E2 Lock Losses Investigation
(Preliminary report - December 1, 2000)
D. Chin, R. Gustafson, K. Riles (Michigan)
W. Butler (Rochester)
Background information:
-
Coarse tidal correction actuation disabled during E2
-
Expect periodic lock loss as dynamic range of longitudinal actuation exceeded
-
Rate of lock loss determined by linear combination of common-mode and differential-mode
tidal slopes (steep slope --> short locked stretches)
-
Pitch compensation reduces effective dynamic range
(must push harder with upper coils than with lower)
-
Residual imbalance in mirror suspension requires significant DC offsets
in some coil currents.
-
Asymmetry leads to saturation sooner in one direction than the other
-
Data taken in recombination configuration (no recycling) and with each
arm individually
-
Actuation configuration:
-
L+ --> Equal push on ETMX and ETMY
-
L- --> Opposite push on ETMX and ETMY
-
l- --> Opposite push on ITMX and ITMY (and on ETMX
and ETMY!)
Observations:
In recombination locks lasted typically an hour with expected (qualitative)
correlation of length with tidal slope inferred from L+ control signal.
Lock durations especially short when common-mode tidal slope falling, suggesting
asymmetric sensitivity. [Figure: pdf
eps]
-
Y arm behavior [Figure: jpg (scanned
figure)]
Transmitted light through Y arm drifted linearly with time during each
lock, as if misalignment steadily worsening
Optical levers for both pitch and yaw also showed linear drift
Interpreted as mistuning of pitch compensation and yaw balance
ETMY output matrix retuned Saturday morning (R. Schofield).
Smaller drifts in both pitch/yaw seen subsequently
Despite these drifts, we don't believe the Y arm was responsible for
most lock losses
-
X arm behavior
Transmitted light through X arm relatively flat during most of lock,
but rapid and non-linear degradation seen in minutes preceding lock loss
Yaw of ETMX showed modest, linear drift
Pitch flatter, but same rapid pre-loss degradation seen as in transmitted
light
Two coils on ETMX with large DC offsets
Symptoms suggested "premature" saturation in one or more ETMX coils,
causing mirror twist
-
Coil currents
Examined long-term trends of ETMX and ETMY coil currents and
found anomalous behavior in upper left (UL) coil of ETMX:
-
"Software rails" seen in data acquisition ADC at expected +/- 2 V for all
ETMX and ETMY coils, but for periods preceding lock loss , evidence of
"hardware rail" seen in ETMX UL --> Odd, symmetric "exponential decay"
of min/max envelope, suggesting internal electronic saturation in UL circuit.
[4-day trend (pdf eps),
6-hour trend (pdf, eps)]
-
During one such "decay", checked monitor points on ETMX controller in X
mid stations with portable oscilloscope. Observed severe, asymmetric railing
at +2 V on UL coil, with moderate railing on UR and very little railing
on LL and LR. Eyeball estimate from scope traces indicated a strong 2-2.5
kHz component. During and after subsequent lock reacquistion, railing nearly
absent, but envelope of signals uncomfortably close to rails --> perhaps
a long-term noise concern.
-
Power spectrum [pdf, eps]
of ETMX UL coil current also peculiar: anomalous peak at about 30 Hz (aliasing
of higher frequency (>2 kHz) oscillation?)
Preliminary conclusions:
-
Necessity of coarse tidal actuation amply confirmed!
-
ETMX UL coil probable culprit in most lock losses
--> saturation
--> mirror twisting
--> degradation in one or more servo gains, with growth of higher
order modes
Open questions (still to be investigated):
-
Is premature UL saturation strictly result of initial DC offset?
-
Or does a faulty electronic circuit in the ETMX controller contribute too?
-
What is the exact mechanism by which gross misalignment leads to lock loss?
(which servo gives way first?) Calibration lines may be helpful here for
understanding gain loss.
Important additional task:
-
Create a complete list of E2 locked stretches longer than, say, 1 minute,
from replay of data.
(Our lock loss monitor wasn't up and running until Friday, and database
triggers weren't generated until Saturday.)
Comments on software tools:
-
Online tools in control room (data viewer and diagnostic test tool) extremely
valuable
-
But data viewer unavailable outside control room, even to other Hanford
computers
--> Hampers off-site analysis
Acknowledgements:
Aside from "official" lock loss investigators, many other physicists
contributed ideas to investigate and helped to carry out studies. A partial
list includes S. Whitcomb, D. Sigg, R. Schofield, R. Drever, D. Ugolini,
R. Flaminio, R. Adhikari.