1. Present Status of EPICS Development at the 3GeV Proton Beam Transport Facility Control System in J-PARC
JAERI
Motoki Oi Tetuya Kai Sinichiro Meigo
Sinichi Sakai Hidetaka Kinoshita Kenji Sakai
Masanori Kaminaga
My talk in about present Status of EPICS Development at the 3GeV Proton Beam Transport Facility Control system in J-PARC

2. Contents Outline of 3NBT monitor system Goal of 3NBT EPICS system
About CC/NET
Data Read & archive performance test
Channel access data missing and Tick frequency
Profile monitor test in KEK
Summary
Contents of my talk is
Outline of 3NBT monitor system, goal of 3NBT EPICS system
About CC/NET, Data Read & archive performace, Channel access data missing and tick frequency
Profile monitor test in KEK
And summary

3. 3GeV Proton Beam Transport Facility(3NBT)
Beam loss : ~1W/m
Length : 310m
Magnet : 108
Proton monitor : 89
3NBT
Material & Life
science Facility
50GeV PS
Experimental Area
(MLF)
J-PARC is consist of three proton accelerater and four experimental facilities. 3GeV proton beam transport facility in short 3NBT is beam transport line from 3GeV synchrotron to Material and Life Science facility. Length of 3NBT is over 300m and there is over 100 magnet and 89 proton monitors. Design criteria of beam loss of 3NBT is under 1W/m. So, proton monitor system and magnet control system is very important.

4. Goal of 3NBT EPICS system
・ Monitor all of the proton beam monitors at 25Hz
・ Archive all of proton beam monitor data with data tag
・ Control and monitor the beam line components
(Magnet power supply, vacuum, etc.)
・Joint 3NBT to CCR and MLF control with EPICS network
Goal of 3NBT EPICS system is monitor all of the proton beam monitors at 25Hz.
Archive all of proton beam monitor data with data tag. This is for proton beam line study and trouble shooting.
Control and monitor the beam line components. Such as Magnet power supply, vacuum and etc.
And joint 3NBT to CCR and MLF control with EPICS network.

5. Outline of 3NBT Monitor system
CCR
MLF
3NBT Control
EPICS OPI
EPICS OPI
EPICS IOC
EPICS OPI
Data Archiver
EPICS IOC
EPICS IOC
LAN
LAN
CAMAC modules
Total 1100 data
DAQ at 25Hz
Data flow
CC/NET
EPICS IOC
Data flow
This figure shows outline of 3NBT monitor system.
We have four type of proton beam monitors. Profile monitor, halo monitor, loss monitor and current monitor.
These signal is lead to ADC on CAMAC module. As a CAMAC controller, we use CC/NET and EPICS IOC is installed on the CC/NET. Monitor data is archived with data archiver and display with EPICS OPI and lead to EPICS IOC. EPICS IOC is also connected to other facilities.
CC/NET: CAMAC controller
ADC
Profile Monitor
Current Monitor
Halo Monitor
Loss Monitor

6. What is CC/NET PC104Plus Single Board Computer
Interface: PCI and CAMAC
CPU:Crusoe500MHz
Memory：310MB
Main Drive: Compact Flush 1GB
Network：100Mbps　Ethernet
OS：2.4 kernel Linux
TRIG-IN
BUSY-OUT
Ethernet
RBG OUT
About CC/NET.CC/NET is network CAMAC controller which consist of PC104 Plus single board computer with PCI interface and CAMAC interface. It also have trigger input and busy-out signal. Of course Ethernet port and RBG port and PS/2 port. As a operating system, Linux kernel 2.4 is installed. So, we can install EPICS base program directly on the CC/NET.
In 3NBT proton beam monitor system, we use many 16ch CAMAC ADC modules. And as a CAMAC controller, Pipe line CAMAC controller CC/NET is selected. CC/NET consists of a PC104Plus-based single bord computer, PCI interface and CAMAC interface. Linux is installed and of course connectable to Ethernet. Signal handling of trigger input and busy out are also implemented.
右の絵に矢印でどんなコネクタがあるかを明記
PS/2
USB

