1 Pulsar Timing for a Balloon-borne TPCan atmospheric muon David A. Smith, CEN de Bordeaux-Gradignan 14 avril 2017

2 Conclusions Recording accurate event times and instrument positions isn’t as easy as you may think. Many major missions goofed. Studying gamma polarization as a function of pulsar rotational phase requires forethought and testing.

3 This talk: Required time (and position) accuracyProblems on previous missions. Fermi LAT end-to-end test: bug found & fixed. A glimpse at GPS receivers.

4 The main gamma peak leads the main radio pulse by M. Lemoine-Goumard, M-H Grondin, CENBG. The main gamma peak leads the main radio pulse by d = * ± 12 ± 12 µs. /P0 = ms/33 ms = in phase. 0.1 < d < 0.3 for most young pulsars. (The little blip is the “classical” radio peak, corresponding to the magnetic axis.) radio cone g-ray fan beam * Erratum posted at

5 First known “black widow” MSP. B1937 ( = J1939+2134) : L. Guillemot, ex-CENBG, now Nançay. B1957 (= J ): First known “black widow” MSP. B1937 ( = J ) : First MSP discovered. Both remain among the fastest: P0 ≈1.6 ms. 0.014 * 1.6 ms = 22 µs peak width. Photon fluxes >100 MeV ≈ Crab/200.

6 Fermi LAT timing precisionNASA requirement: dti <10 µs absolute time accuracy per event. LAT collaboration goal: < 2 µs. We achieved <1 µs (and possibly <100 ns). 100 us would probably be fine for the TPC’s balloon flights… Position localization accuracy: Simply cdti (but in 3-D). Example: 3e8 * 1e-6 = 300 m.

7 Timing failures on 6 missions (1 of 2)USA (X-rays): The GPS often froze on orbit and had to be reset a few times a day. The satellite would go through GPS beams intense enough to confuse the receivers. Also, the speed of the satellite relative to GPS's was far from the design-regime for ground-based GPS's. XMM: Two years elapsed before absolute phases were reliable, after a series of 5 different kinds of electronics problems. Proc. SPIE 5165, (2004). INTEGRAL: Orbital inaccuracies due to ground software caused 300 us problems.  CHANDRA: For the HRC, the time stamp of a given event was that of the previous event. On-board filters remove events, so obtaining the right date for a given event was impossible. The solution is to trigger only on the central CCD chip, to reduce the event rate, to allow sending all events to the ground (“timing mode”). S. Murray et al, ApJ 568: (2002) and references therein.

8 Timing failures on 6 missions (2 of 2)Compton GRO: In the days before GPS. Events were assembled into packets on board, and the packets were grouped into a "major packet", to which a time stamp was afixed. These packets were sent to the ground. But the time stamp was from the preceding packet! And the time was off by over a second. ROSAT: Excerpt from : "A problem was…found…timing individual events, due to…software (Briel et al. 1994). The origin…was the spacecraft clock reset which followed the spacecraft tumbling incident of 1991 Jan. 25. All PSPC data after that time are affected. The problem leads to relative shift of 1s between adjacent PSPC events."  Never quite the same problem twice… GPS issues seem easily avoidable today, not the others… The above problems were either large (100’s and 1000’s of µs) or fatal.

9 I learned about the previous mission failures.I spent at Stanford working on LAT “I&T”*, mostly watching muons go through the calorimeter (=“CAL”). I realized there was no plan to test timing, and that the system is complex. I learned about the previous mission failures. I proposed a simple test. * “I&T” = Integration & Test.

10 GLAST LAT absolute timestamp end-to-end testBottom phototube + scintillator Top phototube + scintillator Track of a single atmospheric muon David Smith, Eric Grove, Denis Dumora, with invaluable support from Neil Johnson & Dave Thompson.

