11 Feb. 2002

Report on Mt. Hopkins activity, 27 Jan.-9 Feb., 2002


	1.)  We refined the alignment of components along the relay path
to Telescope C, as described in an earlier partial report.  The feed
mirror location for the corner position was changed by a few mm to
accommodate a slight error in the location of the port in the vacuum
tank.  The dihedral on the LD2 carriage was shimmed up by 0.083 inch to
match the corresponding lift of the SD2 dihedral, resulting from
adjustments to the SD2 carriage.

	N.B.-- It appears that the dihedral on the SD1 carriage is also
riding a little higher than nominal, such that we ought to shim up the
LD1 dihedral as well, by perhaps 0.030 inch.  We did not get to this, and
the result is that when you look outward for alignment, the beam runs a
bit low (perhaps 0.060") at the corner.  The corner mirror may be tilted
to center the beam on the feed mirror, using the LEDs installed there,
which make this easy.  We also used these LEDs as the target for aligning
the yaw after LD1 motions.

	2.)  An innocent question about running the LD carriages on open
loop (i.e., without the laser-metrology control; just relying on
posotioning by the stepper-motor drive) led to a large amount of work by
Pete Schloerb, Mike Brewer, Angela, and Marc, which eventually put
everything in good shape for both open-loop and closed-loop operation.
The conditions "open-loop" and "closed-loop" are designated in the
control software as "absolute" and "servo" operation, by the way.  In
closed-loop (servo) operation (the normal mode), the closure is achieved
by giving the stepper motor the nominal number of pulses to get to the
desired location, then checking the laser to see where the carriage
actually got to, and iterating with smaller and smaller commands until
the error is tolerable.

	3.)  We set up a simple Michelson interferometer on the
eight-foot optical table, using (but not disturbing) some of the
components feeding the integrated-optics beam combiner.  This is
described in a previous report.  It is quite easy to set up and align,
and thus can give fringes to play with on cloudy nights.  Note, however,
that it feeds a modulated signal into only one of the three IO input
fibers, and uses the IO combiner only as a means of getting the signal to
the detector.  The signal enters on the fiber that would ordinarily
receive the beam from the "Delay1" path, and comes out, in about equal
intensities, on the four outputs that would ordinarily show the two
complementary fringe signals from (Delay1 + Fixed) and (Delay2 + Delay1).
We display real sky fringes by subtracting the two complementary outputs
for each of the three pairs.  For the laboratory fringes, this does not
work, since there is no complementarity; you can only look at one of the
four outputs that show the fringes.

	We tested Ettore's fringe-recognition software on the laboratory
fringes, and it worked well on these.  Using a scan of less than the full
60-micron capability of the piezo scanner, the software can move this
scan around so as to keep the fringes in range, tracking the piston
displacements imposed by the atmosphere (or, in our case, in the
laboratory).  In this way we can maintain the efficiency of a short scan.  
If there is a secular drift, as may occur if the baseline vector is not
known with sufficient accuracy, a correction for this can be fed back to
the short delay line.

	4.)  We noted that there was a confusion in the display of data.
To discuss this, let us summarize a number of things about the data
recording and display:

	First, the six outputs of the integrated-optics system are
recorded on six pixels in a given column of one quadrant of the PICNIC
detector, with four unused pixels between the active ones.  Careful
adjustment of the focus and x-y position of the lens focusing the IO
output onto the detector gives a good concentration of the signals on
individual pixels - of the order of 80% or better.