7. DAQ mode of CC/NET CC/NET have three DAQ mode
Single mode Programmed Dynamic Memory
I/O mode Access mode
Small Small Large
Low Middle Fast
High Low Middle
Over head
DAQ Speed
CPU load
Pipeline method used in PIO and DMA mode
CC/NET have three Data Acquisition mode. So call Single mode, PIO mode and DMA mode.
General performance of these three mode is shown this. Over head of Single and PIO mode is small but DMA mode is large. And this figure shows comparison of access time of PIO mode and DAM mode.
If data frame size is small, PIO mode is faster but the data size is large, DMA mode is fast.
Single mode is simply send one CAMAC command and Receave one data. PIO mode send multiple CAMAC command at onece and Resieve Multiple data at once. DMA mode is almost same as PIO mode but Reseave data in order.
CC/NETは3種類のDAQモードを使用できる。SingleモードとPIOモードとDMAモードである。
一般的に、シングルモードとPIOモードのオーバーヘッドは小さく、DMAのは大きい。
DAQスピードは、SingleよりPIO、PIOよりDMAモードが早いとされている。
この図は、PIOモードとDMAモードの性能を比較した物だが、フレーム数が増えると、PIOモードよりDMAモードの方が早い事が分かる。

8. EPICS DAQ test sequence
START
NIM module
Repeat at 25Hz
Clock generator
CAMAC module
Base clock
Gate signal
ADC
Max 23 ADC
Gate generator
Trig signal : delay 200 nsec
CC/NET EPICS IOC
Max 368 data
CAMAC DAQ
This figure shows EPICS DAQ test sequence.
At first, gate signal is input and monitor data is read with ADC.
Then, Trigger signal input to CC/NET and monitor data becomes EPICS record.
These data are archived with Channel Archiver and Channel access to EPICS IOC.
And repeat this cycle at 25 Hz.
CAMACモジュール1台あたりのデータ点数を書いて、次のテストでの意味付けを明確にする。
各種枠内に入れる文字を再検討
付加する文字についても検討する。
CAMACモジュール1台当り最大368点だが、実際に取り扱うデータは、64点とする可能性もある。
その事を明確にした構成に
Data Archive
Channel Access
Channel Archiver
EPICS IOC
Linux WS
EPICS
Channel Archiver 2.1 2

9. Data Taking 16ch ADC x 23 slot = 368 data / CAMAC module
368 data / waveform is not reasonable.
64 data / Profile monitor
4 data / Halo monitor
4 data / Loss monitor
64 data is reasonable
About data taking.
We use 16ch ADC and 23 ADC for one CAMAC create. So we got 368 data par CAMAC module.
But 368 data par waveform is not reasonable.
Because one profile monitor gives 64 data, halo monitor is 4 data and loss monitor is 4 data.
And also, large data is not easy to take with channel archiver because it takes long time to read.
So, 64 or more less data is reasonable.
Then, change NELM par record and examine CC/NET to EPICS DAQ performance test.
Change the NELM / Record and examine CC/NET-EPICS DAQ performance test

10. Data Read & Archive performance test @25Hz
CPU load
DAQ time
This figure shows the result of Data read and archive performance test at 25Hz.
This means Number of Element by Number of Records. CPU load of CAMAC access with EPICS Channel access. DAQ time is only CAMAC access time.
From this result, if we use 368 data par 1 waveform, DMA mode is the best. But 64 element or more less, PIO mode is better than DMA mode.
For small Number of Element, PIO mode is the better

11. Data missing in Channel Access
In the channel access between EPICS IOC to EPICS IOC,
we got data missing.
As a cause of data missing, I expect Linux tick frequency.
Default tick frequency is 100 Hz.
DAQ frequency is 25 Hz.
These are near in order.
CC/NET
EPICS IOC
Data Missing 0%
Data Missing ~1%
Change Linux tick 1000 Hz
Compare Data missing ratio
Next is about data missing in channel access.
In the previous test, data transfer from CC/NET to Channel archiver was perfect. But data transfer of CC/NET EPICS IOC to Linux WS EPICS IOC lost some data.
As a cause of data missing, I expect Linux tick frequency.
Default tick frequency of Linux 2.4 is 100Hz. On the other hand, our DAQ frequency is 25 Hz. These are near in order and it makes this data missing.
Then, Change Linux tick to 1000 Hz and compare data missing ratio.
ここまでの試験において、EPICS IOCから、Channel　Archiverへのデータ転送では、データロスは確認されていない。しかしながら、EPICS IOC間でデータ転送をした場合に、データMissingが生じている
Linux WS
Channel Archiver
Linux WS
EPICS IOC
CCR, MLF
EPICS IOC