11 Serial communications One of the two redundant antennae.(Two GPS units on spacecraft, here at GD’s “Factory of the Future” in Arizona.) Serial communications One of the two redundant antennae. Power One antenna on each side of spacecraft, to see whole sky, i.e., as many GPS satellites at a time as possible. Receiver handles them both together. ‘Viceroy’ purchased from Motorola by GD (I was told…)

12 Two scintillator paddles.To standalone GPS Atmospheric muon track We went to Arizona. We put the scintillators from CAL tests next to GLAST. The triggerable GPS module from CAT & CELESTE interrupted the VME readout. Muons hit LAT and us within ns’s. Compared dates offline. LAT on spacecraft. Two scintillator paddles. We also tested the GBM dates.

13 Aerial view of LAT and m telescope.Extrapolate TKR tracks to scintillator heights. For small GPS vs LAT time differences, the paddles appear.

14 Something is wrong… 8 half-hour muon runs:4 for satellite side A, LAT config 1, and 4 for side/config B/2. 4 with GPS lock, to test <10 ms absolute time requirement 4 without GPS, to test <0.01 ms/s drift requirement. 0>dT>-1 ms sawtooth with ~290 s period during GPS lock runs. Need to add 1 second to Bordeaux times to match LAT Run ‘14191

15 GPS unlock runs : no wrap-aroundConfig/side 2/B Run ‘14218 Config/side 1/A

16 Prognosis GSFC project office very supportive. Telecons with GD engineers (GNC FSW, where GNC = Guidance and Navigation Control). Sawtooth with GPS lock: “subseconds” output (a 32-bit integer) from S/C GPS shows the 1ms sawtooth. If set to zero in S/C FSW then “should work”. Should not have been wired to input. Ramp without wrap-around when no GPS lock: UDL FSW averages PPS’s over preceding 100 seconds, to give good PPS when GPS fix lost. Lock problem thus propagated to un-lock mode. 1 second offset from UTC: Spacecraft epoch (“MET”) is set to ground PC NTP server time, not to GPS UTC. Estimate of time-to-set can be off by an integer step. Looking into a CCSDC compliant method change. S/C GNC had us repeat the measurements once all the fixes were made, which took months.

17 Verification of Absolute Time AccuracyEric Grove report to Collaboration, 7 months before launch. Verification of Absolute Time Accuracy SC PPS meets spec with and without GPS sync July/Aug and Oct 07 retest demonstrate that SC FSW bug is fixed With GPS sync, SC PPS is in phase with GPS PPS See upper panel Without GPS sync, SC PPS drift rate ~10x better than spec See lower panel Getting the integer seconds right… Our tests amply demonstrate that SC PPS will have correct subseconds Integer seconds are set by procedure at SC power-up Recall that SC time is seconds since reference epoch LAT, GD, and GPO are working together on power-up procedure 300 ns 

18 L. Guillemot thesis, pages 77-78.For the record: L. Guillemot thesis, pages The on-orbit calibration of the Fermi Large Area Telescope Abdo, A. A. et al. 2009, Astropart. Phys., 32, 193   (p.212) Beyond the scope of this talk (rotation ephemerides, software tests). Pulsar Timing for the Fermi Large Area Telescope Smith, D. A. et al. 2008, A&A, 492, 923

19 He says this unit works well. Eric G now flies gamma detectors in U2 “spy” planes to study thunderstorms. He says this unit works well. (U2’s have antennas on roof, and connectors inside.) Trimble is a historical leader in the field… Works above 40,000 feet. Not triggerable, but we don’t care.

20 For EUSO-SPB we are using differential GPS.” –Simon Bacholle.“The GPS for the first flight of EUSO-Balloon is the Motorola Oncore M12. For EUSO-SPB we are using differential GPS.” –Simon Bacholle. Oncore M12 ↓ Old! Circa 1999. Nota bene: some GPS’s disabled above ~40,000 feet.

21 Conclusions Recording accurate event times and instrument positions isn’t as easy as you may think. Many major missions goofed. Studying gamma polarization as a function of pulsar rotational phase requires forethought and testing. What, me worry?

22 an example of why to trust Bordeaux timesRF solutions LS-40EVALR1, 168 euros. Use PPS output to trigger VME GPS “time capture” 50ks run (overnight)  0.5 ms dispersion. 500 ns offset due to cable run of one antenna. Lost satellite fix during 4% of the run.