Timing and RF Distribution NLC ILC Josef Frisch
Timing and RF Distribution NLC -> ILC Josef Frisch 1
History • RF phase and timing distribution concept developed for NLC • Prototype of phase stabilized long fiber links tested • Redundant high reliability design concept developed • Many design concepts transfer to ILC • Note: Presentation is a slightly updated version of the NLC system (without much reference of the current ILC timing / phase distribution designs). • Discussion to focus on Availability / Reliability. 2
Requirements (for purposes of discussion – NOT a specification) • RF phase distribution with stability to ~ 1 picosecond peak to peak. – The compressor has tighter requirements which may require a special system. • Trigger timing distribution with stability to a fraction of a cycle of L-band: 100 ps peak – peak (~30 ps RMS). – This is to allow resynchronization circuits to reliably select a single cycle of L-band. • Single point failure resistant. 3
4
Components: Fiducial Generator • The machine is assumed synchronized to a 5 (or possibly 10) Hz fiducial. The fiducial generator would be enabled under software control, synchronized to the 60 Hz power line, and to the 1. 3 GHz RF. • This is a single point failure, but there is only one unit in the accelerator, so the failure rate is expected to be low. 5
Components: Master Oscillator • This must be low noise 1. 3 Ghz oscillator. • A variety of technologies are available, for example Sapphire disciplined by Rubidium or GPS. • PSI specifies a sapphire based oscillator at – -120 d. Bc/Hz at 100 Hz – -152 d. Bc/Hz at 1 KHz – -160 d. Bc/Hz at 10 KHz and above. • This should meet ILC phase noise requirements. • The Master Oscillator is a single point failure, but there is only one unit, so failure rate is expected to be low. 6
Fiber Links • Point to Point links using standard telecom fiber. • 1550 nm laser diode source, modulated at RF – 357 MHz for NLC test, 1300 MHz OK for ILC. • Fiber spool in oven for fiber length compensation. 7
Phase shift for 10 degree C fiber change, 1 month (note 1 degree X-band = 250 fsec). 8
Components: Fiber Transmitter (1) • Use conventional Telecom laser diode at 1550 nm, directly modulated with RF. – NLC tests done at 357 MHz – Modern diodes OK for direct 1. 3 GHz modulation, and have (20 d. B) lower noise – Best to pulse diode so that reflected power measured with transmitter off. • • Note, must limit transmitter power to ~ 1 m. W, or get nonlinear effects in fiber which degrade performance. Fiber length compensation using ~5 Km spool of fiber in oven – Requires few X 100 Watts: – Continuously cool, heat with fan and wire grid. • Get ~10 second time delay from fiber. (with integration term). • Easy to close feedback loop • • Reflected phase measurement same as for receiver. Use downmix and digitizer system. Fiber transmitter is broad band, so fiducial can be applied as a bipolar phase shift to the RF. System cost low – all conventional components (except oven!). Requires ~6 rack units per transmitter. 9
10
Components – Fiber transmitter (2) • Ovens are “unpopular”: large and consume power • Alternate scheme: Use wavelength tunable fiber working against the fiber dispersion. – Need approximately 4 nm/C tuning range, with 0. 5 pm wavelength resolution. – Available commercially 100 nm range, without mode hops, . 02 pm resolution. • Cost ~$25 K. (From New Focus). • Expanded use of DWDM telecom systems may substantially reduce the price of tunable laser systems. • Scheme was briefly tested for NLC and worked, but at that time wide band, hop free tuning was not available. • Probably this is the technology of choice as the laser costs decrease. 11
12
Fiber Transmitter Reliability • Transmitters are redundant. Auto fail over is performed by the phase comparison unit. • Transmitters can detect broken fibers from reflected signal – In principal can automatically TDR fiber to help quickly find break or reflection. 13
Components: Fiber • Long haul fibers can use standard SMF-28 telecom fiber. • Need low reflections – want fusion splices, not connectors except at transmit and receive chassis. • Note that standard SMF-28 fiber is about as radiation sensitive as a human: a few hundred Rads can degrade its performance. – This varies dramatically with the exact fiber composition. • Need to test transmission system with real installed fiber. 14
Components: Fiber Receiver • • • Simple Converts optical to electrical signal Re-generates fiducial Error checking on optical signals Redundant, fail over in phase comparison unit. 15
16
Phase Comparison Unit • Located at the “crate” level – System design should allow accelerator operation with a failed “crate”. • Local narrow band Phase Locked Loop – Lock to either fiber system – Standby system has phase shifted to match active system – Prevents sudden global phase shifts (MPS issue) • Diagnostics to determine which fiber system is bad • Must be relatively low cost – high multiplicity item. 17
18
Phase Compare Unit – detecting failed channel • If channel signal level drops, or if fiducials are not detected. • Phase noise relative to narrow band VCO • If slow drift is detected, use head / tail monitors, or beam phase measurement to decide • Phase shifter on standby channel allows smooth changeover 19
Phase Control Unit: Low Noise VCO • Need good narrow band noise to allow phase memory between pulses • Need low cost since this is a high multiplicity unit. • Commercial multiplied VCXOs – Integrated phase noise few ps at 1 Hz – Slew rate limit PLL feedback for MPS to prevent sudden beam phase shifts – Can detect noise in fibers at frequencies above ~30 Hz 20
Head / Tail Monitor • Phase detection to compare neighboring sectors. • Used in conjunction with Phase Comparison units to detect failed fiber transmission systems. 21
Beam Phase Monitor • Can use Monopole HOM modes. – (have a hammer, everything looks like a nail!) • Data taken at TTF as part of HOM alignment / BPM experiments – Experiment was primarily looking at Dipole modes – Monopole modes were only used for testing system • Directly digitize cavity HOM signals with fast (5 Gs/s) scope • Look at phase of HOM Monopole modes relative to 1. 3 GHz phase reference • HOM modes are a good detector – high Q and mechanical stability (in helium) give accurate measurement. 22
23
HOM phase results / comments • • 1. 4 ps RMS for each mode. Mode difference 1 ps RMS -> mode noise of 700 fs RMS Can probably do much better with an optimized system. HOM couplers also see 1. 3 GHz signal in cavity. – Provides a direct comparison of 1. 3 GHz vs beam time. • Electronics specific to monopole modes would be low cost (standard down-mix / digitize system) • Effect of Lorentz detuning not known – could be a major problem for this type of measurement. – If so, can always use conventional phase cavities. 24
Triggers • Assume triggers derived from 1. 3 GHz countdowns, reset by Fiducial. – 1. 3 GHz too fast for present day programmable logic – limit few hundred MHz – Can work at a divided down frequency – Expect faster programmable devices by time ILC is constructed • Countdowns similar to SLAC PDUs (now running at 476 Mhz). • Due to need to reset frequency dividers running at 1. 3 GHz, need trigger stability <100 ps! 25
Issues / Conclusions • Base technologies for a redundant phase and timing distribution system for the ILC have been demonstrated – Compressor phase is the exception! Needs R+D. • Much engineering required to build a complete system • Various alternate technologies available – Example is fiber laser based phase / timing distribution system developed at MIT and DESY. – Need to do detailed engineering to evaluate trade-offs. 26
- Slides: 26