	The labeling of each pixel according to what combination it is
actually detecting is done by illuminating successively each input fiber
(identified as Fixed, Delay1, and Delay2), and seeing which pixels light
up.  The result of this exercise can be shown as:

  Pixel                    Beam Designations

    1	    Fixed (less)    Delay1 (more)

    2       Fixed (more)    Delay1 (less)

    3       Fixed (more)                        Delay2 (less)

    4       Fixed (less)                        Delay2 (more)

    5                        Delay1 (less)      Delay2 (more)

    6                        Delay1 (more)      Delay2 (less)

We put in the labels "more" and "less" because there seems to be a
systematic distribution of intensity in the four outputs for a given
input, with the "more" components being perhaps 20% brighter than the
"less" ones.  For each beam, however, if we sum the "more" and "less"
components (say, for the F input, we sum the ones that are to be combined
with D1, and those that are to be compared with D2, and do the
corresponding thing for the other beams), we find that the sums appear to
differ by quite a bit less than 20%.  This should tell us something about
the way in which the components of the integrated optics device perform
their functions.  Each of the inputs is first separated into two parts by
a Y divider, and these dividers apparently all make close to a 50-50
split.  The combiners, however, appear to deliver somewhat unequal parts
of the two signals coming to them into their two complementary outputs,
differing by the observed 20%.  [This 20% should not be taken too
seriously; we did notice a systematic difference, but the number only
comes from glancing at an analog display.]

	The signals from the six pixels can be monitored by displays of
three types, two for setting up, and a final one that shows for each
pixel signal-vs-time, which is to say, signal-vs-optical path difference,
as generated by the piezo-driven path-length scanners.

	The first display shows the whole quadrant that we use, and the
illuminated pixels appear in a horizontal row (although they are
physically in a vertical column).  In this and the next, the strength of
the signal from a pixel is shown on a false-color scale.

	Second, to zoom in on the six pixels, one must click on the end
pixel of those that are illuminated.  If the pixels do not show up with
good signal-to-noise, there is a possibility of an error here, whereby
one can subsequently read out the wrong sequence of pixels, for instance,
missing a good one at one end of the sequence and reading an
unilluminated one at the other end.  Assuming that we zoom in correctly,
the next display shows the column of pixels, oriented vertically, with
the pixel on the right in the previous display being at the top of the
new one.  The signals illuminating each pixel are as described above.

	The display that we monitor during observation shows for each
pixel successive displays of the 256 samples taken during path-length
scans.  The displays are ordered on the screen in the same fashion as the
pixels in the previous display, top to bottom.  Each display is labeled
according to the beam pair being detected on the pixel.  Successive pairs
of pixels show the complementary outputs from a given pair-combiner, and
are labeled as:

	Fixed  - Delay1
	Fixed  + Delay1
	Delay2 - Fixed
	Delay2 + Fixed
	Delay1 - Delay2
	Delay1 + Delay2

The + and - are used arbitrarily, to note the opposite phase of the
fringes in the elements of a pair.  In addition to the six primary
displays we can view the combinations of pairs in the form (a-b)/(a+b),
so that the fringe signal is enhanced, and common-mode signal
fluctuations are rejected.  The power spectra of these combined signals
are also shown.  [N.B. - As of this date, the labeling and placement of
these scan displays may still have some confusion, and should be double
checked.]

	The computer cannot show every scan; something like every third
or fourth scan is shown, in general.  This means that if you search in
path-delay to find the fringes at the zero point, the search should be
slow enough so that the fringes will appear in several successive scans,
and not turn up in only one or two, which by bad luck might be omitted.
Fringe-recognition software, if operating, should actually examine every
scan, and stop the search when it finds fringes.

	Another area of confusion exists in designating the numbers of
loops and reads involved in reading out the pixels.  A loop is the cycle
of starting at the address of the corner of the quadrant, clocking over
to the active column and down to the first active pixel, and reading the
pixels sequentially.  Once at a pixel address, one may read it more than
once, and gain efficiency thereby, since the first read is affected by
settling.  Each read operation takes on the order of 22 microsec;  the
total time spent in clocking is around 300 microsec.  The minimum time
around a loop, with one read only, is therefore around 370 microsec, and
the minimum time to generate a 256-sample scan is just under 100 msec.
The integration time on a pixel is determined by the number of loops one
takes and the number of reads in each loop, to generate one sample. 
Better signal-to-noise is achieved with more reads, at the expense of a
greater lack of simultaneity in recording the six signals.  More loops
just gives more integration time, with no change in differential time
delays.
	The confusion here at present is that the numbers of reads and
loops actually executed does not seem to correspond exactly to the number
requested.  The number seems to be one less than requested, but with a
request for one giving one, and not zero.  This may not be exactly what
is going on; the main thing is to fix and double-check it.

