Although its rare, got the 500 to go nuts on an inflight power cycle.

The DG was spinning around, so my guess is its the fluxgate that's confused, not any of the mems sensors.. Removing cpu0 got my "aided" bit to go down, but the 500 keeps the roll going after I've leveld off from a roll. and I'm still getting "comm" dropouts.

The inflight "full navigator" reset still needs some work. Until it works better, the existing production code is best for clients because we can issue a "do not power cycle inflight" note so it never gets used.

Below is a graph where I power cycled at 3.9 but the GPS didn't lock for over 5 mins @ ~4.48 (veryfyed with CFS's map) then the 425 went into its "nuts" mode.

but its fun to see the little jig in the upper right were the chelton went from dead reconing mode to being GPS corrected.

The problem with the 500-222 is when it sees the "airspeed" packet from CFS 5.0C* the "aided" bit never goes out and I get "com" errors that also flip the "soft" bit that blanks the screen. If I remove the RX wire (airspeed packet) the com errors go away.

The 425 5beta1 code with:

  • Startup attitude estimation has been improved under vibration by incorporating a more sophisticated tilt estimator during the first few seconds of operation and retuning the initial statistical metrics that determine filter authority.
  • The AHRS500 "hard bit" is simulated during power-up. Rate bias estimation must fall within a specified limit before the bit will drop. This will effectively prevent power-up under most dynamic conditions.
  • "GPS soft drop" has been added, effectively letting the full navigation filter ride through GPS outages up to 3 sec. This should reduce attitude wonder after aggressive maneuvers combined with a brief loss of GPS (aileron roll, etc.)
  • GPS down->up transition performance has been improved. If GPS is down for longer than the "soft drop" limit, and is later reacquired under dynamics (slow roll, inverted flight, etc.), the full navigator must be reinitialized. The statistics that determine correction authority in this situation have been tuned to reduce attitude wonder once GPS is reacquired.
  • The World Magnetic Model was updated to reflect the 2005 coefficients just released. This will maintain consistency between GPS data and magnetometer data into the future.

    Did ok on a roll but the inflight power cycle is still unusable.

    Still disapointed in the 500. It doesn't come level after a roll

    but more importantly, it still blacks out into the redX for a second.

    this is what the bits look on a second flight:
    Timestatus
    0.000000 -31574
    0.000000 -31614
    0.000000 -31606
    1.585400 -31614
    1.626200 -31606
    3.271800 -31614
    3.314000 -31606
    4.944000 -31616
    4.985400 -31608
    6.624200 -31616
    6.666400 -31608
    7.646201 -31616
    60.383335 -32640
    ......
    93.102005 -32384Turn
    119.143211 -32640
    121.489166 -32384Turn
    970.255859 -32640
    971.921814 -32384Turn
    982.292358 -32376

    What's intresting with today testing at full 25+ samples/sec is plotting status bits at startup. It was error free as far as startup roll and pitch drift goes.
    Timestatus
    0.000000 -28096
    26.943001 21056
    66.116005 20544
    125.563507 20552
    126.473610 20544
    126.571342 20552
    127.466339 20544
    127.831207 20480
    128.263397 20992
    128.966003 21000
    129.182007 20488
    139.315002 4104

    but I ended up taking off with a bit status of 4104 (but I didn't know it at the time) and they never cleared prior to starting the inflight power cycle:

    Then the bits went from:
    Timestatus
    0.000000 -15552
    0.995192 -32192
    31.069601 16968
    39.385334 16960
    42.376003 16896
    42.776001 16384
    53.876335 0

    so in 54 seconds it thought it was perfect but didn't get close until 700sec. Then a Roll and IMC turn were done:
    Timestatus
    851.391052 -32704 Roll
    852.667053 -32768
    862.649048 0
    909.149536 -32704 IMC turn
    911.658630 -32768
    921.666382 0

    Then another Roll
    Timestatus
    1674.052490 -32704 Roll
    1674.795776 -32768
    1684.833984 0

    and landing.
    Timestatus
    2462.041016 512 landing
    2462.970947 0

    all on a calm day when I tryed to stay level after the upsets.

