Cable Fire at the LER TFB Kicker

1
Cable Fire at the LER TFB Kicker

• Stan Ecklund for investigation committee and
  other helpers

2
Investigation Committee

• Ron Akre
• Stan Ecklund (Chair)
• Keith Jobe
• Joseph Olszewski
• Bob Reek
• Ponciano Rodriguez
• Mike Stanek
• Uli Wienands
• Andrew Young

3
Charge

• 1. Research the sequence of events leading to the
  fire. Determine the cause of the fire, if
  possible, or determine probable causes.
• 2. Identify damage and recommend a course of
  action to repair or replace damaged equipment.
• 3. Based on determined cause, recommend changes
  to this and similar systems that will prevent
  future incidents.
• 4. Coordinate with the ORPS investigation (do not
  duplicate)

4
Location in PEP-II Region 4
5
LER Horizontal Kicker 3040
6
Transverse Kicker Pickup Structure
This picture shows the kicker as a pickup. The
picture below is the time domain response of a
by4 fill pattern. If it was a by2 fill pattern,
the signal in the red circle would move to the
blue circle thus, the leading edge of the next
bunch cancels the trail edge of the previous
bunch.
The round trip time(4.2ns) 2L/c where L 63cm
7
Kicker

• 1.755 radius (44.577 mm)
• 120 deg.
• 24.62 long

8
Photo by David Kharakh on 11 Oct. 2005
9
Photo by Mike Zurawel on Aug. 17 2006
10
Wall Side
11
Breach in Cable
12
Melted center conductor
13
Cable Trays
14
Investigation Activities

• Observe and document situation
• Many Photos at
• V\AD\AreaManagers\MZurawel\LER X Kicker 8-17
• Roped off Site to keep undisturbed during
  investigation
• Calculated beam power to cables (Steve Smith,
  Andrew Young, Anatoly Krasnykh, John Fox)
• Tested another fan, stalled
• Looked at cable/tray damage
• Mined history data

15
Event Sequence (Uli Wienands)

• 2000 mA, 861 bunches, by-4 (034424)
• 1742 arc T3040Ki2 sees temp. rise
• 1746 arc T3040K14, K15 see temp. rise
• 1748 elog LER Abort with LR42 Latched
• 1750 elog Fire Alarm 3114 IR-4 Building
• Starts in tunnel, then B645 (rf), then B640
  (dnstairs)(no time line, just recollection)
• 175132 errlog PR03 174 trips on gnd
  fault175152 errlog PR08 5162 trips mag. Intlk
  2
• 175312 errlog P2HZDSOFF started

16
Current and Temps
Normal Run
At time of fire
HOM
T3040KI2 X- Load Connector
LER Current
T3040K14 Body
T3040K15 Bellows
17
TFB Drive Connections

• Note A Damage found in the output section of the
  amplifier, the output transistor not connected to
  circuit
• Note B Cable center pin was found not mating
  contact with filter. This causes the beam
  reflected signal to see an open at the amplifier
• Note F Fire damage. X- drive cable damage is
  present up to 5 m from the kicker

18
TFB Load Connections

• Note C A cable breach was discovered, the center
  conductor copper clad was melted and showed signs
  of arching on the opposite side of the conductor.
• Note D The short pigtail that went from the long
  haul cable to the attenuator showed signs of
  melting and had an impedence of 25 ohms at the
  min.
• Note E The beam was misaligned by 1 cm to the x
  side
• Note G x load cable damage is present up to 5 m
  from the kicker

19
Power to Cable

• Strip-line kicker, like directional coupler
• Upstream connector has beam induced signal
• No bunch-to-bunch cancellation in by 4 bucket
  spacing.
• Power calculated for centered beam (S. Smith, A.
  Young)
• 774 to 1200 Watts to cable
• Loss 4.7 W/ft
• Observed 1 cm beam offset doubles power
• Loss 9.4 W/ft
• Insulation Melts at 7 W/ft
• LDF4 (1/2 in heliax) spec is 5 W/ft
• Conclude cable should have failed as it did.