	5.)  Coming finally to the main business of detecting fringes: As
previously reported, we found fringes on the N-S baseline (telescopes A
and B at the 15-m stations on the north and south arms, respectively) on
two stars, iota Aur and alpha Lyn, running the short delay line, SD1,
with an offset of 2.83-2.85 cm.  This was close to the predicted value of
2,7 cm, derived from previous measurements of 2.0 cm, with a correction
of 0.7 cm resulting from the exchange of the SD1 and SD2 dihedrals.
During these successful searches, we saw occasional bursts of noise in
excess of what the atmosphere seemed to be giving us, and these seemed
not to be white noise, but rather concentrated in the range of several
hundred Hz (i.e., in the higher range of frequencies that we could detect
in our scans of 256 points in something over 100 msec).  These bursts
were not crippling, but when we brought the second carriage, SD2, away
from home (where it had previously been), in order to search for fringes
btween telescopes C and B, we saw much greater bursts of noise, including
large spikes as well as enhancement in the range near 1 kHz.  This noise
was too large to tolerate, and after some dithering around, we set out to
find correlations of the noise occurrence with other parameters.  As
reported earlier, we found that the noise was a periodic function of the
position of SD2, with a period of about 6 cm.  Noise occurred, e.g., when
the carriage was in the intervals 101-100 cm, 95-94 cm, 89-88 cm, etc,
(followed down to 30 cm or so).  The noise happened during slew,
tracking, and even (though possibly less) while standing still in those
ranges.  After opening up the vacuum tank, we found that the only
obvious structural periodicity along the track, near 6 cm, was the
spacing between pairs of permanent magnets on the motor track, which
was measured to be about 6.08 cm.  The spacing between the fasteners on 
the side of the granite, that caused bulges in the top surface, is about 
13.0 cm.  Another spacing of interest is that between the home positions
of SD1 and SD2, which we measured to be about 48.6 cm.  Knowing this, we
should see whether the lesser bursts of noise observed during SD1
operation occur at the same track locations.  We did not get around to
doing this.
	We found that the noise occurrence during SD2 operation was well
correlated with signals that we could observe, connected with its motor.
We monitored one phase of the motor current and also the power-supply
voltage at the point where it was feeding the SD2 motor-driver (Eltrol)
card.  At both of these locations we could see high-frequency signals (in
the range ~1MHz to ~12MHz) whose amplitude correlated well with the
camera-noise occurrence.  The correlation with the 12MHz component seemed
the strongest, with the greatest interference coming when this drove the
high-frequency ripples on the power supply up to 3v or so, riding on 28v.
	After observing these correlations, we turned to some study of
the 1kHz signals.  We did not, as I remember, notice variations in this
range in the motor signals, but this warrants further exploration.  We
had noticed, however, that some component in this range was present in
the camera output even when the SD carriage drives were turned off (and
indeed, on further investigation, when practically everything in the room
was turned off).  This residual component was at a low level (perhaps
20-30 adu rms), but still sufficient to dominate the camera noise under
dark conditions, giving the noise a distinctly non-random appearance.
The main thing that we then found was that this component essentially
disappeared when we isolated the whole camera system, dewar and all, from
the optical table.  We did not get back, while I was present, to seeing
how the isolated system responded to the motor-correlated noise signals.

	N.B.- In all investigations involving the camera signals,
displayed as 256-point scans, note that the displays are auto-scaling, so
you can be deceived as to what is going on, if you do not keep an eye on
the scale changes.  We probably want the capability of fixing the scale
of the displays for diagnostic purposes.