    There were some bumps decending to pattern alt. The 425 was removed and I will return to the address xbow forwards to me.

    What upset me more was after doing magalign on the 500-222, I saw the same takeoff and landing roll error as before, but now my dropouts(temp?) that happened only within the first 5mins, now accured for 143 seconds spread out across a 1800 second flight.. worse than it was 6weeks ago... sigh... I'm going to need another 500 or something, but I'll get the bits so you know what's wrong with it.. (The long span at 0 was a power cycle, the rest of the times the line is at 0 is when the screen has a red X on it :-(

    I've permently wired the 425's GPS to the IDU. dliccardo's theory that the GPS unlocked in the 3G IMC turn is lickly valid as several 2G turns didn't unlock the 425's GPS while it did on the gps18. Once level, there were no deviations from level flight displayed.


    Confrence call Thu Jan 6 13:11:26 EST 2005 identifyed these pris:
  • (hi) IMC turn. +-10 for 180sec
  • (hi) Inflight power cycle. +-30 for 450sec
  • (med) Wiggle mode. +-2 over 10sec
  • (med) cold start. +oo
  • (med) $GPRMC once a sec at 9600b
  • (low) airspeed aiding incase of loss of gps (CFS would lickly also be unlocked so its kinda pointless) {gps is a major part of "airspeed aiding"}
  • (low) jtag data loader

    I'm running model nav420ca-200 version 5020-0121-04_Beta2

    a 180 like someone would do if they entered IMC unexpectidly


    the "wiggle mode" with a 10sec cycle (no turbulence has a 10sec cycle) is clearly shown between 5.1 and 5.4 (smooth air, no Gs) but is visible in all the 425 graphs. Plus its a bummer it takes almost 3mins to recover from a 10sec turn.

    This startup error was the second flight of the day, so its less lickly a tempature issue, but perhaps a vibration or startup code issue... The DG was also spinning with huge error. In all cases a power cycle fixes it.

    Action Items:
    Rob: capture raw accel data on startup problem.
    Daren: soften the 10sec cycle more like a 500.
    Kirk: Permit Rob to post a "do not remove power from a 425 inflight" and include a letter in the box of all 425.

    The production 425 does a much better job at wind crab angles than the pre-production model. Its still "hunts" much more than the 500 in straight and level flight. compare times 1.35 to 1.56 with the 500 graph far below.

    After an in-flight power cycle at time 1.56, it drifts back into straight and level flight than then goes into the same "unusable" state as the pre-production 425.

    I've never had the want to power cycle a 500, but the 425 often needs a cycle on the ground before flight as it inits in a wildly strange angles.

    And the inflight "hunting" makes one wonder if they should reset the 425 inflight if they are looking closely. (don't!)

    Recommended Action: Issue some kinda safety of flight notice to never remove power from a 425 inflight.

    This flight was mainly spent on AP, with a stright GPS track.. but when there is a huge heading crab error for 700 seconds, like between 2.0 and 2.7, I did like what any pilot would do, cycle the power on the ahrs..

    ahrs was power cycled at ~2.73, ~2.83, and ~3.5

    and on my trip home I tryed two power cycles:


    that's what a 500 looks like with the same vibrations (Gs). note how much more the 425 "hunts" for stright and level even when its "working".

    Is there any way to make the ahars come online faster than 700 seconds?

    as always, the flight data is here.

    How did the 425 AHRS flight(s) go??

    Not well, but after looking at the data, most of the problems were gps18 related. Not only does it unlock:

    but it reported a position over a mile off after the second single 3g turn.

    After removing the gps18 errors, the only 425 issue was a large pitch and roll error (vibration?)

    while landing.

    Did Daren get you the pinout to wire the GPS to the IDU?

    No, there is lots more work to be done.. Like changing my mounting plate from phanalic to AL for vibration. The Fe cal is done progressively rather than a map table, (interesting) and programming is done by sending strings, rather than bits. I was able to confirm the gps lock (bit 6) in under 1min. Faster than the gps18 on cold starts. Much more work to be done, but the outcome of today is:

    Return all gps18s.... They are unsafe for flight.

    as always, the flight data is here.