12. Tick and DAQ frequency Channel Access Data missing CC/NET to Linux IOC
DAQ Tick (Hz) 100Hz 1000Hz
1Hz 0.1% 0.1%
10Hz 0.2% 0.2%
25Hz 0.8% 0.2%
33Hz 0.7% 0.2%
50Hz 0.6% 0.2%
PIO mode Waveform 368 element
This is result of the difference of tick frequency and DAQ frequency.
DAQ frequency of 1Hz to 50 Hz was examined. For 1Hz and 10Hz there is almost no difference between 100 Hz tick and 1000 Hz tick.but when DAQ frequency becomes larger than 25Hz, 1000 Hz tick frequency shows little bit better performance.
This Channel access result is not perfect and not enough to data archive, but enough for monitoring with I/O display.
CA performance is a little bit improved.
This CA result is not perfect but enough
for monitoring with I/O display

13. Profile Monitor Test in KEK
3NBT line transports 1 MW power proton beam, we have to develop long life profile monitor.
SiC wire type profile monitor is developed.
Profile monitor performance is tested in KEK proton beam dump line.
Next is about profile monitor test in KEK
3NBT line transports 1MW power proton beam, and we have to develop long long profile monitor.
Thenm We develop SiC wire type profile monitor and its performance is tested in KEK proton beam dump line

14. KEK NML Beam Dump Proton Energy : 500 MeV Proton current : 182x1010ppb
from KEK-PS
Proton Energy : 500 MeV
Proton current : 182x1010ppb
Frequency : 0.45 Hz
Beam port : Beam dump line
to Beam dump
Profile monitor
NML Beam dump
to KENS
About KEK NML beam dump line. This figure shows around Proton beam dump line and profile monitor is installed here. And this is profile monitor chamber. Proton energy is 500 MeV proton current is 182 to 10 ppb. Beam frequency is 0.45 Hz. This is due to the limit of this beam dump.
Beam dump

15. Profile monitor test component
Prototype profile monitor
Oscilloscope
Signal check
INV AMP
Profile monitoring
32ch
ADC
CAMAC
CC/NET EPICS IOC
This figure shows prototype profile monitor tested in KEK. There is 16 wires in horizontal and 16 wires in vertical.
At first, signal from this wire is checked with oscilloscope and lead to inverter amp. Then AD converted with ADC.
The data is read by CC/NET and display with MEDM on another Linux WS.
Linux WS
SiC clad tungsten wire f100mm
Data Archiver
MEDM
H:16 V:16 wires

16. Results
Raw signal
Successfully get proton beam profile with SiC wire profile monitor!!
MEDM window
x32ch
Horizontal
Vertical
Inv AMP
This is the result of measurement, this is raw signal from SiC wire and this is measured proton beam profile.
As a results we got successflly proton beam profile with SiC wire profile monitor.
ADC
ADC count
EPICS IOC
Channel No
Channel No

17. Summary ・EPICS for 3NBT control system is under construction
・Data archive at 25Hz from CAMAC-CC/NET to Channel archiver on Linux WS is successful, but data transport among EPICS IOC has still data missing.
・ 25Hz DAQ data taking becomes better with changing the linux tick to 1000Hz.
・SiC wire proton beam profile monitor test was succeeded.
Summary
EPICS for 3NBT control system is under construction.
Data archive at 25Hz from CAMAC on CC/NET to Channel archiver on Linux WS is successful but data transport among EPICS IOCS has still data missing.
25Hz DAQ data taking becomes better with changing the linux tick to 1000 Hz
SiC wire proton beam profile monitor test was succeeded.

18. Future plan ・Improve the EPICS data missing problem.
・PLC (FA-M3) data I/O operation test.
・Design 3NBT OPI display.
・Design and test the DAQ and data archive system with full scale system.