

at Stage 1



MOPS108
-
katagiri.doc

Fig.1
-
2

Photograph of cPCI digital FB system



GS
1andDRFS_
2nd
.tif

Fig.1
-
3

Schematic of the LLRF system at Stage 2


STF_RFmonitor.ai

Fig.1
-
4 Digital LLRF configuration at Stage 2.



MOPS108
-
katagiri.doc

Fig.
1
-
5

Photograph of power
-
meters used at S1Global.



Fig.1
-
6

Measured S11 for cable
calibration

MOPS108
-
katagiri.doc





MOPS108
-
katagiri.doc

Fig.
2
-
1

uTCA based digital LLRF system at Stage 3.


S1G_Study110208.ppt

Fig.
2
-
2

Photograph of the LLRF rack layout at Stage 3.



DST
-
LMK
8
ch clock distribution
1300
/
32
=
40
.
625
MHz
1300
/
128
=
10
.
156
MHz
1300
/
128
=
10
.
156
MHz
90
deg
.
Delay line
10
MHz LPF
10
MHz LPF
IQ modulator
I component
Q component
1300
MHz
FPGA board
(
clk
)
Down converter
(
10
dBm
)
Amplifier
CW powermeter
Fig.
2
-
3

Clock generation at Stage 3.

DRFS101020D.doc


DRFS101031.doc

Fig.
2
-
4

Schematic of the LLRF system at Stage 3


S1G_Study110125.ppt

Fig.
3
-
1

Quench detector







S1G_Study110125.ppt

Fig.
4
-
1

On
-
line

detuning monitor used at Stage 2.




IPACv4.docx

Fig
4
-
2

RF waveforms (Vcav, Vfor

and Vref), calculated detuning and loaded Q at Stage 3.