20
Observations and Conclusions

• Hottest part of fire inside ring, near fan
• Two Candidates for Ignition source.
• Fan due to malfunction
• RF cable from kicker
• Tested identical fan with stall did not ignite,
  circuit opened
• Autopsy of fan consistent with it being victim,
  not cause
• Calculation of Beam induced Power into load cable
  exceeded cable rating
• Beam in a by 4 pattern. No Cancellation as in by
  2 pattern.
• Beam offset by 1 cm. Doubles power from centered
  beam.
• Bad Cable connection at x- drive could also
  increased reflected power.
• Cable to load measured low Z25 would also
  reflect power.
• Conclusion
• Cause of ignition is exceeding power capacity of
  Load cable
• Fan provided fuel, accelerating fire.

21
Repairs

• Clean Area done
• Build platform for cable repairs done
• Cable repairs -
• EWP, LOTO - done
• Start DC cables
• Complete DC cables
• RF cables
• Checkout

22
Corrective Actions

• Additional Measures
• Higher power rating on RF Load cables (7/8 inch
  vs. ½ inch)
• Replace RF cables with those with Fire retardant
  jacket.
• Add monitoring (Machine Protection) to TFB
  systems
• Thermocouple on cable
• Forward and reverse power couplers/diodes
• Operational Limit on current, specific to pattern
  (by 4)
• Understand and correct orbit offset

23
EndExtra Slides
24
Ref orbit During MD 17 Aug 2006 1343
25
Orbit Consistently off x -9 mm
26
XCOR making bump
27
Field at wall for 1 cm offset of beam
28
Power to x electrode doubles at -1 cm offset
29
PEP-IITransverse KickerBeam Power Estimate

• Steve Smith
• Sept. 19, 2006

30
Procedure

• Foam Coax Rated for 5 Watt/ft loss
• Independent of size
• Size affects loss, not max loss
• Estimate beam power from idealized stripline
  kicker model
• Compare prediction to observed waveforms

31
The Andrew's Cable average power handling specs.
for this cable are 1.91KW _at_ 300MHz and 532W
_at_3GHz. Using these numbers the limiting power
density of the cable 5W/ft with inner
conductor of 100oC. The melting temperature of
the foam dielectric is 120 -130oC
32
Andrew LDF4 Cable
Estimate Loss Constant 2.2 dB/100/?GHz Limiti
ng power loss per length 5.3 W/ft
33
Andrew FSJ1 (1/4) Cable
Estimate Loss Constant 6 dB/100/?GHz Limiting
power loss per length 5.1 W/ft
34
Kicker Model

• Parameters
• Length
• Width
• Impedance
• Shunt capacitance
• Beam charge
• (Bunch length)

35
Beam Spectrum

• For zero bunch length, the beam current spectrum
  is a series of delta functions at harmonics of
  the bunch spacing frequency.
• For the by-4 fill pattern the frequencies are
• The current in in each line is given by
• For finite bunch length the spectrum gets rolled
  off at high frequencies.
• Assuming Gaussian bunch shape with st 38 ps
• The beam-induced power is down by about 3 dB at
  sf.

36
Stripline Kicker

• Assuming
• weak coupling between striplines and
• matched impedances at both ends
• (Modify later for shunting capacitance)
• impulse response
• two impulses of opposite polarity separated by
• Impulse
• Frequency response

37
Paddle Capacitance

• Paddle area A 2.8 in2
• Paddle gap d 0.25 in
• Paddle capacitance C eA/d 2.6 pF
• Fc 2.25 GHz

38
Power Disspation in LDF4

• Assumes
• Weakly-coupled striplines of Z50 Ohms
• underestimates power
• Odd mode is 50 Ohm
• Even mode is 64 Ohm??
• Directionality
• Divide by 2 for even split
• No waveguide modes
• Reduces power
• No HOM power from other sources
• Increases power
• Centered beam

39
Power Disspation in FSJ1

• Assumes
• Weakly-coupled striplines of Z50 Ohms
• underestimates power
• Odd mode is 50 Ohm
• Even mode is 64 Ohm??
• Directionality
• Divide by 2 for even split
• No waveguide modes
• Reduces power
• No HOM power from other sources
• Increases power
• Centered beam

40
Predicted vs Measured Beam signals
41
Tentative Conclusions

• Observed bandwidth about half of calculated
• Limitations of model
• Directionality vs frequency
• Calculate marginally enough power loss in LDF4 to
  melt dielectric
• Only requires 7 W/ft loss
• Losses in FSJ1 much higher!
• FSJ1 failure may have increased losses in LDF4
• Did not include position sensitivity