The Initial LIGO Timing System Nearly Fatally Flawed

The Initial LIGO Timing System Nearly Fatally Flawed CDS Meeting, January 30, 2008 Daniel Sigg, LIGO Hanford Observatory G 080022 -00 -D CDS Meeting

Timing System Requirements q q The ability to reconstruct the arrival time of a gravitational wave signal with infinite signal-to-noise ratio shall be within 10 μs of UTC. The timing jitter on the converter clocks shall not degrade their performance. Timing glitches, clock errors, servo system hick-ups and malfunctions shall be recorded. All oscillator frequencies shall be locked to UTC. 2 N sampling clock → Impossible to buy off the shelf G 080022 -00 -D CDS Meeting 2

Version 1 GPS Brandywine 223 Fan-out Slave 211 -222 1 PPS 1 GPS master per (L)VEA q Timing distribution signals are differential ECL q Converter clock in most cases oversampled q Converter clock starts at 1 second boundary No true synchronization Start-up problems (every time you boot, timing jumps) Counting errors can go unnoticed for days, weeks, … q G 080022 -00 -D CDS Meeting 3

Version 2 GPS Brandywine Atomic Clock 223 Fan-out Slave 211 -222 1 PPS =? 107 1 PPS UTCG fiber TCT 1 atomic clock per site q 1 timing comparator per (L)VEA Doesn’t solve converter clocking problems q G 080022 -00 -D CDS Meeting 4

Transport Problem AS ADC LSC Computer Skew? 4 km fiber SUS Computer ETM DAC Overall delay? Recording delays? q q Skew counters Currently ~200 ms not accounted for in the calibration! G 080022 -00 -D CDS Meeting 5

Version 2½ GPS Brandywine Atomic Clock q 223 Fan-out Slave 211 -222 1 PPS =? 107 1 PPS UTCG fiber TCT Replace ECL lines with fiber G 080022 -00 -D CDS Meeting 6

Version 3 GPS & OCXO Atomic Clock fiber Fan-out VCXO fiber Slave VCXO 211 -222 =? 107 1 PPS UTCG fiber TCT 1 GPS/OCXO master per site (Symmetricom 58503 B) q Each fan-out and slave has its own VCXO clock q Return fiber indicates the state of the slave clock Timing errors and glitches are recognized and can be recorded q G 080022 -00 -D CDS Meeting 7

Version 3+ q Time stamp with Duo. Tone Ø 960 Hz and 961 Hz (can resolve 1 PPS) Ø Ramp, IRIG-B, pulse don’t work (harmonics) q Only AS port ADC, DACs not covered AWG DAC PEM ADC G 080022 -00 -D AS port ADC Duo. Tone Ramp Slave CDS Meeting 8

Fiber Protocol i. LIGO: q 223 Hz clock q Positive edges are used to lock VCXO in slave q Negative edges are used to encode 1 PPS a. LIGO: q Encode GPS time with negative edges q Encode status and error bits with negative edges G 080022 -00 -D CDS Meeting 9

Version 4. A (a. LIGO) q Fiber based timing system Ø Ø Ø q q New FPGA based All diagnostics sent back to master Master interfaces a “EPICS” computer Master integrates GPS receiver for time stamp Master controls OCXO by digital servo Atomic clock as independent check (reuse current) Dedicated slaves Ø Converter clocking Ø Duo. Tone generator & 1 PPS comparator Ø XO locking G 080022 -00 -D CDS Meeting 10

Version 4. B (a. LIGO) PCI-E q Digital IO board Full time stamps Ø Integrate timing and converter Ø Duo. Tone on all boards/read GPS time Converter Clocking DAC board fiber Duo. Tone ADC board G 080022 -00 -D Duo. Tone Generator CDS Meeting fiber 11

G 080022 -00 -D CDS Meeting 12

Timing Slave (6” x 3. 5”) G 080022 -00 -D CDS Meeting 13

Conclusion q Persistent problems: Ø Ø q q Timing jumps on reboot (problem resurfaced in fast ADC) Lack of timing diagnostics Lack of reliable time stamps Lack of qualification 200 ms unexplained calibration delay Photon calibrator analysis only now started Full time stamps and reliable diagnostics are necessary for the next generation timing system G 080022 -00 -D CDS Meeting 14