Beam Losses and Machine Protection (real life) By Kay Wittenburg, Deutsches Elektronen Synchrotron DESY, Hamburg, Germany Experiences from HERA (accidental losses) Beam Dump: why?

Loss-Mechanisms 1)After an equipment failure (e.g. Power supply(ies) trip) the beam starts to oscillate (position or size) with an exponential growing amplitude. First losses occur after a time (length) which depends on the failure typ – then the beam “explodes” within a very short time. 2) Mislead beam. Might be very fast (< 1 turn). Reasons: Kicker, Operation, … t 0 X or Y Failure First losses Total Beam loss Aper- tur long short  t,  l

The Detector Two PIN-PDs in coincidence to count charged particles Signal (in Si): dE/dx = 3.7 MeV/cm 3.7 eV/e-hole pair => C/MIP => e - /MIP Small dimensions: Area: 2.75 · 2.75 mm 2 or 20 · 7.5 mm 2 Costs: 1 $ for small PD 100 $ for big PD BPW34 PIN Photodiodes to satisfy the special conditions in HERA

The Amplifier Sensitive and fast amplifier with low noise and with a fast coincidence following DESY BLM with lead hat (removed) on top of a sc quadrupole Diodes Pre-ampl. Video ampl. Comperator +5V +24 V Bias TTL driver -5V Threshold -5V Efficiency to charged particles: 30% TTL output for counting Very low noise: Dark count rate < 0.01 Hz max. count rate > 10.4 MHz Very high dynamic range: >10 9 Insensitive to synchrotron radiation: Efficiency to  : 3.5 · Coincidence + lead: 1 Hz at 1.5 Gy/h (e- ring at max.) Pulse shape of the BLM output Response to MIPs: Blue line: Single diode; Green line: Coincidence; TTL compatible (90  )

The counting module - Integration time: 5.2 ms (to be shorter than the cryogenic time constant of about 20 ms) - Short mode buffer: 128 · 5.2 ms = 666 ms - Long mode buffer: 128 · mean short = 85 s - Stop data taking in case of alarm - Archiving - Function check Counter im „Grab“; immer neben Controller

No quench Dump due to losses HERA BLM Alarm System

Alarmtest: Set threshold 1 or 2. Will be overwritten after some seconds! Should show 30 or 5

Note: A quench in HERA is not a disaster! It takes typ. 1-2 h to recover from cryogenic  = 189 Quenches More statistics HERA experience with

5 ms event, PS failure, HF failure 92 BLM Alarms

Old HERA Beam-Loss-Alarm- Topology ALIs Alarm loop-Zentrale Alarmloop DUMP BLMs + BPMs +Alarm-modules “Alarm-Loop- Interface” Internal Power- Supply- Alarm Galv. Trenn. HF failure input

Clean Dump due to HF alarm

start after 5 month shutdown

start after 5 month shutdown (Lumi upgrade) All by 5 ms PS failure events

What is a critical PS?

Alarm timing during failure of a critical magnet power supply t 0 X oder Y Power supply failure Aper- tur BLM- Alarm ACCT- Alarm Improved and faster internal Power-Supply- Alarm Magnet- current- Alarm Total- loss Too late No faster BLM Alarms due to spiky background!

ALIs Alarm loop-Zentrale Alarmloop DUMP BLMs + BPMs +Alarm-modules “Alarm-Loop- Interface” Beam-Loss-Alarm-Topology Internal Power- Supply- Alarm Galv. Trenn. faster Active New ACCT-Alarm DCCT-Alarm Faster clock rate Magnet current- Alarm More Failure inputs: PS, HF, …

Improvements Beam-Dump: before: 570  s after: 10  s Alarm at 0 Turn by turn current of bunch #1 DCCT beam current

BPM SL345 had wrong readings. => local Bump at one Quad. => < 4 BLM- alarms

Story (1): Statement: In HERA each cold Quad has a BPM. Instruction: Install a BLM close to each BPM to cover all cold Quads. DONE Events: Quenches of one Magnet in the middle of the arc during ramp. Observation: No Orbit distortion, no beam losses. ????? After a few days, some tries, some quenches: Observation 2: The correction coils in this area showed higher values Calculations: The correction coils drive a local closed bump. WHY THE BPM and BLM DIDN’T SHOW ANYTHING???? Observation 3: There is no BPM (because there is a cold-box. No BPM foreseen) Observation 4: Therefore there is no BLM (see above) Analysis: The automatic Orbit correction makes the local bump by accident. Consequence: Now we installed a BLM! => flexible system

Alarm Zentrale failure: Threshold went from 5 to 30

Story (2): Due to a wrong cabling, the alarms of 20 BLMs were subtracted and not added Story (3): Fieldbus-commands for other modules on the bus were interpreted by the ALZ

Injection: 10 bunches injected into first Dipoles

Collimators went too far into the beam.=> Losses in the magnets behind. (no quench but happened in earlier years. Very high Collimator BLM thresholds) protons

Operating (1): Wrong rampfile was chosen by operator.

Operating (2): Fast switch-on of magnets. Alarm loop (A1) was still disabled

Diverse: Hitting a cable during drilling (no quench)

Experiments PS-manipulation coasting beam Operating diverse diverses Operating dump? Remarks: What was first? Transient recorders most helpful. Here: p-beam was lost 8 ms before e-beam. Quench

Dump of 19% coasting beam is not a problem in HERA.

Some loss induced quenches were not documented in the Logbook?!?!?

< 4 BLMs Still 4 loss induced quenches in 2004:

The End now open for discussion

Solution 2: Proton and Positron ring in Lumi-Optic - BPM thresholds reduced to around 3 mm. Clean dump - no detectable current loss before dump triggered. Congratulations! Logbook comment to the test: “ In dem Quencharchiv steht die Schwelle in der Alarmloopzentrale auf 30 und es werden 30 anstehende Alarme angezeigt. Wenn das Setzen von 40 Monitoren noetig war, dann deutet dies darauf hin, dass in den Alarmkassetten nicht alle BPMs Scharf geschaltet sind (leider sind dies Jumper im Tunnel unter Beton). Naja, wenn 3/4scharf sind, geht das ja noch. “